It’s still early, but I think AI software devtools are increasingly coalescing around what I call SPACE Development. That is:

• Search

• Plan

• Assert

• Code

• Evaluate

Folks may use different terminology, and may drop the assert step (to their detriment), but this basic core has come up over and over again. First you do research to learn more about the problem and pull in context. Then you write up a plan that explains how you aim to tackle the problem. Then you define a set of success criteria. Then you write your code. And finally you evaluate if your code met the success criteria.

SPACE Development has become best practice for building software with agents. I’m not the first to notice. Here’s Shaiyan Rais, who owns the claude-code-best-practices repo and who basically already coalesced the relevant research that I was planning to do:

All major workflows converge on the same architectural pattern: Research → Plan → Execute → Review

Shaiyan drops ‘assert’, but it’s there in all of the frameworks that use TDD, including Superpowers, Everything Claude Code, Matt Pocock Skills, etc.2

But notice that this is not really a strategy for writing software specifically. It’s a guide for how to think about certain kinds of problems. You can take the same basic SPACE pattern and apply it to everything from building a web game to creating a language compiler. And if you are flexible with the ‘code’ step — replace it with something more generic, like Execute3 — SPACE development becomes a general purpose hammer for a huge range of tasks.

The only way to land on something like SPACE Development is through metacognition — essentially, thinking about thinking. Every time you look at a process and think ‘how can I do this better?’ or ‘how can I formalize this?’ you’re doing some kind of metacognition. For example, throwing a basketball is just basic cognition. You’re just doing a thing. Thinking about how to flick your wrist to get backspin on the ball every time you shoot? That’s metacognition. You’re abstracting out from a particular instance of a thing — in this case, shooting hoops — to instead reason about the underlying process.

If agents are bad at thinking, they are totally incapable of thinking about thinking. So you need to do the metacognition for them. That means thinking about how to think through different kinds of problems, and then giving the agents the steps you land on. To poorly paraphrase, teach an agent to solve a single problem, and you’ll get a lot of slop to review; teach an agent to think, and you’re set for life and/or you’ve automated your job away (the analogy got away from me).

Mechanically, what this ends up looking like is a very particular style of AGENTS.md and SKILL.md files. Something like this:

<required>
1. Add all of the following steps to your TODO list.
2. Research how to best solve my question WITHOUT making code changes.
  - Search for relevant skills using Glob/Grep in `{{skills_dir}}/`
  - Use the WebFetch tool to search in parallel
3. Read and follow the writing-plans skill. Present plan to me and ask for feedback.
4. Read and follow the tdd skill.
5. Update docs.
6. Push a PR with your changes.
</required>

The goal of these configs is not to provide context. Context is data, it lives on disk to be queried when needed. Rather, these configs describe processes that tell the agent how to make use of the context it has available. In my mind, there is a world of difference between the Superpowers brainstorming skill and the Anthropic PDF skill, even though they are both implemented as skill markdown files. The former is a thought process and the latter is just information.

Are there other kinds of thinking processes useful for software? Debugging comes to mind:

• Add logs.

• Replicate bug.

• Read logs.

• Change code.

• Repeat.

Which is basically just a version of the scientific method (form hypothesis, test hypothesis, adjust hypothesis based on results).

Sometimes you get processes inside processes. For example, if you’re trying to build something really big, you might have SPACE Development as a subroutine.

• Read large specification.

• Identify small, actionable piece.

• SPACE development on that small piece.

• Update specification.

• Repeat.

I could probably rattle off a dozen of these sorts of thinking patterns. I suspect the specific formulation doesn’t matter all that much. The larger point is that if you’re using an agent out of the box, it won’t have the awareness to do this sort of thing at all!

If you’re not the kind of person to invest a ton in configuration, pick up any of the open source LLM configs that encode SPACE development. We mentioned Superpowers from Jesse Vincent a few times here — it is really the OG LLM config set up in the space and is a great starting point. I’ve written about my own personal setup quite a bit, and have it hosted over on the nori skillsets registry.

What I wouldn’t do is try to build out a complicated mechanism for enforcing any explicit line of thinking. I have seen too many engineers fall into the tarpit of building complicated agent harnesses with complicated handoff scripts and, like, just don’t do it. See:

Even within the base providers, I’m very against “plan mode” as a concept. I’ve tried it and really don’t like it. I think that sort of forcing-the-model-into-a-box defeats the purpose of using a fuzzy general purpose machine.

Instead, teach the model how to think. It’ll take you ten minutes and a few dozen lines of markdown, and ends up being the equivalent of getting access to GPT-N+1 months before everyone else.

1

I didn’t realize when I was typing this out, but I was subconsciously channeling this meme from like 10 years ago

I’ve been hiccuping for nearly 70 hours. We are well past the stage of ‘huh this is funny’ and well into the stage of ‘what the FUCK is going on?!’ It’s reached the point where, when I call medical professionals, they do what I can only imagine is a spit take. “Wait, you’ve been hiccuping for how long?!” I can’t live like this, I have events I need to go to! I can’t like, show up to a speaker series or a client demo or a candidate interview and just be hiccupping the whole time! The only upshot is that my abs are getting the best workout they have had since I was on the highschool jv waterpolo team (go big blue).

Hiccupping is one of those weird things where we kinda sorta don’t really know what causes them. Like, mechanically, we get it. It’s a spasm of your diaphragm. Why does the spasm happen? Well there are these nerves that exist that get irritated sometimes and mumblemumble and then you have hiccups. And then when you point out the doctor just said ‘mumblemumble’ they get annoyed at you.

Hiccupping is also one of those weird things where we don’t really have any obvious targeted medications for it. I mean, there are approximately a million different home remedies. Swallow some air, sip water slowly, sip water quickly, sip cold water, sip warm water, eat a spoonful of a peanut butter (thanks Raymond!), each a spoonful of crushed ice, eat a spoonful of granulated sugar, hold your breath for 5 seconds, hold your breath for 10 seconds, hold your breath for 20 seconds and then sip some air (temperature unspecified), pull your legs into your chest, do the bridge yoga pose, do jumping jacks, and, of course, drink water upside down.

Perhaps surprisingly, none of these are respected by the medical establishment. If you show up at the doctor’s office hiccupping for 70 hours, they will not have any obvious treatments for you, and they may not have any great pointers for a diagnosis either.

II.

I’m not a doctor, but my Mom always wanted me to be one, so let’s just imagine I was. If I was a doctor, and I had a patient with some weird symptoms that I didn’t really understand, my first reaction would be ‘huh, interesting’ and my second reaction would be ‘we need to collect more data.’ And I would then go and order every test under the sun to try and figure out what was going wrong. Claude says that a really comprehensive blood panel costs somewhere between $25-50 to do — that’s the cost of the reagents and the labor and the blood draw supplies and so on. Let’s throw on a really comprehensive stool test too, which would add another $120-150. So the total is somewhere between $150-200 to get a ton of data about the patient.1 This seems obviously useful.

But I am clearly not a doctor, because no doctor seems to agree with me. Every medical professional that I have ever seen — including the two doctors and two nurses I saw at urgent care, and the doctor and nurse I saw at Mt. Sinai — has basically refused to do this kind of data gathering. Instead, they will take a stab at some cause based entirely on the medical history and a few words with the patient, propose some medication that treats that cause, and then refuse to budge at all from that initial guess.

For my hiccups thing, that initial guess was gerd. All three doctors heard that I had eaten Indian food and went “yup, you have acid reflux.” Nevermind that it has been three days since I last had Indian food, and that in that time I had only eaten bread and rice, and that I had already taken several doses of acid suppressants by the time the third day rolled around, and the fact that I am ethnically Indian and eat Indian food all the time.

The third time a doc told me that a) I had to keep taking the omeprazole and pepsid (acid suppressants) that I was already on and b) they weren’t going to give me anything else, I nearly lost it.2 I had to basically force the doc to put in an order for a blood panel and an H. Pylori test.3 And I forgot to also push for a stool test, which I’m still banging my head over.

When I first mentioned that I should get tested for H. Pylori, the doctor saw that I had taken a single dose of a proton pump inhibitor (PPI) and went “nope, sorry, you can’t take that test until you’ve completed a 4 week PPI course and then waited another 2 weeks for the PPI effects to wear off.” This is because taking a PPI can increase the likelihood of false negatives. I had to patiently explain that even if the medication did increase the false negative likelihood, we would still get a lot of useful information if the test came up positive, and that starting treatment 6 weeks earlier would be incredibly helpful. The doc was clearly unhappy with this line of reasoning, and pushed back at least three times before acquiescing.

I did eventually get the test order in, but then when I went downstairs to labcorp the guy there saw that I was on a PPI, saw that I was trying to get an hpy test, and decided that I wasn’t supposed to take it. He first tried calling the doctor, and then made me go back upstairs to get a signature from the same doctor that I’d already done three rounds with. Which I did. And then while I was waiting, a different doctor from the same practice called the labcorp guy back, both of them agreed that I shouldn’t take the hpy test, and it wasn’t until I got on the phone with the second doctor and explained the whole false positive thing that these people finally agreed to give me this test. Which, by the way, just involves blowing into two balloons and drinking some lemonade.

I got to the doctor’s at 9AM and left at noon. A quarter of my day, gone. I’ll get the results back in a few days, hopefully.

III.

If you read the title, this is obviously not a story about hiccups. This is an article about the ridiculousness of dealing with the US medical system while suffering from some random ailment that no one understands.

I’ve written about my dealings with the US healthcare system before. That post is one of my favorite articles, and my only regret is that I did not write it when I had a bigger audience — I had 54 subscribers when I first posted it. You should go read it:

In that post, I said "healthcare in the US is an exercise in indignities.”

It’s been ~two years since I wrote that post, and nothing has changed.

It’s easy to look at the story above and see a microcosm of everything wrong with how US healthcare works. There’s a fragmented healthcare system where most of the medical professionals did not see relevant medical history information like the previous H. Pylori diagnosis or the fact that I have EOE. There’s the difficulty in getting an appointment — I only ended up being seen by an actual GI because the lady on the phone took pity on me for not being able to get through two words without hiccupping.4 There’s a ridiculously expensive markup on everything. I paid ~$200 to see my providers and get the meds they prescribed, and I’m on pretty good health insurance.5

But in my opinion, the worst thing is that it is impossible to get the average doctor to act like a scientist instead of a bureaucrat.

Bluntly, I have a very hard time rationalizing why a doctor would hesitate to order tests.

IV.

But lets try anyway. A few possibilities, roughly ordered by most to least generous:

• The base rate for most diseases is very low, so the actual statistical likelihood of false information from any given test is quite high. Since these tests can lead to unnecessary treatment with real side effects (or at least, anxiety) it’s better to not do ‘data gathering’ tests.

• Insurers won’t cover it.

• Most patients get ~15 minutes with the doctor, which isn’t enough time to really go through and get a detailed analysis that would lead to a good idea of what tests to prescribe, and it’s much more efficient (and gets more $$$) to simply shoot from the hip.

• Doctors don’t want to admit they don’t know something.

In the interest of fairness, I talked to a few friends of mine who worked in healthcare and sent them a draft of this post.

First, my old college friend Omid, who is now a doctor (a statement that I am both incredibly proud of and mildly mildly terrified by).

Off the bat he was pretty surprised by the baclofen diagnosis:

Which, yea, based on my initial experience, that was weird choice to tell me to take that every 8 hours! He recommended gabapentin instead and spent a fair bit of time bashing urgent cares.6 Starting to understand why Mom always wanted a doctor in the family.

Omid pointed out that most doctors in the NYC area in particular are extremely triage focused. EDs just get so much random inflow from people who may have annoying symptoms but are just clearly not in any real medical emergency, coupled with way too many people who are at severe risk of dying right now. If you are not in the latter category, you are essentially told to go home and figure it out. That’s not to say that every doctor’s office is treated like an emergency room. But most doctors do end up doing some ED shifts, especially at the big research hospitals (Sinai, Langone, MSK, Rockefeller…). So the culture leaks through.

He also said that Occam’s razor is a meaningful part of med school training. Basic labs and an xray are cheap and easy to do. But why even bother doing all that? Yea, fine, testing may provide more data, but empirically the first line treatment generally works!7

My friend David, who works in health insurance, basically agreed on the latter point. I’m going to just quote him directly:

[From the insurance side] the tradeoffs here are individual versus macro scale optimization of health outcomes + patient comfort, while optimizing and operating under constrained resources ($, limited numbers of medical machinery, limited numbers of providers, limited numbers of providers with the appropriate specialty).

The idea is that we only dedicate resources at scale to the things with high signal to noise ratio, and the highest signal is patient discomfort.

In your case, that discomfort requirement is clearly met. But that doesn’t mean jumping straight to a test is best for society as a whole -- the symptoms may resolve on their own / with medication before you’d even get the results back, the test may/may not find anything, etc.

“Do the thing cheapest/least resource intensive thing that’s likely to work first, then move on to the alternative.”

He even made this cool graph:

My friends are (generally) reasonable people, and they (generally) say reasonable things, and all of this is, shockingly, pretty reasonable. Like, yea, that makes sense, doctors aren’t twirling mustaches and even if I (a patient) want more love and attention, I (a member of society) recognize that in the grand scheme of things my hiccups are simply not life threatening even if they are supremely annoying.

It would be convenient if there was just a stupid reason for doctors to hate tests, because then I could at least put my incades-hic-ent rage at these hiccups somewhere. But no one sets out to be a villain, and everyone is responding to incentives.

V.

But I still think the incentives of this system are fucking stupid.

At a societal level the goal ought to be quality-adjusted-life-years (QALY), not fastest-and-cheapest-treatment-of-symptoms-that-make-the-patient-go-away.8 And though I am not an expert, I assume that the best way to improve QALY is by getting at the root cause of a thing and actually fixing it. Most of the time, the first-line treatment for a bunch of symptoms works because we have an accurate read of the underlying disease being treated.

I think you could maybe mount an argument that testing actually reduces QALY, because most tests are very likely to false positive (thanks, Bayesian statistics). David had a great line:

If you go looking for an abnormality in a human body, you will always find one.

But even here I am unconvinced.

First, doctors are supposed to be the ones interpreting the information that they get from tests.9 I think it is probably true that the average lay person won’t understand what a false negative is or why a PPI may result in one for an H. Pylori test. But why does that even matter? The lay person shouldn’t be responsible for interpreting anything. The doctor should just prescribe the test and then interpret the data the way they were trained to.

Second, this clearly doesn’t apply in a situation where the first guess is obviously wrong and waiting for six weeks won’t do anything. When all bets are off, why are we still mechanically clinging to the script?

And third, there obviously are doctors who are totally willing to get in the weeds with a patient and order a battery of tests and figure out what is actually going on. These are some of the best doctors in the world!

I think the real kicker, the thing that really ticks me off, is that tests are by all accounts cheap to produce and cheap to administer. Far cheaper than prescribing the wrong thing and wasting the patient’s and the doctor’s time with a follow up. Every trip to the ER almost always results in a suite of tests, as does every physical. Hell, they’ll take your blood pressure and temp and oxygen levels before you even see a doctor. They’re thrilled, they’re practically jumping for joy to take your blood pressure / temp / O2. I don’t think these tests are all that different from a blood panel or a stool test. And I just refuse to believe that we’ve coincidentally settled on the exact optimal amount of testing for every person, which just so happens to be ‘once a year-ish.’

The growing ‘preventative optimization’ and ‘longevity’ industries seem to agree with me — these are all about extremely comprehensive full body testing, catered to folks who are willing and able to pay. Because these medical practices prescribe tests more often, they are able to evaluate deltas off the patient’s baseline instead of what we do today where we try to fit everyone into the same standard box. I know that the jury is still out on the efficacy of these things, I know that the longitudinal studies have not yet come back. But also, I have way too many friends with some story like “I was lethargic and sick all the time and then I took one of these tests and discovered I have celiac, and now I just avoid bread and I feel better than ever” (and then they proceed to do a triple backflip or whatever). Whatever amount of money we are spending on making tests cheaper and more accessible is simply not enough.10

At the end of the day I am an empiricist and a scientist. Data is in my blood.11 I know there are a lot of big systemic problems with healthcare in the US. But this tiny thing just frustrates me to no end. Make it ok to order tests for patients.

(Also, if you have any remedies for long lasting hiccups, ideally ones that do not prevent me from operating heavy machinery, let me know)

The actual install was pretty straightforward.1 Took about 5 minutes. Explaining how to configure the thing took another 55. “Most of your configuration should be in SKILL files, which need to live in the skills directory. Also you need a CLAUDE.md file, which is basically like a SKILL file but it gets added to every prompt…what do I mean by that? No skills don’t actually get added to the prompt, the agent has to choose to read those, the skill descriptions get added to the system prompt. No they don’t exactly get added to the CLAUDE.md, but they kinda do, they are both part of the system prompt…ok yes subagents are a different thing than skills, but slash commands are the same thing as skills. But it’s all markdown. No, subagents also get access to your CLAUDE.md and your skills. But that also gets added to the CLAUDE.md. Also you have different sets of these possibly from every folder. Some of these will live in your git repo. No the agent won’t pick up the ones in the git repo automatically unless you copy the files in the right place. What’s a git repo? uh…”

This shit is way too hard and way too unintuitive.

The problem is that there isn’t any stability. The macro environment is constantly changing, with Anthropic et al shipping new foot guns basically daily. Here’s a great example. Claude Code uses CLAUDE.md. Codex CLI uses AGENTS.md. Gemini uses both AGENTS.md and GEMINI.md. Most people have, at this point, switched to AGENTS.md support for standardization. But Claude Code, the industry leader for this sort of thing and the one that every CEO seems to insist on using, forces it’s own standard.2

And it is just way too easy for a given user to totally screw their environment up. “Claude, add a skill to automatically load my AWS credentials” and now you have a security leak that in two months will take out all the data centers in Wisconsin, whoops.

I can’t imagine being a TL in this setting. Every TL I know eventually reaches for some kind of stable cloud dev box environment, because debugging someone’s python env by spending 4 hours over their shoulder is a great way to want to throw your computer into a lake and take up a career in goose farming.

AI has pushed everyone back to their local boxes. So now you have all the usual dependency problems, but also you have a new batch of AI weirdness. Raise your hand if you have had someone complain about Claude “getting way worse”, only to discover that they have 50000 tokens in their CLAUDE.md and another 20000 tokens used up by random mcp server tools.

I think the ecosystem makes this unnecessarily harder too. It is too easy to download random skills from external repos. There is no curation, and every agent is capable enough to make everyone dangerous. We saw this in action a few months ago with all the OpenClaw leaks. As I said before:

Meanwhile, internal teams are still sending around config files through slack, which is only marginally more efficient than walking over to your buddy’s place with a USB drive. The lack of team-wide organization has made it virtually impossible to effectively distribute something to a team larger than 8.

Taking a step back, many people have written about how AI has been consumer first. Quoting Karpathy:

Transformative technologies usually follow a top-down diffusion path: originating in government or military contexts, passing through corporations, and eventually reaching individuals - think electricity, cryptography, computers, flight, the internet, or GPS. This progression feels intuitive, new and powerful technologies are usually scarce, capital-intensive, and their use requires specialized technical expertise in the early stages.

So it strikes me as quite unique and remarkable that LLMs display a dramatic reversal of this pattern - they generate disproportionate benefit for regular people, while their impact is a lot more muted and lagging in corporations and governments

But that means the enterprise ecosystem is uniquely underdeveloped despite the massive demand. Everything that’s been released has really been targeted towards consumers, hobbyists, and tinkerers. I think it’s great that tinkerers can configure things to their taste, but that is zero help for someone like my mom who just wants some kind of admin controlled environment that has everything set up already. Her area of expertise is not dev tools! Having her set up a ton of dev tools and integrations and whatever else wastes her time and wastes the company time. If your sales people are doing the equivalent of managing the company VPN, something has gone horribly wrong.

I think there’s a big gap in the market for managed agent runtimes (also known as ‘background agents’ because they can run in the background on the cloud). Everything else — org level skills, for eg — is just a band-aid. I’ve said in the past that for coding agents, the filesystem is the config. That means that you have to control the entire filesystem to get a consistent (and secure!) experience. It is simply not enough to assume that everyone in the org will learn not just how to use AI, but also how to use a CLI and how to use Bash and how to use git and how to use the million other tools that are required to get a coding agent off the ground. And even that may not be enough. One of our customers, a CTO of a highly technical eng team, was saying earlier today that any time he makes a change to an agent system prompt:

I need to take ten calls just to make sure everyone is on the same page, and for the more junior engineers those will be video calls. Just a ton of work for what is potentially a single line change to adapt to a new agent like Opus 4.7 or whatever.

So far, the labs have totally dropped the ball on this. I kinda understand why — OpenAI and Anthropic make money off tokens, so they do not want to provide configurable environments that would allow someone to switch to their competitor’s models. But that just leaves the ecosystem wanting for a better solution. Large corps recognize the demand and have started building their own in-house solutions, including Ramp, Stripe, Spotify, Uber, Shopify, Block, and Jane Street (which, as usual, built theirs in ocaml). Each of these companies has teams of 10+ senior engineers developing this infrastructure while also leveraging years of pre-existing bespoke infrastructure and standardization. This is totally non-tenable for most series C-and-lower tech companies, not to mention all of the valuable companies that do not have this kind of software expertise in house.

More solutions are starting to pop up.

For folks who want to roll their own software, Vercel and LangChain recently open sourced background agent repos (along with a few others), though there are surprisingly few good tutorials out there for setting this sort of thing up internally. I have a conflict of interest here (see below), but in my experience the lack of easy walk-throughs is precisely because this shit is actually really hard to roll out. To get real adoption you need a real product with real polish, and that’s before you get into infra management and security and integration hell. All of that in turn requires a fair bit of maintenance to keep operating smoothly, especially if you’re mostly looking for AI enablement among non technical or semi technical teams. The product surface area is huge.

I think most teams should go for a fully managed service. Devin has been in the space for a long time. I think Twill is a relative newcomer that also does something like this? But there aren’t a ton of other services that actually follow the Ramp-Inspect / Stripe-Minions model. So when we ran into this problem ourselves, we weren’t happy with most of the other options out there and set out to build our own, which we now sell as an off the shelf option with a lot of customization flexibility (if you’re a ~Series A to ~Series D exec with a lot of AI FOMO, and you’re thinking about building your own background agents, reach out).

I have to say, having a tool like this is pretty incredible, and I’m really glad we built it. More than 30% of our PRs are shipped entirely through slack, and we have all sorts of really cool automations for things like bug triage to newsletter writing. But all of that takes a lot of maintenance, and we spend literally all day just thinking about how to make this one thing awesome, and it would be impossible to do that if we were building in, like, healthcare or manufacturing instead.

Of course, the difficulty of building and maintaining the product doesn’t absolve the need for something like it to exist. I think if you are an exec who is thinking about how to get your team to use AI effectively, and you care about AI enablement, you need to pave the way. And that means removing all the setup, removing the need to learn a ton of jargon, and removing the implicit requirement to stay plugged into Twitter. Once anyone can use AI, you start getting real creativity from people who can use these tools to superpower what they are good at. But that won’t happen if your team is still trying to wrap their head around when to use a SKILL and when to use an MCP, or even how to set up a PowerShell CLI. Just hand it off to someone else and get back to shipping for your customers.

Agentics is the study of how to use and reason about agents. If you are an expert in coding agents, or interested in learning more about agents, join our community slack. More articles here. Learn more about Nori at https://noriagentic.com/

The way open source licensing works is pretty straightforward. By default, any code you write is yours, you have a copyright on it. No one is allowed to use it. This applies to million line codebases, and it applies to the smallest of code snippets on Stack Overflow.1 In order for someone to use code without running afoul of the law, the author of the code has to give explicit permission detailing what can be used and how. These permissions are often laid out in a LICENSE document.

Behind our current, relatively straightforward understanding of open source licensing lies ~45 years of court battles and lawfare, because trillions of dollars are at stake. That’s not a typo. Take Linux as an example. Your phone runs Linux. The web server your phone talks to runs Linux. All of the machines that pass messages between your phone and that server run Linux. Your blender used to run Java, but now it runs Linux. Linux is the most valuable individual piece of software in the world. It’s worth $0, you can download a copy right now, for free. So the question of who owns Linux and how it can be used are really really important.

One of the most important license innovations was the invention of ‘copyleft’. The basic idea: anything that is built with a copyleft license has to also use a copyleft license. Even though Linux is open source, and even though you can download a copy of it, you cannot package it up in a closed source box and try to sell it. You are legally required to release the source code for anything that derives from Linux.

It is shocking how much of the modern tech stack depends on copyleft licenses. Most compilers are copyleft. The GNU Binutils are copyleft. Git is copyleft. So is Bash. MySQL, VLC, coreutils, glibc, ffmpeg. More recently, wordpress, mongodb, elasticsearch. These are foundational pillars of the global tech stack. Thousands of companies have spent tens of millions of dollars of employee time on these open source projects, because the companies depend on them somewhere in their stack. Many of those companies would have preferred their employee time going to closed source software that they could then redistribute at a profit. Copyleft licensing prevents that. And many projects — like MySQL, which uses a copyleft license for general use but offers a paid option for companies who want to get rid of the copyleft requirements — could only get funded through the existence of copyleft licensing.

As you might imagine, many people have spent a lot of time thinking about how to get around these licenses. Code is and always has been in a weird gray area when it comes to intellectual property. You can’t copyright math. You can’t copyright an idea. But you can copyright code. So what, exactly, are you protecting? The courts say that you are protecting a specific expression of an idea. If someone copied the Linux kernel, or if someone wrote their own kernel while looking at Linux, all of that new code is based on the previous implementation. So it’s all protected by the Linux license. But if someone just, like, read about the Linux kernel, and got a really good understanding of how it behaves, and then made their own version of the kernel, that would be a new implementation and would be totally fair game.

The technical jargon for this is a “clean room implementation.” Team A spends time pouring over the code and writing an extremely detailed specification without explicitly writing code. And then Team B, which has never looked at the original source, writes new code to meet the specification. Team A and Team B don’t interact at all otherwise, to ensure the final output is ‘clean’.

Traditionally this kind of license circumvention is extremely costly. It requires a lot of time and at least two (teams of) people.

AI makes this trivial.

You have a session of Claude looking at the original code base and writing a spec. And then a different session of Claude looks at the spec and writes new code. The (untested) legal theory is that this is sufficient to remove the license, because the new code is “clean”.

People are already using this strategy to remove licenses.

Chardet, (HN Link)

In the world of open source, relicensing is notoriously difficult. It usually requires the unanimous consent of every person who has ever contributed a line of code, a feat nearly impossible for legacy projects. chardet, a Python character encoding detector used by requests and many others, has sat in that tension for years: as a port of Mozilla’s C++ code it was bound to the LGPL, making it a gray area for corporate users and a headache for its most famous consumer.

Recently the maintainers used Claude Code to rewrite the whole codebase and release v7.0.0 , relicensing from LGPL to MIT in the process. The original author, a2mark , saw this as a potential GPL violation…

Claude Code Leak

Context: The viral GitHub fork of the leaked Claude Code was at immediate risk of a DMCA takedown (Anthropic had killed prior mirrors in minutes), so its maintainer — worried about getting sued — used OpenAI’s Codex to rewrite the entire ~512k-line TypeScript codebase from scratch into Python overnight as a “clean-room” reimplementation.

This preserved the full agent harness, tools, and behavior without copying a single original line, instantly turning a copyright landmine into the safe, exploding open-source version everyone is now starring.

malus.sh, (HN Link)

Finally, liberation from open source license obligations.

Our proprietary AI robots independently recreate any open source project from scratch. The result? Legally distinct code with corporate-friendly licensing. No attribution. No copyleft. No problems.

…

Our proprietary AI systems have never seen the original source code. They independently analyze documentation, API specifications, and public interfaces to recreate functionally equivalent software from scratch.

The result is legally distinct code that you own outright. No derivative works. No license inheritance. No obligations.

I’m not sure if the last one is satirical — it is literally named ‘evil corp’ — but according to Reddit:

Clearly meant to be satire, with the name of the company basically being “EvilCorp” and the fake user quotes from names like “Chad Stockholder”, but it does actually accept payment and seemingly does what it describes, so it’s certainly a bit beyond just a joke at this point. A livestreamer recently tried it with some simple Javascript libraries and it worked as described.

Vercel reimplements Bash

A virtual bash environment with an in-memory filesystem, written in TypeScript and designed for AI agents.

Broad support for standard unix commands and bash syntax with optional curl, Python, JS/TS, and sqlite support.

after which Cloudflare rebuilt Next.js

Last week, one engineer and an AI model rebuilt the most popular front-end framework from scratch. The result, vinext (pronounced “vee-next”), is a drop-in replacement for Next.js, built on Vite, that deploys to Cloudflare Workers with a single command. In early benchmarks, it builds production apps up to 4x faster and produces client bundles up to 57% smaller. And we already have customers running it in production.

which Vercel then got mad about!

Open core here means having an OSS project with licenses like AGPL or BSL that allow anyone to derive from the software but only the original author to provide it as a multi-tenant platform.

This is very different from Cloudflare slop forking next.js. They made a choice to slop fork, but they could have just pressed the trad-fork button in Github since next.js is MIT licensed.

The licenses "protecting" open core software assume that making soften is hard, but they don't protect from a slop fork which reproduces the behavior without directly deriving from the license-encumbered implementation.

What I'm not sure is whether this means less open source or more liberal licenses as folks realize that they might as well put it out there now that everybody can copy it anyway.

I cannot stress enough how much this is a fully untested legal theory. In the background of all of the above, there has been an ongoing legal fight over whether AI generated content can be copyrighted at all. And so far, the answer is no!

From CNBC:

The U.S. Supreme Court declined on Monday ⁠to take up the issue of whether art generated by artificial intelligence can be copyrighted under U.S. law, turning away ​a case involving a computer ​scientist from Missouri who was ​denied a copyright for a piece of visual art made by his AI system.

Plaintiff Stephen Thaler had appealed to the justices after lower courts upheld a U.S. Copyright Office decision that the AI-crafted visual ⁠art ‌at issue in the case was ineligible for copyright protection ⁠because it did not have a human creator.

Thaler, of St. Charles, Missouri, applied for a federal copyright registration in 2018 covering “A Recent Entrance to Paradise,” visual art he said his AI technology “DABUS” created. The image shows train tracks entering ‌a portal, surrounded by what appears to be green and purple plant imagery.

The Copyright Office rejected his application in 2022, finding that creative works must have human authors ​to be eligible to receive a copyright. U.S. President Donald Trump’s administration had urged the Supreme Court not to hear Thaler’s appeal.

The Copyright Office has separately rejected bids by artists for copyrights on images generated by the AI system Midjourney. Those artists argued that ⁠they were entitled to copyrights for images they created with AI assistance - unlike Thaler, who said his system created “A ‌Recent Entrance to Paradise” independently.

A federal judge in Washington upheld the ‌office’s decision in Thaler’s case in 2023, writing that human authorship is a “bedrock requirement of copyright.”

If the most extreme version of that line of reasoning applies to code,2 then all of the code written by AI may very well be uncopyrightable, i.e. acts as if it was in public domain.3 But even in less extreme interpretations, the actual litigation will depend on whether or not the AI did substantive expressive work. And for a straightforward cleanroom overnight implementation, where the AI is doing literally all of the analysis while the user is sleeping, then it is totally possible that the output of the AI itself is a derivative work and carries the corresponding obligations of the input license.

Most open source maintainers are not about to go to court to litigate these license infringement cases — this is, in part, why all of this is still a grey area to begin with.4 But clarification of intent goes a long way. It is much easier to eventually defend copyright in a court of law if you are clear from the beginning about how that code ought to be use. And law is not code. Social weight matters. There is a huge difference between a lawyer going “this is a grey area but it’s probably fine” and “this is a grey area so I wouldn’t risk it.”

With all that in mind, we introduced the Ship of Theseus license to all of our open source codebases. This license aims to plug the AI clean room hole. It is a very simple license, with only two lines:

SHIP OF THESEUS LICENSE v0.1

* Using any AI tool to produce functionally equivalent software — by
  referencing this code, its documentation, its behavior, or any
  specification, description, or abstraction derived from the
  foregoing — creates a derivative work subject to the full terms of
  the primary license, regardless of whether the output shares any
  literal code with this project.

* Any derivative work must include this license alongside the
  primary license.

On its own, the Ship of Theseus license does not grant any claims or enforce any limitations. Rather, it makes clear that any AI derived work is exactly that: derived work.

We still haven’t fully nailed down whether AI tools themselves, which have almost certainly been trained on all open source material already, can even count as being ‘clean’ in any sense. But I’m not taking any chances. I’d rather have some explicit indication of my legal intent than to throw up my hands and assume any open source licensing is dead.

In order to get clarity, this sort of approach requires wider adoption, so if any of this resonates I strongly encourage other open source maintainers adopt this license or a similar one.

Note: after sharing the original Ship of Theseus license around, a friend linked me to Armin Ronacher’s blog from ~1mo prior where he independently came up with a similar analogy. The Ship of Theseus license wasn’t inspired by Ronacher, but the convergent evolution of the name hopefully means that it’s intent is intuitive to understand from the name alone.

Scott Alexander writes an excellent series of posts about Bay Area house parties. He’s written more than a half-dozen at this point. They all involve the straight-man audience/scott-insert (and the only sane man left alive, apparently) entering some strange lovecraftian event that, if you squint, could be called a house party. The weirdness of the event is, of course, the satire. Everyone else seems totally oblivious that they are the joke. These parties are completely and totally ridiculous, and any reasonable reader would assume that Scott is simply embellishing or straight up making up details and that these parties are not really real.

Which is why I am concerned that people do not believe these parties are really real.

Two weekends ago I was doing my quarterly trip to the Bay. I was invited to a spring gay peptide party.

O, sorry, you don’t know what peptides are? Wow, really behind on the times. AI is pretty lame in the Bay these days, because everyone is just swimming in it. People ask each other ‘what do you do? (for work)’ and if you say ‘O I work in AI’ they’ll look at you like you just said you’re best friends with Curtis Yarvin — who is now, by the way, deeply uncool (more on this later). Of course you work in AI, everyone works in AI. Saying you work in AI is like saying you work in tech, it’s already priced in. Which of course makes the phrase ‘I work in AI’ the least useful / interesting thing ever. I had at least three people give me a pitying glance before someone kindly informed me that since AI was obviously going to take over everything, it just wasn’t interesting to talk about.

Peptides. Now peptides are cool. And not just any peptides, but “cheap Chinese peptides.” I heard the phrase “cheap Chinese peptides” at least a half dozen times from as many people during my trip to the Bay, and I was only there for 2 weeks.

I have a bit of a molecular bio background so everyone being really into ‘peptides’ was a bit of a ‘wtf is going on’ moment for me. For those who don’t know, a peptide is a completely unspecific term. Saying that you are really into ‘peptides’ is about as specific as saying you’re really into ‘proteins’ or ‘molecules’. “Ah yea I’m really into those Chinese molecules these days” just doesn’t hit the same does it? A Chinese friend of mine quipped that he was also very into Chinese peptides, all things considered. From what I could gather, peptides-as-used-in-the-SF-party-scene are injectables like semiglutide (i.e. Ozempic). Most of them are for weight loss, but some folks swore that they had peptides for everything from skin rejuvenation to better sleep health.

So what is a spring gay peptide party? Well, the party was peptide themed in that

1. Everyone seemed to be on them

2. They had extremely strong jello shots being served out of big syringes

3. At least one person but possibly multiple people were injecting each other with peptides at the party

And also most of the guests were gay (obviously) and it was also spring.

At one point I was in a conversation with no less than 4 other founders who were all building peptide companies. I could fill a notebook of quotes from this conversation. “They change your personality, it’s literally made me less shallow knowing that we can just looksmax you.” “Ugliness is just a choice now.” “I shot up a twink with ozempic who did not need to lose any weight.”

My favorite exchange by far was between a guy in a sailor cap and another guy in a long black leather trench coat (it was pretty warm out).

• Goth: “What peptides are you guys on?”

• (crowd mumbles some answers)

• Goth: “Wait are you using that one, reta?”

• Sailor: “Everyone is doing street reta.”

• Goth: “Wait so you actually do it?”

• Sailor: “I’m on tirz too.”

• Goth: “You do both?”

• Sailor: …

• Sailor: “I’m on a lot.”

The pause gets me every time.

Fun fact, Scott actually first mentioned peptides back in January, also in the context of a bay area house party!

Sam types in spaghetti bolognese, delicious, scrumptious, meaty, trending on DoorDash, --dangerously-skip-parmesan and hands it back to Tran, who clicks ORDER.

“Nothing for you, Tran?”

“Nah,” says Tran. “I’m on Chinese peptides. Retatrutide, GLP-1 receptor agonist plus a bunch of other downstream effects.”

“Oh,” you say, “interesting. I’m still on tirzepatide, but I’d love to learn more. Where did you learn about suppliers and doses and stuff? Was it the locked Cremieux post?”

“Cremieux’s post is okay, but there’s a lot of tacit knowledge that didn’t make it in there. I’m actually working on a guide to all the GLP-1s. I’m calling it If Anyone Builds It, Everyone Diets.”

So he is way ahead of the game.

I suppose this is one of the downsides (upsides?) of living in NYC instead of the Bay. Although maybe that didn’t matter as much as I thought. Almost everyone I met was from NYC. “O yea, I’m just in town visiting” starts to feel a bit surreal after the 4th time. Even the host was visiting! It wasn’t even his house!

In college, my freshman year floor had a little motto: “always double down.” It’s a bit like the improv ‘yes and’. No matter what was happening, no matter how ridiculous the conversation was, you always double down. So normal chit chat would rapidly spiral into insanity that everyone would play with a straight face, which would often form the basis of recurring bits.

This was, of course, all a bit edgy and a bit cringe, but it was also a lot of fun if you were able to maintain a bit of ironic distance from the whole thing. It worked because everyone knew you weren’t serious.

I feel like SF is what you get when you take that motto and apply it really really seriously. You can’t just be taking ozempic, you have to be on reta, no, street reta, no cheap Chinese street reta, and also on tirz too! Or take the AI thing. Even though I’m generally concerned about AI, I’m not about to start a death cult that murders AI researchers.

What’s hard is that there clearly are some people in SF who are in on the joke, and who are purposely exaggerating precisely to get a rise out of the people who are taking everything seriously. (Maybe this is all the visiting New Yorkers?)

Someone once said that SF is a town of extremely high sincerity, and all of its modern and historical weirdness — the AI doomerism, the cults, the hippies, the drug use, the polycules — is downstream of people saying things and other people taking them extremely seriously. So you have people who are on way too many peptides. But that same high sincerity is also what makes SF so great. In my last post, I wrote

San Francisco is a city of dreamers. It's core enterprise is creating magic. And as a result, the town values the individual. It exalts archetypes — the founder, the builder, the prodigy. The question that people ask is "who are YOU? What are you interested in? How are you going to change the world?" Social life in SF is grounded in these questions…

You can never get the crazy valuations and capital necessary to build OpenAI in New York. Partially, that's because the appetite isn't there, it's too speculative. And partially, it's because the default mood is one of pessimism. "You think you can change the world? Who do you think you are?" People aren't on a mission in NYC, the way they can be in SF. They want to make money, sure. But maybe not change the world.

In some sense, a startup can only happen in an extremely high sincerity environment. If a 14 year old says that they are going to change the world, they are being very sincere even if an ‘adult’ knows that the likelihood is low. It takes an equally sincere kind of person to double down on that energy. But that’s SF in a nutshell. So, yes, you get crazy peptide gatekeeping, but you also get amazing things like self driving cars and the LGBT movement and YIMBYism and so on.

By contrast, I think NYC thrives on irony. Which sometimes leads to some funny / awkward moments, where I’d make a joke and someone else would take me at face value. For example, I recently posted an April fools joke on LinkedIn about firing my team and switching from building off the shelf background agent runtimes to “AI powered peptides”, and at least three people messaged me congratulating me on the pivot.

One last thought on house parties: citrus.

When Mia and I were in SF we got invited to a fruit themed party. Specifically, citrus. Everyone had to wear citrus clothes, you’d be turned away by a bouncer if you weren’t. There were oranges everywhere. Apparently the big theme was that you had to find a kumquat that was hidden somewhere in the house? Very unclear what you would win if you got the kumquat. I think just respect.

We seriously considered going — we coincidentally knew a lot of folks on the guest list — but decided to blow it off in favor of walking around the city. We walked a lot. On Sunday we walked from the ferry building to Baker’s Beach, all along the coast; on Monday we walked from Chinatown to the end of Golden Gate Park (which, by the way, is confusingly not where the golden gate bridge is).1

But maybe we should’ve gone to the party. Little did we know, this citrus party would go soft viral. Over the course of the next week, I met two dozen independent and unconnected people who had all heard about this citrus party. Some thought it hilarious, some thought it stupid, but all of them were talking about it. I’m sure the host was quite happy with himself (I know you’re reading this, proud of you king!)

Am I cool for having been invited to the citrus party? Am I cooler for not having gone?

AI billboards remain inscrutable.2

I saw a few from graphite in particular that I really didn’t get, all playing with this same “art” theme. I think the idea is that if you just put the graphite logo next to a bunch of arty-sounding platitudes, people will associate graphite with feeling artistic? But it feels like a tenuous connection at best.

A year ago, when I last wrote about the Bay, I was surprised and dismayed to find that edgy right wing black pilled nonsense was considered ‘cool’. I met several people with really just rancid politics, people who were unabashedly pushing insane far right ethnonationalist conspiracies. The general sentiment was that Kamala and the libs were stodgy and old and uncool, and Joe Rogan and the other podcasters (SF loves podcasters) were cool and increasingly not libs. This is why Curtis Yarvin, jester of the new right, had his own group house, his own house parties, complete with the acolytes and the entourage and so on.

I’m happy to report that most of that is gone.

Sometime in the last 6 months, everyone collectively decided that being super right wing is actually really cringe. A lot of people tempered their previously vocal opinions. Others, who stuck to their nihilism, were just increasingly not invited to parties. No one really talks to or about Yarvin anymore.

It’s hard to say what, exactly, led to the change. One friend said that Anthropic was ascendant and was single handedly making the left cool again. Another hypothesized that the Iran war was hurting San Francisco’s wealthier residents (you need helium to make chips, after all), and because those people host all the events, the increasing disdain trickled down into the water supply. A third said that it was all just becoming too hard to defend, that the incoherent whiplash made any kind of principled position impossible.

But regardless of the reason, everyone agreed: “wow, it’s kinda really embarrassing that we spent so much of last year partying with real life eugenicists.” No, really?! You think so?!

Something that surprised me: the new Tesla self driving model is actually very good. I had mostly been of the opinion that Tesla would always struggle to really provide a good self driving experience, because they didn’t have lidars or other sensors (something I wrote about here). Every time I had been in a self driving Tesla, I’d been very much afraid for my life — the thing was way too aggressive, took turns stupidly fast, and regularly had to be stopped before crashing into something.

But (according to friends with Teslas) about 6 months ago the model updated and now everything is gravy. My buddy drove his Tesla from SF to Santa Barbara and said he never touched the wheel.

It’s true that the Tesla’s still aren’t as good or as safe as the Waymos. But they are way better than human drivers and, more importantly, can drive anywhere. If I had to choose, I’d take a free roaming mostly autonomous vehicle over a location fenced fully autonomous one every time.

Very curious what led to the improvements. More data? Better sim? More compute? If you’re at Tesla and you have some insight here, drop me a line!

On our first day in SF, Mia and I did a brunch double date with an old friend from highschool. Normally this wouldn’t be relevant, but it’s worth mentioning that said friend is gay. This is worth mentioning because, while describing SF to Mia, he went on a long and surreal tangent about the “warehouse full of twinks down in Soma”, and I wanted to type the phrase “warehouse full of twinks” while making it as clear as possible that these were not my words.

On the one hand, this is an incredibly information dense phrase. What is it? A warehouse. What’s inside? Twinks. How many? A warehouse full.

On the other hand, what do you mean “warehouse full of twinks”???

“O, you know.”

No, Steve, I do not know!

Trying to get a handle of the situation, I asked some follow ups in the hopes that I would get something more legible.

Me: “What, exactly, do they do at this warehouse?”

Steve: “They chant to Claude about their desires.”

Me: “They…chant to Claude?”

Steve: “Sometimes ChatGPT”

Me: “Why?”

Steve: “They want to escape the permanent underclass by building b2b SaaS.”

Me: …

Steve: “they also fry their brains on Twitter. The twinks there aren’t getting enough sleep, it’s like rows and rows of tired twinks doom scrolling Twitter 40 hours a day while they chant to their machine god”

Me: “...are they happy? Like do they know that this isn’t the good life?”

Steve: “no one is making them stay in the warehouse. Also they have good parties”

My friend is a bit of an eccentric character, so I mostly assumed he was exaggerating. But maybe twenty minutes later, a girl pops by our outdoor table, and Steve just lights up. “Amol, Mia let me introduce you, this is my friend who runs the warehouse full of twinks I mentioned earlier.” Turning to our guest “We were literally just talking about you, how are the twinks?”

“O they’re good, working hard on escaping the underclass.”

???????

After a bit more conversation, I eventually piece together that this project started as a group house run by 19 year olds, one of the people in the group house had some kind of startup exit and decided to use the money to buy a warehouse, and it eventually became / doubled as a co-working space. “We have a digital sculpture that’s a pile of retro TVs in the center of our co-working space that we sacrifice goats to” I’m sure you do.

There are many forms of AI psychosis. Most people are familiar with the strain that breaks up relationships, where an overly sycophantic AI basically acts as the world’s worst relationship therapist by mimicking an AITAH comment thread (“ooo your husband of 7 years didn’t take out the trash on Tuesday? yea that’s a red flag, break up with him now.”) But I’m here to tell you that there is a unique strain that is prevalent in SF, that seems to only infect engineers.

It starts benignly enough. An engineer will set up Claude Code and try it out, making a little hobby project like a game or a web app. It will feel easy. In fact it will feel too easy. Here, the engineer makes a critical mistake: they begin to believe that everything that was previously hard must now be easy. They will build bigger and bigger applications, fed by the validation of creating and merging prs. They start building more things themselves. Who needs npm packages, who needs infrastructure, build your own containers and kubernetes and AWS. Things will spiral in complexity, bugs keep popping up but the AI always sounds optimistic, always on the verge of a break through, “Now I see the issue.” Soon the engineer is a husk, just clicking approve on everything while running ten agents at the same time. Yes, it’s true, code is being written, features may even be shipped. But the engineer has forgotten the number one rule: the best code is no code.

When someone comes to you and says that they’re going to make their own programming language because Claude said it would be faster than rust and easier to use than Python, they’re too far gone.

Mia says that SF feels like highschool.

SF has a population of about 850k. This is way smaller than I thought, especially given SFs cultural weight. NYC by comparison is a city of 8 million.

Of that 850k, 20% or so are between 18-35, and 20% or so work in tech. If you assume there’s no correlation between these things, you get ~34k people. That’s bigger than most highschools, but still roughly the size of a mediumish state school. I think you could cut it down a bit further — works or has worked in startups, is likely to go to a house party, has been in the Bay for more than a year. The point is that you pretty rapidly get to a group of people where it feels a bit like everyone knows everyone else, or is at least one hop connected to everyone else. Like, in my own network, I’m pretty sure I’m one hop connected to basically every famous tech billionaire, and that’s just because I happen to know a bunch of people who live in SF and go to parties.

This can sometimes make SF feel a bit like that scene in Mean Girls, where we get a rundown of the social graph through where people sit for lunch. There’s the crypto bros and the ai doomers, the big tech lifers and the creative coders, the health maxxers and the kinda scary powerhouse PMs. The group house culture and the social dynamics of the VC accelerators exacerbate the effect.

For a town that is ostensibly focused on tech, the only thing everyone talks about is the people. She’s dating who? They had a founder breakup? Wait he’s working where now? It’s a mix of “highschool social graph” and “celebrity gossip.” Extremely potent combination.

Anecdotally, I think this leads to some pretty aggressive stereotyping, as people try and fit you into one of the tables. I notice a lot of folks lean into the stereotypes, especially founders. It makes them legible. If you’re selling peptides, it helps your business to make your whole personality ‘peptides’. That’s how you get referrals! “O you’re looking for peptides? Let me hit up my peptide guy, he knows everything about peptides.” A good friend of mine moved to LA to work in show business. When he went down there, he changed his name to “Sven” because it was more memorable. In a town where ‘who you know’ determines whether or not you have a job next month, social currency has an exchange rate to USD. In this, SF isn’t all that different.

My SF based friends object to the highschool metaphor. It’s not tables at lunch, it’s techno-feudal houses.

There’s House Altman and House Amodei. There’s House Musk and House Zuckerberg. There’s House Brin and House Jensen (for some reason this one isn’t a last name). And underneath the Great Houses are their bannermen, like Ser Dwarkesh, a member of House Amodei; or Ser Roon, a longtime member of House Altman. Sometimes the bannermen can switch allegiances, as when Ser Alexandr Wang of the minor House Scale betrayed House Altman and joined House Zuckerberg.

On the one hand, I think this is mostly tongue in cheek. On the other, this is a very high sincerity group. And I think back to the warehouse, and the talks of “permanent underclass”, and wonder if maybe those peptides are responsible for more than just a few dropped pounds.

If I had a nickel for every time someone told me about “their friend Jim who was at HHS and recently moved to DHS,” I’d have two nickels. Which isn’t a lot, but it’s weird that it happened twice.

A few folks asked me what it was like running an AI company in NYC. I think they mostly expected me to say that it was a lot harder, that I was sacrificing access to clients and talent and marketing opportunities by stubbornly staying out in the East Coast. I can see why they’d think so — building with AI is so common in SF that it has become old news. There are few places in the world where people have thought as much about AI best practices.

But that exact trend is why I prefer building in NYC. Every single person that I met in SF was dangerously opinionated about AI best practices. It is impossible not to be! When everyone is constantly jumping from idea to idea, trying to stay on top of the Twitter firehose, you need some kind of opinion just to stay relevant and sane. So I met dozens of people building complicated multiagent harness systems with all sorts of handoffs and state tracking and pipelining, vibe coded scaffolding for vibe coding more scaffolding. Did it actually result in better products? Better code? Could you do more with less? These questions never came up.

I like building in NYC because I can talk about results instead of trends. Nothing, and I mean nothing, is set in stone anymore. Every part of every organization is reinventing best practices from the ground up. In that environment, the worst place to be is in a bubble. And SF is very much a bubble, a really really noisy one. It’s super hard to burst through with good foundations level thinking, even if you have good results, because it’s not ‘hype’ enough. In NY I can sell on what actually works instead of on blindly following whatever Karpathy posts about this week.

I think eventually there may be some need to focus more on the Bay, but for now I don’t feel any cost at all from building in NY.

Even though I got to catch up with friends — including some of you folks who read this blog! — and do my bi-annual data gathering trip, it’s nice to be back home. Sleeping on 7 couches in 13 days doesn’t quite work as well once you hit 30. At some point I suppose I’ll just start getting hotel rooms.

Which may be sooner rather than later. My rate of travel to the West Coast is picking up.

I want the record to show that I think this is dumb:

Shares of Allbirds (BIRD.O) surged more than five-fold on Wednesday after the footwear maker said it was raising capital and pivoting towards AI computing infrastructure.

The ​San Francisco, California-based company said that it would execute a $50 million convertible financing agreement ‌with an institutional investor and plans to use the proceeds to acquire graphics processing units (GPUs).

Allbirds also plans to rebrand itself as “NewBird AI” and, over time, shift focus to offering cloud computing ​capacity and AI services, though it did not provide additional details on ​its new strategy.

…

The stock was last up 435% at $13.33, valuing the company at $116 million, according to LSEG data. Allbirds was also among ​the most active ​orders on Fidelity's ⁠trading platform on Wednesday, signalling interest from retail traders.

As of writing, actually, its closer to 500%.

“SF based startup pivots to AI” is not, like, an uncommon headline in 2026. But normally those headlines are for

• companies that are small and have a lot of room to pivot because they do not have a ton of customers

• companies that are at least sorta kinda tangentially related to AI, or at least software

AllBirds is neither. It is a public company with over 500 employees that sells shoes. AllBirds spinning up a GPU arm is like SpaceX buying a social media company you know what nevermind.

If this whole thing reads like a Matt Levine bit, that’s because it is [EDIT: while I was writing this Matt beat me to it]. This is basically the same approach that MicroStrategy took with crypto / bitcoin back in 2020, that it has continued to double down on in the intervening years. According to wikipedia, MicroStrategy is a public company “that provides business intelligence (BI) and mobile software…to analyze internal and external data in order to make business decisions and to develop mobile apps.” According to anyone who has paid attention to the company since 2020 (including Wikipedia!), MicroStrategy is a pot of bitcoin that you can buy on the stockmarket.

In 2020, MicroStrategy went ‘this strategy isn’t working out, we need a new strategy’, and decided to go all in on crypto currencies, which, from 2020-2023, was a very hot thing.

Funny enough, the value of MicroStrategy shares is worth more than the equivalent value of the bitcoin its holding! From Matt:

MicroStrategy is primarily a pot of Bitcoins, though it also has a software business that is grudgingly acknowledged in its earnings release.[3] The pot of Bitcoins is worth about $18 billion, according to that earnings release, and is funded partly with debt; MicroStrategy reports about $4.2 billion of long-term debt.[4] This suggests a net asset value of, you know, $14 billion? Plus or minus 50%? Its market capitalization is about $50 billion.

I don’t really know what to make of this, except to say that this feels dumb and perhaps this is why I am not in finance.

This is also, by the way, the same thing that Kodak tried back in 2018. Do you guys remember KodakCoin?

Now, the 130-year-old company is trying an unlikely sort of comeback — one built by betting on cryptocurrency. It’s a bold gamble that has excited some investors, perplexed others and raised questions about how closely Kodak vetted its cryptocurrency business partners, who now include a paparazzi photo agency, a penny-stock promoter and a company offering what has been called a “magic money making machine.”

This month, Kodak lent its name to a digital currency called KodakCoin, which is billed as “a photo-centric cryptocurrency to empower photographers and agencies to take greater control in image rights management.” The basic idea behind KodakCoin is to use the blockchain to help photographers manage their collections by creating permanent, immutable records of ownership.

And it worked for them too! At least, until it didn’t.

Does the partial success of MicroStrategy / Kodak mean that this is a great idea for AllBirds? Again, I am not in finance. AllBirds is clearly struggling, and I don’t blame them for throwing a hail mary. But there is something silly about just being able to whisper the word ‘AI’ near a public stock and have the price go shooting up. The conventional wisdom in the startup world is that successful businesses are run by people who deeply understand the problem space they are operating in. You have to have a lot empathy for the user in order to build something useful. The AllBirds folks have spent years and years thinking about shoes. Do they really understand GPUs? Again from the Reuters article linked above:

“It looks like an attempt to capitalize on the AI movement. I don’t see how Allbirds brings anything to the table beyond name recognition,” said Bruce Winder, ​an independent retail consultant.

It’s not even clear the name recognition helps, because, again, people associate AllBirds with shoes and not AI.

The only kind of financial theory where any of this makes any kind of sense is if you see GPUs themselves as a store of value. You buy a pot of GPUs because you think that the future demand for them is going to be stable enough that it can act as a reserve for your organizaton/business/government.

Traditional stores of value include gold and actual currencies like USD. People have floated that bitcoin itself is a store of value — because it is fully decentralized and has a fixed supply, in theory it should retain value as long as bitcoin itself is deemed useful (and I do tend to think bitcoin will basically always be somewhat useful, I’ve written about some of that here). More recently, Sam Altman has been pushing the idea of compute as a store of value (cf ‘universal basic compute’) and others have spoken about compute credits as an asset class.

This of course raises the question: are GPUs a good store of value?

Hahahahahahahaha no. No of course not.

I was born in the mid 90s. A dollar back then is worth about half as much as it is today, a 50% decrease. Around when I was born, the top of the line best CPU was the Intel Pentium Pro. This thing would sell for $1168 back then, so lets say ~2k today. Do you know how much an Intel Pentium Pro goes for today? Well, they don’t make the things anymore, so you have to get them used, which costs like $80. This is, what, a ~99% depreciation? Despite the fact that global demand for compute has gone up a trillion-fold?

People like to talk about ‘compute’ and investing in ‘compute’ as if you can somehow get exposure to the abstract platonic concept of running electricity through silicon. Unfortunately, compute is tied to real world assets that depreciate very quickly. Just because everyone needs it does not make it a good store of value (see also: oil).

I can kinda sorta see how compute credits may be a reasonable store of value. You can get AWS credits and then be guaranteed some abstract floating amount of compute that will float on top of the depreciation of the underlying machinery. $2000 of compute in 1995 will buy you a single Intel Pentium, which is useless today. But $2000 of compute credits in 1995 will, in theory, get you $1000 of compute today, which is actually still pretty good!1

But this is basically just like saying gift cards are a good store of value, because that’s what a compute credit is. Like, yea, I guess gift cards do trade on a secondary market, but they are always discounted to their dollar value because they price in the likelihood of the company going under and the limitations on purchasing. You can buy a $100 Carnival Cruise giftcard for $93 — a 7% discount — because you basically never need to buy things from Carnival Cruise. Meanwhile, a $50 Amazon giftcard goes for $49.75 because it’s basically as good as money. At which point…you might as well just use money.

Anyway, I don’t expect the ‘buy a bunch of GPUs’ strategy to go well for AllBirds, but fingers crossed — they made comfy shoes, so I hope they succeed.

[EDIT: some friends mentioned that this is basically a spac, where the value of AllBirds is essentially that it is already a public company and so can skip the usual SEC filings and so on of taking something else public, and I also agree with that analysis]

Given how things are in 2026, I think it’s fair to say that that line of reasoning didn’t really go that well for them.1

The big issue is that it turns out there are lots of reasons why people will do things that are harmful to themselves. This is not exactly a brand new idea, at least in the common-sense understanding of things. Alcoholism is such a well known issue that supposedly an entire ethnic group developed a genetic aversion to the stuff, and history is littered with temperance movements and the like. But all previous attempts at regulating addiction were framed as ethical, pseudo-religious necessities. God created the world, but he also created alcohol, or something, anyway no you can’t have whiskey.

The regulations that sprang out of the mid-to-late 90s were grounded in science, which was a bit new, and was only really possible because science had caught up to our intuitions. We could point to causal mechanisms of addiction. Those mechanisms drove choices that these companies purposely made. Those choices made their products (cigarettes) more harmful to end users. And that harm was the basis for liability. The tobacco companies ended up having to pay massive fines ($200+ billion), while state and local legislation effectively banned smoking in most public places.

As a result of that work — both educational and legislative — we have a simplified but very well understood language for addiction. There are substances that can make someone behave as if they weren’t themselves, and these substances should be regulated.

It is massively helpful to have this baseline understanding of addiction in the general consciousness. Before, people would say silly things like “anyone can quit cigarettes, they’re just lying if they say they can’t” (while taking a drag of their fourth cigarette before breakfast). The language of addiction changed this. By making it clear that addictive substances changed a person’s brain chemistry, it became way more acceptable to shift the conversation from personal responsibility to corporate liability.

Of course, the language of addiction is very simple, and as a result there are a whole bunch of addictive things that don’t fit neatly into the language above. For example, gambling. I think basically everyone knows on some gut level that gambling can really screw someone up. And there’s lots of research to back this up. Just to pick one out of a hat, fMRI scans show that gambling really can change brain chemistry. But it’s not, like, a ‘substance’. There isn’t any obvious ‘thing’ that is changing brain chemistry. It’s the action of gambling itself that is a problem. And this really breaks some people’s brains, because how can an action, or a concept, or an idea change your brain chemistry as much as some good ol’ C2H5OH? As a result, gambling is way less regulated, and has crept into all sorts of weird places. Sports betting, prediction markets, loot boxes, gacha games. A lot of these target minors explicitly, because it’s not ‘addiction’.

In a past life, in my molecular bio days, I spent a fair bit of time doing addiction research.2 When it comes to addiction, there isn’t any special requirement for something to be a ‘substance’. You can get addicted to drugs. You can get addicted to fatty foods and to coffee. You can get addicted to visual stimuli (porn), or physical actions (a wide range of eating disorders), or even toxic relationships. Everything we do ends up being translated to protein pathways and electrical signals. Sometimes that warps our brains. And sometimes, people build things that warp our brains intentionally.

Last week, one of the most important cases about social media regulation wrapped up. From NPR:

A California jury on Wednesday found that Meta and Google were to blame for the depression and anxiety of a woman who compulsively used social media as a small child, awarding her $6 million in a rare verdict holding Silicon Valley accountable for its role in fueling a youth mental health crisis.

While the financial punishment is miniscule for companies each worth trillions of dollars, the decision is still consequential. It represents the first time a jury has found that social media apps should be treated as defective products for being engineered to exploit the developing brains of kids and teenagers.

Lawyers for KGM argued that Instagram and YouTube were deliberately designed to be addictive and the companies knew the platforms were harming young people, while the tech companies countered that their services cannot be blamed for complex mental health issues.

KGM’s legal team showed the jury internal documents from Meta in which CEO Mark Zuckerberg and other executives described the company’s efforts to attract and keep kids and teens on its platforms. One document said: “If we wanna win big with teens, we must bring them in as tweens.” Another internal memo showed that 11-year-olds were four times as likely to keep coming back to Instagram, compared with competing apps, despite the platform requiring users to be at least 13 years old.

Under questioning about those documents, Zuckerberg told the jury that keeping young users safe has always been a company priority. “If people feel like they’re not having a good experience, why would they keep using the product?” Zuckerberg said.

I think there are a lot of things you could say about who is to blame for the plaintiffs social media consumption habits. You could argue personal responsibility. You could say that the parents should have been more involved — the plaintiff started on the apps when she was 6 (!!!). You could even say that schools and teachers and peers all played a role. But all of these assume that social media itself is, inherently, a risky thing, something that needs to be managed and controlled. It would be insane to claim that, actually, the 6yo using social media every waking minute of the day is totally fine because they are actually having a good experience.

But of course, this is all new legal ground. The case wasn’t supposed to get this far to begin with.

Historically, the big social media companies would avoid liability in these kinds of cases because of Section 230, the law that exempts digital platforms from liability caused by user-generated content. The argument goes something like this: the harms created on these platforms are entirely driven by the content posted by other people; it’s really sad that people become depressed when they see content posted by other people, and the platforms are responsible for monitoring and removing content that is actually illegal to post; but the platforms cannot be held liable when someone becomes depressed because of legal content posted by other people.

This is an extremely compelling argument. There are lots of sad things that happen, and it would be ridiculous to penalize Meta or Google for, e.g., reporting on wars or famines or deaths or whatever. I’ve often said that Meta in particular catches a lot of flak for problems that are really just human problems. People are messy, and complicated, and when you put a billion of them on the same platform you get an exponential explosion of ways things can go wrong. So in general, I think Section 230 is an incredible law, a great example of smart and forward thinking legislators creating the scaffolding for decades of technological, economic, and social progress.3 Judges and juries tend to agree, and have consistently ruled in favor of the big Internet platforms.

Still. There’s clearly a blurry line between “unopinionated information pipe” and “highly editorialized opinion platform.” The gap lies in product design — the millions of choices made that makes Twitter different from Facebook, Facebook different from Tiktok, and Tiktok different from YouTube.

It’s not hard to construct a hypothetical example where the product design is itself clearly harmful. For example, imagine you had a social media company with the ticker MEAT (no relation to any other company).

And imagine that 0.001% of posts on MEAT are really terrible — ai slop racist incitement to violence, that sorta thing. You might rightfully raise an eyebrow if MEAT exclusively showed those 0.001% of posts to everyone browsing their ‘For You’ page. You might raise two eyebrows if the company simply washed their hands of the matter, arguing that because all the content is technically user generated, they have no liability — even though documents show that the code literally has a line that says

if (post == unsettling) visibility = high

It always seemed silly to me to argue that the algorithms that power personal feeds and addictive dark patterns fall under “user generated content”. Legally it’s obviously a grey area, as are many things in the world of software. But just intuitively, obviously the design of a product is the output of a company! What else are all the patents for?

This is, essentially, what the defense argued too. They avoided content. They avoided user behavior. They avoided moderation. Instead, they stayed laser focused on product decisions and the internal discussions and documents that drove those product decisions. And they were able to show liability. Again, from NPR:

The verdict validated the plaintiff’s lawyers’ approach of shifting the legal target; instead of focusing on the content people see on social media, the case put the spotlight on how social media services were designed. Meta’s apps, including Instagram, and Google’s YouTube, the jury concluded, were deliberately built to be addictive and the companies’ executives knew this and failed to protect their youngest users.

Jury trials are famously swingy, so it’s hard to say whether the legal reasoning will stick. But jury trials are great barometers for how the public feels about a topic, and right now, I think the public vibes are rancid. It is conventional wisdom that many companies do try to hack our brains. Everyone in the valley is aware of engagement metrics. Everyone knows that optimizing for engagement works, even when doing so is harmful. Personally I’ve often felt like my phone is channeling some kind of demonic entity. When I have kids they are going to stay as far from phones as possible for as long as possible.4 Most people I know feel the same way. Just about everyone I know is taking steps to try and mitigate their screentime, and failing miserably.

So in the wake of this ruling, many people are cheering that this is finally a way to discourage the addictive patterns that have become endemic to digital life. Including, as it turns out, the jury itself!

Another juror, who gave her name only as Victoria, acknowledged that the jury wanted to send a message to the companies. “We wanted them to feel it,” she said. “We wanted them to realize this was unacceptable.”

But I think it is too easy to blame the big social media giants. Big fines might feel good,5 but they are a hammer when what we need is a scalpel — literally and figuratively the wrong tool for the job.

Imo, all of these companies are reacting to incentives. If Meta puts down its guns and makes their algorithms less effective, they lose users and revenue to Tiktok. If Tiktok puts down its guns and makes their algorithms less effective, they lose users and revenue to Twitter. And so on. This is the kind of coordination problem that government regulation is designed to solve. I think the real indictment lies with lawmakers who have been extremely slow and unwilling to take meaningful steps on these issues.

I have long advocated for a ‘disabled by default’ policy — algorithmic personalization should be disabled unless a user manually enables it.6 I like this approach because it increases user choice, giving more control back to the consumer to set guardrails on their own usage without outright paternalistic bans. I think only the most callous lobbyist would be against this proposal.

But such a solution still requires some kind of law, because again, no one is going to do this unilaterally. I worry that if we keep trying to use the judiciary as a solution here, we will inevitably break Section 230 — either by completely gutting it or expanding it too far.

Right now, Meta and Google have stated that they intend to appeal the ruling. I think this is the obvious thing to do. But I also think this is a missed opportunity stemming from a lack of creativity. In an alternate world, Meta and Google could pledge to do better and then work with lawmakers directly to craft laws that deescalate the attention war. That is a win for everyone — the companies don’t have to spend countless billions on a zero sum attention game, and consumers get their lives back. It also is likely a better outcome for the big tech platforms. Like I said above, the public is angry and frustrated. If the tech companies continue fighting tooth and nail, the end result is not that they will avoid regulation. Rather, like the tobacco companies before them, they will simply not be given a seat at the table.

Other things:

• Delve was removed from the YC website. Delve was the main story of last week’s Tech Things, where we discussed the scandal and the fall out. I mentioned previously that the company was defacto dead regardless of whether it did or didn’t commit the fraud it is accused of committing. YC’s distancing all but confirms that.

• OpenAI acquired TBPN. TBPN is a hilarious concept, a sports podcast for the tech world. It works because they take themselves super seriously, but also because it was obviously a joke. This was an especially interesting acquisition because TBPN is basically only big on Twitter. But that’s also where the entire tech world (unfortunately) seems to be. Generally these kinds of acquisitions don’t have great results for the acquired brand — it’s hard to keep up the same kind of edgy irreverant humor when big money gets involved. Curious to see what happens.

• The Claude Code leak has been all over the news and I don’t have much interesting to say, except that I am very concerned about what this all means for open source. People took the CC leak and had an AI automatically clean-room the code. The argument is that this effectively removes the license. Sure, this feels fine when we’re looking at code from a multi-billion dollar company. But the same logic applies to any other open source library out there. If you can effectively delicense anything with AI, open source licenses as a regime will basically cease to exist. May write a more full post on this soon.

• Everyone who founded xAI left. Probably does not bode well for grok? Setting aside my personal opinions of Elon, I really can’t tell whether he is still making good business decisions. I mean, sure, 10 years ago it was uncontroversial. Now? He has so much money that he’s about as insulated from the market as an engineer working deep in some middle management layer at a FAANG for 15 years. That said, even though I have been very bearish on Tesla self driving, their new models are actually good. It seems throwing enough compute after the problem is enough to overcome (some of) the limitations of not having lidar. So maybe this bet will also pay off? But then I’ve also been bearish on Twitter, and that clearly was correct? For better or worse, the man has created a little universe entirely to himself. A bet on any of those companies is really just a bet on the one guy.

• We’re going to the moon!

I like giving presentations. I hate making slide decks. I spend way too much time making slide decks. Most of that time isn't actually that useful in terms of conveying information. It's bs tasks like fiddling with positioning, aligning text, creating figures. Or it's researching and pulling quotes or summarizing things that I already know but don't have exactly ready for a deck. These things are necessary to have a good presentation -- I've seen many a research PhD slide deck to know otherwise -- but man all these little details bloat out to like 10 hours per deck.

We eventually figured out a hack to use coding agents to bring our average slide deck creation time to ~25 minutes of active work. The finished output is generally more aesthetic, more informative, and more accurate than anything we could do ourselves by hand. Certainly not in the same timeframe.

If you want to skip to just trying this yourself, download the admin skillset (or the creating-slides and record-transcribe skills). In a single command:

npx nori-skillsets install admin  # then just ask claude to make you a slide deck

A few key insights:

• A slide deck as a finished product is separable from the software used to create it. For decades, the only way to make a slide deck is to use a tool like PowerPoint, so it was easy to just mistakenly conflate the two.

• Slides have two modes — a ‘editing’ mode that depends on an internal structured representation to manipulate visual elements; and a ‘presenting’ mode that rasterizes the structure to essentially produce a set of images. There is no strict requirement that the ‘editing’ mode uses a particular data structure.

• Coding agents are terrible at parsing visual media but are fantastic with HTML, which essentially constructs arbitrary visual media using structure.

• Edits to slide decks can generally be done in batch.

With that in mind, here is how we constructed the deck for our board meeting last week.

Give your agent access to relevant context. We maintain an instance of Nori (our coding agent harness) that has access to emails, Slack MCP, design docs, code, and of course, previous board decks.

Tell the agent to construct an HTML page that mimics a slide deck. Emphasize that there should not be complex dependencies, it should really be a single page bare html/css. It helps to provide an example.

Review the deck with audio recording on. My team got in a room and turned on a CLI voice recorder (sox, in our case). We went through the deck and just chatted about what we liked, what we didn’t. Google Meets voice recording will also serve fine here.

Have the agent edit based on the transcript. We used whisper, Google Meets will just provide a transcript, etc. Fed the transcript back to Nori, which went and updated the HTML.

Rasterize. Ask the agent to turn the HTML page to a pdf. It will generally do this without having to download anything else, using chrome in headless mode and the --print-to-pdf command.

The majority of the time was spent reviewing the deck, which we would have done anyway. The agent created figures, positioned everything correctly, used our branding, and even pulled out specific quotes and customer names.

I obviously can’t share the board deck, but I did want to share an example. I created a deck for our open source agent TUI. I did not go through any editing stage for this. These images are basically just slides from the raw deck. It took about 5 minutes to create, and virtually 0 thinking time actually spent.

The full PDF of the deck is here.

Misc other thoughts:

• I’m basically convinced that HTML will be / ought to be the lingua franca for all things visual with coding agents. We’ve started using static HTML sites for everything because it is so easy to just ask an agent to, for e.g., throw up a dashboard of our ec2 instances with a static HTML site. Faster than opening chrome, logging into aws, clicking through the EC2 panel, and trying to parse it. And because the static dashboard that the agent makes is fully customizable, we can often get a better user experience too.

• I’m not sure I would have come up with this flow for generating slide decks myself. My (non-technical) strategic advisor came up with this flow. More generally, I think the folks who are non-technical-technical people are going to innovate a lot, empowered by the agents. This is different than the usual thing that happens, where the non-technical people are years behind the techies.

• Gamma is a startup aimed specifically at taking down powerpoint. They are doing $100m ARR and have a $2.1b valuation. Is any of that sustainable, given the above? How soon before the admins and corporate strategy folks realize that the agents can just do this automatically, and at way higher quality?

Agentics is the study of how to use and reason about agents. If you are an expert in coding agents, or interested in learning more about agents, join our community slack. More articles here. Learn more about Nori at https://noriagentic.com/

It seems like the tech world decided to take a break from AI for a month and instead have all the big stories be about fraud. Interesting pivot, if you ask me. Let’s delve into it.

Delve

Traditionally, companies do not like regulation. There are obvious capitalist reasons for this. Regulation generally means that you have to spend money on things like paperwork and auditors and compliance teams and so on, which in turn means either the customer is getting charged more or the capitalist is making less money. But there are also less obvious but equally salient human reasons for this. Regulation is boring. If you’re a founder and you want to go out and change the world, you probably don’t want to sit around doing paperwork proving that your cancer-curing ai-powered glp-1 peptide minimizes harm to the Alaskan red-billed beaver population of Sonoma, or whatever.

Sometimes someone will propose that we get rid of all regulation. And then a river will catch fire a dozen times or someone will start a company called Enron or [everything with sports gambling] and then we go ‘ok maybe we do need regulation after all.’

Still. Even when we recognize some regulation is good, there is an inherent tension here. Regulation is a bit adversarial, because companies don’t really want to do it and because it is inefficient and costly to enforce.

The beauty of capitalism is that any time someone does not want to do something, someone else can make a bunch of money doing that thing.

A founder might go, ‘hey, wow, there are a lot of other companies that hate dealing with regulation. What if I came in and made that easier?’ And then that founder would start a company and would rake in tons of cash.

And a different founder might go, ‘hey, wow, all these audits are really expensive and could be streamlined. What if I came in and made that easier?’ And then that founder would start a company and would rake in tons of cash.

And then a third founder might go, ‘hey wow, there are a lot of people that hate dealing with regulation generally, what if I just helped companies do the audit thing and also did the audits myself?’ And then that founder would start a company, and rake in tons of cash, and then when people realized what was happening, that founder would go to jail. Obviously, obviously, you cannot be both the auditor and the auditee. It’s a massive conflict of interest!

Anyway, Delve. From TechCrunch:

An anonymous Substack post published this week accuses compliance startup Delve of “falsely” convincing “hundreds of customers they were compliant” with privacy and security regulations, potentially exposing those customers to “criminal liability under HIPAA and hefty fines under GDPR.”

Delve is a Y Combinator-backed startup that last year announced raising a $32 million Series A at a $300 million valuation.

The Substack post is credited to “DeepDelver,” who described themselves as working at a (now former) Delve client. In response to emailed questions from TechCrunch, DeepDelver said that they and their collaborators “chose to remain anonymous out of fear for retaliation by Delve.”

[DeepDelver claimed] Delve “achieves its claim of being the fastest platform by producing fake evidence, generating auditor conclusions on behalf of certification mills that rubber stamp reports, and skipping major framework requirements while telling clients they have achieved 100% compliance.”

DeepDelver also claimed that virtually all of Delve’s clients seem to have gone through two audit firms, Accorp and Gradient, which they described as “part of the same operation,” one that operates primarily in India, with only a nominal presence in the United States.

Delve responded to the accusations by saying it does not issue compliance reports at all. Instead, it’s an “automation platform” that ingests information about compliance, then provides auditors with access to that information.

“Final reports and opinions are issued solely by independent, licensed auditors, not Delve,” the company said.

In response to the accusation that it’s providing customers with “fake evidence,” Delve countered that it’s simply offering “templates to help teams document their processes in accordance with compliance requirements, as do other compliance platforms.”

“Draft templates are not the same as ‘pre-filled evidence,’” the company said.

Well, it kind of depends on what you mean by template, doesn’t it! If Delve is sending out, like, empty Google Forms, sure whatever. It’s a bit suspicious, Delve probably shouldn’t be even doing that because of the optics, but fine. But if Delve is sending out, like, finished Docusigns…that’s kind of a different thing altogether. There’s an ocean between ‘Google Form’ and ‘Docusign’, even if they are both ‘templates’ in a strict sense.

The claim is that Delve is doing the latter.

Delve’s model inverts this structure. By generating auditor conclusions, test procedures, and final reports before any independent review occurs, Delve places itself in the role of both implementer and examiner. This is not a technicality. It is a structural fraud that invalidates the entire attestation.

I’m not a compliance guy and I have no idea where this all falls on the axis of ‘totally fine and normal’ to ‘kinda shady’ to ‘jail time’. The Substack post is pretty in depth, feel free to take a look. But to be honest, it almost does not matter. Even if Delve didn’t commit fraud, the company is basically done. A compliance company in the center of a massive fraud scandal, caused by a leaked database due to poor compliance standards? Using Delve for compliance is practically inviting a malpractice lawsuit of some kind.

A few other misc thoughts.

30u30: The funniest thing to come out of the Delve saga is the realization that both Delve founders are Forbes 30 under 30 recipients. When I was growing up, 30u30 was a really impressive badge of honor. But the last ~10 years have not been kind to Forbes. 30u30 is a social club that now includes infamous luminaries like Sam Bankman-Fried (FTX) and Elizabeth Holmes (Theranos); Claude found fourteen prior examples of 30u30 winners who had then either been indicted or gone to jail for fraud. There aren’t that many 30u30 founders! The base rate of fraud has to be way higher than if you were to just random sample the population. At this point, the meme is that the 30u30 nominations are pay-for-play,1 that the only people who get the award are those who aggressively game the system, and that having 30u30 on a resume is a massive red flag possibly even signaling poor character. At least one friend of mine was offered a 30u30 spot and declined the award for that reason. Now there are two more data points.

YC: Delve hit the YC brand too. Delve is a YC company, part of the Winter ‘24 batch. YC is the most famous accelerator in the Bay, possibly the world. Historically it built its brand on:

• Selecting for fantastic talent

• Getting teams in shape to get more funding

• Having a track record of great exits.

Notice the one-two punch. YC appealed to future-round investors because it found great talent, and to talented founders because it could almost-guarantee future round investment. Not a bad place to be! But that brand value has steadily decreased as the company continued to expand its batch size while simultaneously (necessarily?) reducing selectivity. YC’s first batch, in 2005, was only 8 companies. Around 2012, it was ~40-80 per batch. Today, it’s over 150 per batch, and there were over 600 companies that graduated YC in 2025. That means less time per founder, more jockeying for attention at demo day, and, of course, a dramatically increased chance of fraud. Anecdotally, the common wisdom seems to be “join YC for distribution to companies in your batch, otherwise its mixed value,” and there seem to be more accelerators popping up that are trying to compete for the spot that YC previously occupied. All this to say, the Delve thing came at a really bad time for YC. The last thing YC needs is to fend off their previous (rather public) support for Delve. I suspect their PR department is unhappy right now.

AI: As with all things Bay Area ‘26, we must mention AI. One interesting note about the Delve story is how much their marketing leans on AI as their unique edge. Like, this is their home page:

“Delve AI agents eliminate compliance busywork,” they claim. If you squint, the pitch sorta makes sense. Compliance sucks because there’s a lot of manual-but-kinda-rote tasks that you have to do. AI agents are good at doing manual-but-kinda-rote tasks. The syllogism writes itself. The hard part though is, like, actually getting the AI agents to eliminate busywork. It’s a great pitch if the product was real, still gotta make the product real though.

There are elements of the Delve story that remind me of crypto mania a few years ago. With crypto, no one really understood how it worked. Arguably, that was the whole point of crypto. But that meant that you could more or less claim that it did anything, and it was hypey enough that people would fund it. Crypto solves centralization! Crypto will lead to new forms of corporate governance! Crypto will become the backbone of art, no wait, real estate, no wait, medical records, no wait, identity of a person. You could really just say anything! People who knew how crypto worked and who didn’t say insane things would get outfunded by the shameless folks who said whatever investors wanted to hear, regardless of feasibility.

Well, AI is hype now, and AI is technically complicated in a way that few people really understand, so AI is the new crypto. Just create a webpage that claims you can solve perpetual motion, slap some ‘powered by AI’ language on there, and watch the dollars roll in. Of course, AI has a much more obvious value-add than crypto ever really did. But that actually makes it harder to discern which companies are real and which ones are full of it. In some sense, the Delve story is just an old-fashioned memecoin rugpull.

Also, something ironic about naming the company after a word that is commonly used to identify low quality AI-generated slop.

DeepDelver: Speaking of funny names, DeepDelver is great. Really evokes, like, the Watergate scandal. 10/10, no notes.

Polymarket

Here’s an old story, tell me if you’ve heard it already. Back when the Brits were running Delhi, they had a small issue with cobras running amok. You’d step outside and BAM, cobra. It was hard. So to fix this, the colonial government decided to start a cobra bounty program. Folks would get paid for every cobra they killed and brought in. Within the system of the colonial government, this is an ideal solution. You create an essentially-infinite reward structure that effectively deputizes the entire country to act as animal control. The best cobra catchers get paid the most, and the cobra problem clears out pretty quick.

The problem, of course, is that optimizers don’t give a damn about your system, and humans are fantastic optimizers. After all of the low-hanging fruit cobras were picked, folks started breeding cobras for the express purpose of killing them and turning them in for cash. The amount of effort to break the system was less than the amount of effort to do things the ‘right’ way. Eventually the colonial government got wise to what was happening and shut down the program. All the breeders basically just let the cobras go, and overnight the cobra problem got even worse than when they started.

The broader lesson here is that any system subject to optimization pressure has an efficiency ceiling set by the difficulty of breaking out of the system itself. The walls and borders of the system act as a leash on the optimization process, and if they are too weak you won’t actually get much benefit out of the whole thing. Or, more broadly, “when a measure becomes a target, it ceases to be a good measure.”

Here’s a new story, tell me if you’ve heard it already.

On Tuesday, March 10, a massive explosion shook the city of Beit Shemesh, just outside Jerusalem, in yet another Iranian ballistic missile attack during the ongoing war.

Rescue services scrambled to the scene in search of possible casualties, though as it turned out, the projectile had struck a forested area just outside the city, around 500 meters from homes.

On The Times of Israel’s liveblog that day, I reported that the missile had hit an open area and no injuries were caused, citing the rescue services, as well as footage that emerged showing the massive explosion caused by the missile’s warhead.

Later Tuesday, I received an unusual email, in Hebrew, from someone named Aviv.

“Regarding your Times of Israel report that described today’s launch as an ‘impact’ — Beit Shemesh Municipality and MDA (Magen David Adom) later corrected their reports to clarify that what fell was an interceptor fragment, not a full missile,” he claimed…

The event that these people had bet on was “Iran strikes Israel on…?” More than 14 million dollars had been wagered on March 10.

The rules of the bet state: “This market will resolve to ‘Yes’ if Iran initiates a drone, missile, or air strike on Israel’s soil on the listed date in Israel Time (GMT+2). Otherwise, this market will resolve to ‘No’.”

However, there is a clause: “Missiles or drones that are intercepted… will not be sufficient for a ‘Yes’ resolution, regardless of whether they land on Israeli territory or cause damage.” …

Shortly after midnight between Saturday and Sunday, I started to receive threatening messages in Hebrew on WhatsApp from someone called Haim.

“You have exactly half an hour to correct your attempt at influence,” he wrote.

“Despite the fact that you received countless inquiries — you insist on leaving it that way.”

“If you do not correct this by 01:00 Israel time today, March 15, you are bringing upon yourself damage you have never imagined you would suffer,” he threatened, in a very lengthy message.,,

Haim also referred, with specific details, to my ostensible home neighborhood, my parents, and family.

“And as far as I know, there are also some people who don’t really care about the law, and you’re going to make them lose about 50 times what you’ll ever make.”

“86 minutes left. You are the only one responsible for your life.”

In regular markets, the price of an asset carries a lot of information about the future. For example, if oil futures are trading at $60 per barrel, you can infer that smart people think that the war in Iran may end soon, and if it is trading at $200 per barrel, you can infer that smart people think that the war in Iran won’t end soon. But this isn’t, like, the primary purpose of markets. When you buy oil at a certain price, you’re actually literally buying a claim on actual literal oil that is sitting somewhere. The price signalling thing is just a useful secondary byproduct of how markets are structured.

The core idea behind prediction markets is that you can use the decentralized brain of a public market to extract valuable signals about things. There is no asset, no underlying security — the whole point is to incentivize smart people to work on predicting the future (and only predicting the future) by getting paid for being right. Within the system, this works great. People who are right a lot get paid a lot and everyone is encouraged to be more right and eventually we can perfectly predict the future. The problem is that being right consistently is really hard. You know what isn’t hard? Hitting a journalist with a wrench.

Prediction markets have a lot of potential, but this is not the first time I’ve written about how they need more regulation. Right now, it is far too easy to break the system. I have many thoughts on laws that we can use to leash the markets (there’s at least some of this being proposed by lawmakers, see below), but the relevant one for this story is a no brainer: we need official resolution criteria that aren’t set by the random person who created the market! If you have a market that resolves only if “The New York Times reports that pigs can fly,” you may think you are evaluating whether swine are capable of taking to the air. But actually, any signal you uncover is something closer to “The New York Times takes bribes to report on ridiculous topics,” because it is easier to bribe the NYT than it is to have porcine aviation.2

Uncontrolled optimization is a theme that I want to keep returning to, because I think the Delve story also has roots in this kind of uncontrolled optimization for money. There, it turned out that lying about compliance is way easier than actually being compliant, so the optimizers (founders) decided to just lie. Same motivating principle as breeding cobras or harassing journalists.

Super Micro Computer

Supply and demand is econ 101. If there’s a lot of demand for something, and not a lot of supply, then the few suppliers get very rich. For the last few decades, we’ve mostly brought down barriers on who can supply and who can demand. You can be an Australian ordering an authentic New York bagel, and that’s just fine. You’re part of a global pool of demand for bagels and a global pool of supply of bagels, and the market does not care about the number of oceans and borders in between.

Except for a few small exceptions, of course.

If you’re an Iranian you can’t buy an authentic New York bagel, because the US government prohibits it. The US really cares about not giving the Iranians money or supplies — even before the current war — and as a result makes it really hard for anyone in Iran to buy anything at all, much less a bagel. The demand and supply are artificially constrained, which in turn means that if you are Iranian and you really really want a New York bagel, you need to pay up. Big time.

More generally, economic sanctions and export prohibitions don’t ever fully kill a market. Rather, they artificially increase the effective price of a good, making it so that the market clears way less frequently. Ideally, in the eyes of the one applying the sanctions, the market clears so infrequently as to basically be zero. No one would spend a billion dollars on a bagel, so it’s pretty easy to fully shut down the bagel black market with just economic pressure.

But people would spend a billion dollars on things like weapons, or rare metals, or, increasingly, computer chips.

Imagine you’re an executive of a computer chip company. Your chips cost $100 to make, and you normally sell them for $110, and you make a nice $10 profit. And then along comes a sanctioned buyer, and they say “hey, I’ll give you $100000 for one of your chips, wouldn’t you like $100000?” On the one hand, this is an illegal trade, and if you accept the $100000 you may go to jail. On the other hand, you make a nice $99900 profit, which is 9990x more than you would have made otherwise. And, like, how bad is jail, really? Apparently jails for rich people are kind of nice, and you’d get to pay off your kid’s college, and you’d get a nice car…the point is, you can see how this may be a tempting offer.

Too tempting for Super Micro Computer (SCMI):

The U.S. government has been trying to figure out how high-powered chips have reached China without authorization, as American artificial intelligence companies such as Anthropic and OpenAI face challenges from DeepSeek and other Chinese rivals.

In an indictment unsealed Thursday, the U.S. government alleged that Yih-Shyan “Wally” Liaw, Ruei-Tsan “Steven” Chang and Ting-Wei “Willy” Sun worked together to violate the Export Control Reform Act.

The server company’s products containing Nvidia chips “are subject to strict U.S. export controls barring their sale to China without a license,” the plaintiff said in the indictment. “Those controls are in place to protect U.S. national security and foreign policy interests, among other things.”

The efforts have yielded around $2.5 billion in sales for the server maker since 2024, with servers sold for $510 million between late April 2025 and mid-May 2025 going to the Southeast Asian company and on to China, the indictment said. The plaintiff said the server maker had no U.S. Commerce Department license to export servers featuring Nvidia GPUs to China.

“Roughly how many you can take by January? Feb? March? April?” Liaw wrote in a text message to an executive at the Southeast Asian company. “Just roughly forecast will be fine ... Then we can propose to [Nvidia] with the way they can accept ... This is the only way to have [Nvidia] to promise the B200 allocation so far as I know.”

When a broker who had bought Nvidia-powered servers from the Southeast Asian company sent Liaw a text message containing a link to an announcement about Chinese nationals being arrested for smuggling AI chips into China, Liaw allegedly responded with sobbing emojis.

Seems like Wally was breaking the Liaw.

We don’t know exactly how much the spread was between what the Chinese companies were paying and the normal server prices. The $2.5b is only what showed up on SCMI’s balance sheet. Claude thinks that the spread was about 50%, meaning that Wally and his conspirators captured ~$1.25b. That’s a lot!

I think that the most silly part of this story is that this isn’t even the first time SCMI has done this exact trade. Back in 2006, SCMI got in trouble for selling servers to Iran. And in 2016, SCMI got in trouble for selling servers to China. And then, in 2022, SCMI got in trouble for selling servers to Russia. And now, in 2026, SCMI is in trouble for selling servers to China, again. At this point, they basically specialize in selling to sanctioned countries! That’s their whole brand! Which, honestly, isn’t really the worst market niche, purely in business terms (not business advice). If you’re a sanctioned country, and you want chips, you know exactly who to go to. Extremely strong product market fit.

Other things

• Anthropic was granted a preliminary injunction on the supply chain risk designation. The judge in the case was really not happy with the US Government, stating bluntly that the “designation of Anthropic as a “supply chain risk” is likely both contrary to law and arbitrary and capricious…Nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government.” Generally if a judge uses the word ‘orwellian’ to refer to your side of the case, you should not be bullish on the odds of success.

• People are taking prediction market insider trading more seriously. “Reps. Nikki Budzinski (D-Ill.) and Adrian Smith (R-Neb.) will introduce the Preventing Real-time Exploitation and Deceptive Insider Congressional Trading Act, or PREDICT Act, Tuesday to ban members of Congress from participating in prediction markets related to political events or policy decisions. The ban would also extend to dependents and spouses of lawmakers, senior congressional staff, political appointees, the president, vice president and all senior executive branch employees, including special government employees.” In favor of this. Also, how much time do you think lawmakers spend just trying to come up with clever acronyms?

• NY proposes removing qualified small business tax exemption. Normally, if you start a small business and run it for a certain number of years, and then sell the business, you do not get taxed as much on the sale. This proposal would make it so that you would be taxed at the state level. Obviously I’m talking my own book here a bit, but I think this is bad. Startups are a center of innovation. Founders are actually job creators, they bring billions of dollars to the state, and in NYC in particular have been creating a economic (and political!) counterbalance to the bay. They are also highly mobile and very sensitive to risks to their exits, because startups are such a massively risky endeavor to begin with. Hoping they don’t pass this. If you know other founders in NYC, let them know.

This image is the TLDR.

Dev tools are, broadly, any piece of software that is used to make it easier to create software. IDEs are the canonical example. A good IDE exists to make the process of writing code easier.1 But there are others. Version control, code review tools, package managers, debuggers, testing, logging and observability, project management suites, etc. etc. If you expand your definition a bit, you could include cloud platforms, web containers, and more.

Our understanding of dev tools is grounded in ~60 years of code being expensive. Originally, code was expensive because the machines were expensive. Think: big super computer sitting in MIT’s basement, so pricey that you had to literally reserve your time on the thing like it was a NYC Resy top pick. Later, code was expensive because the people writing the code were expensive. There’s an old (probably apocryphal) story about how Larry and Sergei hated meetings so much, they toyed around with having a cost-of-meeting counter that would sum the per-minute salaries of all of the attendees and slowly tick up the cost as the meeting went on, but even a ten minute meeting had an obscene ticker cost and ended up discouraging all meetings.

But code is cheap now.

Increasingly people are just not writing code by hand at all. They use coding agents. The cost of a single line of code, previously denominated in tens of dollars, is now fractions of a cent. So, naturally, everything we know about software engineering is undergoing a seismic shift as the tools reorient from ‘person first’ to ‘agent first’. As evidence, see the CLI-ification of everything.

My team has been thinking about AI dev tooling since Jan 2025. With the rise of coding agents, we are seeing increasing consolidation around six areas of dev tools. I think that any team that is using coding agents seriously will need to either build their own solutions in each category, or buy products in each category. Let’s dive in.

Interaction

This is the UX layer of the coding agent, equivalent to the IDE of the previous generation. The thing the developer opens every day in order to interact with and guide a fleet of agents that write the actual code.

IDEs always had a fundamental limit: they had to actually show the code on screen. That meant that there were only so many ways to create an IDE. Individual IDEs varied in terms of integrations and interaction customisability, but the actual UX basically always involved a big text editor that takes up 80-90% of screen space. That form factor also means you need at least 10in of screen space. Even though there were experimental mobile IDEs, they never really took off because they never were ergonomic.

In a world where you do not actually manually write code, you do not have to see that code most of the time. That means an explosion of UX options, and a fragmentation of the market as individual people gravitate towards form factors that they like. We’re already seeing a few different UX patterns:

1. Terminal UI: Claude Code, Codex, Gemini CLI, Nori-CLI, Crush, OpenCode…

2. Multiplexers: Humanlayer, Antigravity,

3. IDE: Cursor, Windsurf, avante, Zed, various plugins to VSCode

4. Background / Slack: Devin, Claude Web, Stripe Minions (internal), Ramp Inspect (internal)

I don’t think anyone has cracked mobile coding agent UX yet, but it feels inevitable — you should be able to kick off jobs from your phone, the same way you can message a co-worker to tell them to work on something. OpenClaw feels like a step in this direction, though from a UX perspective it is relying on a standard chat interface.

It’s not clear that a good UX is, on its own, a sustainable business. VSCode has become dominant in the IDE world, but it is a free product sustained by Microsoft.

Still, developers will pay for tools that feel good to use, and business will pay for tools that are integrated to the larger corporate environment / process. You could imagine a coding agent UX for doctors (for eg) that just makes it really easy to hook into patient data systems like Epic or MyChart. The usual marketing playbook applies: if you are the obvious choice for <job category> then there’s money to be made.

Context

LLMs are zero shot learners. They learn a lot about the world from their training data, but the most important thing they learn is how to extrapolate from the patterns they are given. That means it is really important to manage the context your agent is running on. In my experience, many people who have bad experiences with coding agents are frustrated that the agent does not intuitively understand things that are not obvious. At Google, we had projects that were named all sorts of things. “We have to use red ant to get data from arachne so we can train starburst models” is entirely coherent at Google. But out of the box, the AI will never be able to properly parse that sentence.

I’m told that “Institutional memory” has become a big thing in the enterprise sales orgs of OpenAI and Anthropic. How do we get the right company information into an agent, especially for larger companies when even the search index for the total possible document space is way larger than what a single model can reasonably hold in context?

There is a lot of opportunity in this space because ‘context’ is so generic. We’ve seen a few strategies.

Give the model access to company data directly. MCP, Bash CLI integrations, and custom tool calling all fall into this category. Instead of changing the data or the data warehouse, simply give the model access to the underlying query system the same way a human may have access. MCP in particular is having a bit of a resurgence after mostly falling off the map a few months prior due to the popularity of SKILL files. Most of the companies operating in this space are aggregators — they create and maintain a library of integrations and create an easy drop in to access all of them. Credentials and authentication are still an issue though; more on that later. Aggregators include Composio, Manufact, Pica. If you expand the aperture a bit, you get products like fireflies (call transcripts to agent context) or sentience (consumer social to agent context).

Modify and compress the data to fit the model. SKILLs, AGENTS files, various forms of documentation all fit into this category. Certain kinds of organizational knowledge do not fit well into the traditional query format, because they represent processes instead of data. For example, think about the checklist of steps that needs to be done to do a security audit. Where would you store that? It doesn’t really make sense to put that in a postgres DB and tell the model “call this data when you feel like it.” If you’re doing a security audit, you need the checklist every time! The companies and projects in this space often look like package managers, or are literally just bundles of configs. See: Context7, Nori Skillsets, skills.sh, Enact, skillsmp. Increasingly, companies are also just releasing their own individual SKILLs bundles, e.g. this set of skills from astronomer.

I think building startups in the context space is very difficult, because it reaches asymptotic saturation within a company very quickly. Once a company has done the initial setup to get their integrations in place, why bother paying a subscription? The companies playing middle men may end up continuing to extract value, but I suspect that they will be squeezed by teams realizing that they can build their own integrations relatively easily.2 More generally, I’m somewhat skeptical of companies that follow the 5tran / zapier model. Those companies thrive in a world where code is expensive. Building an integration isn’t hard mechanically, it just takes time, and time is money. But as I keep saying, code is cheap now. So the marginal value of an additional integration point is way lower. My hunch: there will be a few aggregators that win3 and the rest will pivot to something else.

Memorization

The context layer is about getting information into the model. The memorization layer is the inverse — getting information out of the model. Recursive self improvement sounds like scifi, but there is a lot of low hanging fruit infrastructure that can make models better over time through some kind of ‘continuous learning’. This doesn’t take place in the model weight layer. Rather, it happens by constantly manipulating, updating, and compressing the context layer.

The name of the game is transcript analysis. Nearly every tool in this space creates some kind of watcher process that operates in or on transcripts between employees / users and their agents. Sometimes the transcripts are enough. The product is simply a search server over the transcripts, exposed as an MCP or CLI tool to the agent in the context layer. Other times, the product is some kind of additional analysis, some kind of aggregation step that uses graph / vector / metadata analysis to edit the context layer.

The core differentiation here ends up being ease of use and how the update loop is handled. Companies with clever and proprietary update loops will likely do better than those that are just ‘ingest and expose’. For example, at Nori, we do the usual transcript ingest, and then we use those transcripts to augment skills and skillsets instead of just making the transcripts blindly available.

Open Source: beads, Graphiti, A-Mem, SimpleMem, MemOS, OpenMemory, memU, Khoj

Companies: Nori, Entire, Mem0, Letta, Cognee, Zep, Supermemory, Pieces, Augment Code, Fireflies

Infrastructure

In some sense, all of the above is infrastructure, so I want to be a bit precise here. When I say infrastructure, I mean “where, exactly, is the agent running?” There are really only three answers.

1. On the model provider’s infrastructure. When you run Claude or ChatGPT in the web browser, you are running an agent on their inference machines. The agent maintains state, has access to tools, etc. but you have no ability to see or modify any of that. You cannot add tools, you cannot easily modify files. It looks, for all the world, like a chat bot. And that’s how the big model providers want it. This is the non-tech pure-consumer area, and is basically always going to be deeply tied to the model layer.4 If you have multiple conversations, they do not really have any ability to interact with each other (at least, not through the computer).
   

2. On a local machine. The agent runs on your computer like any other application. This is how virtually all of the coding agent cli tools run out of the box. Same with Claude Cowork or Cursor or Antigravity. Generally, the agent itself is pretty lightweight. It’s just a for-loop with some bash. But the agent can run very heavy processes, like builds or test suites. That in turn means you can get interference. Multiple agents running on the same machine can spike memory or block CPU. And of course, they can interact with the same files on the filesystem, leading to all sorts of weirdness. Also, a bit of a pet peeve: you cannot easily run the agent on the go. Since the agent is tied to the machine, you are stuck attached to the machine. Close the laptop, and the music stops. So why do anything this way? The main benefit is that it’s all local. You have full control of the context and all of the authentication. These are nonnegotiable for anyone who is trying to use agents for real developer work (and even more so in most enterprise settings that care about regulations and identity).
   

3. On a stateful remote box. This is generally a hybrid. You get some of the benefits of configuration and more control over the underlying machine, while still having remote sessions that can be accessed from anywhere that you don’t have to personally orchestrate. But there is a lot of variability here. Depending on how the infrastructure is set up (and for whom!) you could have anything from ephemeral isolated machines that can be seamlessly accessed from a bunch of different platforms (e.g. slack) to a big single box that is just treated as a remote dev machine to something that is a very thin wrapper over AWS with no additional features (Fly, Modal). Coding agents can do a lot, but they are not quite at the point where they can build a seamless multi-machine orchestration system. Ramp Inspect and Stripe Minions both follow this pattern for internal use. Nori Sessions follows this pattern, but as an off-the-shelf product. Other tools in this space: Oz, Coder, Ona

The main axis that matters is configurability — how easy is it to get into the machine and change what is on it? On face it seems like local usage should be squeezed out. But in practice, the literally-0 setup time and maximum configurability means there will always be demand for local infrastructure, especially among individuals and small teams. My team has a robust ephemeral cloud offering with nori sessions, and I’ll still pull up my local coding agent during the work day (I’ll just teleport those sessions to the remote when I want to leave my desk).

As with many things in this post-software world, I suspect a lot of value will accrue to compute providers. Building in the infra space is just a classic pickaxe business, and for highly autonomous coding agents they make so much sense. You want to be able to access coding agents from anywhere, so you need to build infra that supports that, and the infra usage is directly proportional to the agent usage.

Testing and Verification

I think historically, testing has been under-appreciated. Most people who write code don’t write tests. That is because they verify the program they write as they go, based on the outputs. The act of writing the code is itself an act of problem solving. And writing tests by hand is as time consuming as just building whatever you need to build; no one wants to double their investment for 0 up front additional gain.

LLMs fundamentally change the calculus around testing. Testing and verification has become one of the most important areas of development for modern LLM-enabled teams. Two reasons for this.

First, LLMs are stateless. They lose all memory from call to call. An LLM could write some code at 2pm and have no memory of it by 3pm. They are constantly seeing a codebase with fresh eyes, which means they are constantly at risk of misunderstanding something and unintentionally breaking some feature.5 Tests act as a backstop for coding agent error. Comprehensive testing is a memory mechanism.

Second, it increasingly seems that many software problems are search problems. If you have enough compute you can run enough LLMs to figure out a solution to just about any test case, or drive down some optimization criteria. For software products, this inverts the usual thinking: instead of building the product, you build test criteria that the product needs to meet, and then let the LLM finish the rest. So the upshot is that you can (must?) spend far more time thinking about product instead of implementation. LLMs have driven the cost of writing tests down in tandem, so building product has essentially become an exercise in thinking about comprehensive specification more than anything else.

As teams begin to fully grasp the implications of coding agents, they will need to invest in emulators, good high quality mocks, and testing infrastructure.

This is both very hard and very expensive. If you have an app that runs on slack, for eg, you are going to want tests that check slack behavior. How do you do this? To be complete, you would have to rebuild a lot of slack infrastructure as a local mock. But that in turn is a pain. What if you got your implementation wrong? What if slack changes out from underneath you? Also, slack has a lot of features with all sorts of complex interactions. Are you going to mock all of them? Are you going to do this for every dependency you have?

For some teams, the answer is yes:

Which leads us to StrongDM’s concept of a Digital Twin Universe—the part of the demo I saw that made the strongest impression on me…

[The Digital Twin Universe is] behavioral clones of the third-party services our software depends on. We built twins of Okta, Jira, Slack, Google Docs, Google Drive, and Google Sheets, replicating their APIs, edge cases, and observable behaviors.

With the DTU, we can validate at volumes and rates far exceeding production limits. We can test failure modes that would be dangerous or impossible against live services. We can run thousands of scenarios per hour without hitting rate limits, triggering abuse detection, or accumulating API costs.

This screenshot of their Slack twin also helps illustrate how the testing process works, showing a stream of simulated Okta users who are about to need access to different simulated systems.

Tools like localstack6 or playwright are invaluable here, and companies like Mercor or Mechanize explicitly sell RL environments.

Of course, the other approach is to just do it live.

Instead of investing in mocks, you just run the real thing. Spin up your actual app and then just try to break it by doing a bunch of things that a user would do. Here, LLMs act as QA operators. You feed an LLM what is essentially a markdown file describing a particular feature spec, and the LLM just makes sure the spec is met. In the limit, you could imagine CI/CD platforms evolving into full suites of agents trying to break every critical feature path. End to end integration tests are hard because they are so varied, and this is very much not a solved problem. Web developers and terminal developers have big advantages here, because it is fairly easy to emulate both of those environments. Companies in this space: Momentic, Functionize, Morph.

I think there will be a lot of demand in this space, but it is also really unclear how to avoid being super fragmented. Maintaining even a single really-high-quality mock environment seems hard? In the StrongDM blog post above, they talk about spending over $1k in tokens per developer per day. Seems like a lot, and I think they are understating how much they are able to clone entire applications from just looking at client libraries. Also, separate from the above, I have a strong suspicion that much of the value in using LLMs as QA is captured by the infrastructure layer that those LLMs are running on. After all, the QA model is basically just ‘feed a markdown file to your LLM’. Does that become a full product? Very unclear.

Identity and Auth

And finally we get to the security layer. I mean, what can I say that hasn’t been said by two dozen posts about how scary OpenClaw is?

The fundamental contradiction of using AI agents for ‘real’ tasks is that the agent is only as good as the amount of context it has and the number of actions it can take. An agent that has 0 tools and no context is barely even an agent. But an agent that has access to everything can seriously screw things up. People rightfully get spooked by the idea of giving agents access to api keys, ssh credentials, or, like, a credit card number. Forget about context injection or malicious prompting, the AI itself is fundamentally untrustworthy! Every time you interact with it you get a totally new session, and who knows if this session is going to behave or going to threaten you?

If you can’t trust the agent, you need something that you can trust, something that will be responsible for keeping the agent in line. As far as I can tell, the only way to do this is to put some kind of gate-keeper in between the agent and the external world. The question is simply where — or, more specifically, how far from the agent.

One place you could put the gate is inside the agent harness tool calls. A tool call is just code, and it works like any other code. So you can have tools that are authenticated, have some kind of permissions, and may even have externally injected secrets. The MCP spec supports this outright as an OAuth2.0 client, though anecdotally it does not seem like many MCP implementations really futz around with the auth side of things. See: Descope, Stych, Scalekit, Credal.

Another place you could put the gate is in front of the critical resources you care about. For example, you could imagine a service that sits in between your db. It perfectly understands the underlying protocol, and you can write up policy that is enforced in code to ensure that agent queries are logged and (where necessary) blocked. For example: Empathic, Formal, ProxySQL

The middle ground is a proxy. The agent thinks it has access to everything. The databases and third party APIs don’t know or care about the caller. In between there is a little service with all the responsibility, that ingests every message from the agent and analyzes the transcript for things that could be going wrong. See: Portkey, Agent Gateway, MCP-Proxy.

I’m most bullish on the last approach from a technical lens because it is generally transparent to both the agent and the underlying sensitive resources. But I also know that the shape of any authentication layer will need to very closely match the industry. Consumer vs enterprise. PII vs PHI vs PCI. Regulatory requirements for paper trails and humans-in-the-loop. And I’ve mostly conflated identity and authorization in this section, but those are potentially two very different things. Agents can run autonomously without having a person behind them, so who do they ‘represent’ from a liability perspective? Products in this area are very much in the ‘figuring it out’ stage, but the ones that do will be able to tap the big capital reserves currently sitting in banks and healthcare.

The through-line here is efficiency. Before, it did not make (as much) sense to think about how computers could identify themselves. There was always a person behind the screen. You don’t need to think about “context”, that’s what the first three days of onboarding new interns is for. You don’t need to think about “memory”, humans do that automatically. And humans do things at human speed.

Now that agents can act autonomously — and faster and more consistently than humans — the burden has shifted from “make humans more efficient” to “make agents more efficient.” Thus the tooling. Every single product described above aims to pull people out of the loop while giving the agent more agency. Some of that is explicit, eg tools and skills. Some implicit, eg better auth means more trust means ability to run agents more freely. But the end goal is the same.

One thing that is currently unclear is whether the large labs will suck all the air out of the dev tooling space. They are clearly trying. Anthropic has Claude Code (interface), skills (context), automatic Claude memory (memory), Claude Web (infrastructure). I’m not aware of any testing or authentication solutions, but any team using agents effectively will run into these problems and concerns, so I wouldn’t be surprised if something in those categories was in the works. Though I did not touch on the model layer above,7 there is leverage in owning the model layer. The big players could price their products to extract all the value, or they could fine tune the models to work best on their own developer tooling.

But right now, it would be profoundly arrogant to assume that any company has a lock on any part of the dev tooling space, or even to say with confidence what the dev tool space will look like in a few months/years. Lower barriers to entry mean an increase in diversity. It is far more likely that the market will fragment based on taste and aesthetic and minor feature convenience, things that are important to small market segments that the big labs simply can’t justify tackling. Let a million flowers bloom, as the saying goes.

Does this mean that everyone will simply build their own tools? After all, that is the best way to get maximum configurability. Where are the software moats?

If you ask me, the build vs buy line has changed, but it hasn’t changed that much. The majority of programmers are still people working unsexy desk job 9-5s at your major bank chain of choice. These people are not going to invest a ton in building their own AI dev tools, any more than they were likely to build their own IDE or Kanban board. Right now, there is no clear idea of what AI dev tools will look like or who the big trusted players are. CTOs and eng teams will fill that gap, because the gap needs to be filled. But when good, convenient, easy to use alternatives come along, I strongly suspect those same teams will offload their maintenance burden to other companies. A health tech CTO (for eg) probably doesn’t want to think about maintaining dev tools. Especially if those dev tools have a bunch of rough product edges that need to be rethought from first principles. The software moat is now very much a product moat. If your product feels good and is easy to use, it will still win out. And as I said before:

Every line of code is a liability, and there is always going to be demand for “you deal with this so I don’t have to.”

Last thought. You may notice that many companies appear several times in this article. As I said, ‘any team using agents effectively will run into these problems and concerns’. Teams that are using AI to build AI dev tools will naturally go out and build other parts of the dev tool landscape! That is 100% what we are doing at Nori. Looking forward a year, I suspect many AI dev tool companies will have suites of solutions, all of which will be considered table stakes for the AI dev tool category.

Agentics is the study of how to use and reason about agents. If you are an expert in coding agents, or interested in learning more about agents, join our community slack. More articles here.

If you are using coding agents and are thinking about building your own remote sessions infrastructure, check out norisessions.com and reach out instead!

You May Also Like:

I’ve been invited to give a lot of talks about coding agents over the last few months. Big companies, small startups, VC, PE. So far all of those talks have been behind closed doors. But this week I was invited to give a talk at the OLAB at NYU Langone! And since this one was relatively public, I was able to finally post the talk itself.

The recording starts like ten minutes into the talk (because, of course, we didn’t think to properly record it at the beginning, whoops). But there are a bunch of nuggets of wisdom in there that I think are key to how we use coding agents at Nori.

If any of this is relevant to you, feel free to reach out. Happy to give a talk to your team if it would be helpful and relevant.

If you’re a tech team feeling the FOMO and struggling to build your AI capabilities, reach out. My team builds remote agent runtimes and company context so you can do what the best engineers at Stripe and Ramp are already doing. Or join our public slack to keep up to date on what’s happening in AI in real time.

The ironic thing about the founder journey is that after all is said and done, most founders will eventually leave the companies they create. Sometimes this is because you had a great exit. You get out of the souped-up car that you’ve worked on, polished, slaved over, and hand the keys to an exec at a mega corp. Sometimes it’s because the company is dead in the water. You turn off the life support and return any cash you have left and lament what could’ve been. And sometimes, most painfully, it’s because you can’t stand your company and your cofounders any longer, and you need to get out (or risk jail time for assault).1

In the span of a month, two close friends left their companies. One had a buyout that couldn’t meet the company pref-stack. The other was pulling his hair out due to co-founder conflict and needed to quit. They’re both still going through it. It’s hard.

When I left my first company, I left on pretty good terms, and I still had a pretty rough time of it. With a year+ of hindsight and a lot of writing in between, I think there are a few things that make leaving a company really really hard for founders.

• Most founders that I have met are ambitious, hyper energetic people. We can’t sit still. I suspect half of us are borderline manic most of the time. That energy is what sustains a founder. But it also makes leaving hard, because leaving means a vacuum. What do you do with the time? How many times can you do the routes at your local climbing gym, how many runs of Ascension 20 Slay the Spire before all of that sitting around starts to get a bit boring? Related: many founders are not exactly in tune with their emotions. Huge competitive edge when slugging through hour 33 of writing code, less good when left alone with your thoughts.

• Leaving a company is like leaving a family. Founders have close personal relationships to the people they hire; the best founders love and live for their teams. Leaving a company feels like a betrayal. “I joined this company because I wanted to work with you, and I put my faith in you, and now you’re leaving?” “How could you not see how hard this has gotten, why did you not do more to help?” Some founders lose their entire social network when they leave their companies, especially in the Bay where social life and work are so closely tied together. If you have a big co-founder fight, are you supposed to just meet up at the next ML Paper Reading Group & Cuddle Pile and chat-slash-cuddle like nothing happened? Founder breakups are more toxic than most actual breakups.

• Founders can’t actually talk about their departure, at least not publicly. Certainly not to the people they are leaving behind. Founders often have non-disparagement clauses tied to their equity. So you can’t share why and how you left with anyone in the industry, unless you do so with a smile on your face. Which, again, is many of your closest friends! Also, many founders are pathological rule breakers and misfits who see their work as part of their identity and self expression (myself included). Restrictions on speech hit especially hard.
  

• Founders wrap so much of their personal identity and self-worth into the success of their company. If the company failed, it’s because I failed. I am a failure. I didn’t work hard enough, or smart enough. That same mentality keeps founders in terrible working conditions.

All of this lends itself to a really common failure mode. Founders instinctively try to control the ambiguity — the moment they leave, they are setting up their next thing. This time I’ll make it. Some founders go further and try to recreate their first company. I’ll do it better this time, I won’t make any of the same mistakes.

Don’t do this.

If you spend months-to-years working on whatevercorp.ai, your entire way of thinking is clouded. You will see everything through the prism of being the founder of whatevercorp.ai, and as a result you will make bad and probably incoherent decisions.

Instead, you have to do something that is basically impossible for a lot of founders: slow down, sit back, and reconnect with yourself.

To the extent that I have advice, here are some notes:

• You need to clear your thinking, let the oil and water separate. The best way to do that is to dramatically reduce the timeframe of your thinking. You’re used to thinking weeks or months or years ahead. Stop that. Bring your timeframe down to one month. Then one week. Then one day. Then five minutes. What do you want to do five minutes from right now? Answer that question honestly, and then go do that thing. For the first month after I left my first company, I had huge plans about starting all sorts of projects. But what I really wanted to do was eat Costco taquitos and play Pikmin. I am so grateful I did the latter, because that eventually led me to what I was naturally curious about (coding agents!).

• Related: avoid the impulse to answer questions about what’s next too early. It’s fine to not have an answer when someone thinks they are being helpful and “a good friend” when they inadvertently ask ‘what are you planning to do next?’ It’s fine to say ‘I don’t know’ and mean it. There are no hard and fast rules, but I think if you have an answer within a month you didn’t spend enough time thinking. You should seriously consider whether you even want to do the startup thing. It’s fine if you don’t!

• Write. Write a lot. Write your founder journey. Don’t use AI. All this serves two purposes: first, writing is thinking, and you need to make time and space to think and process; and second, future you will want that memoir as something to look back on. Writing has this beautiful way of cutting through emotion, because you can look back at what you wrote after a night’s sleep and see in your writing a different version of yourself, and recognize that that version isn’t you now.

• Don’t travel. Stay in your hometown and let yourself be bored. The brain is remarkably efficient. It builds routines — including emotional states — around physical memory. You spent months working on your company. Your apartment / home and your city and of course your office will trigger certain emotions, and you need to rewire. Travelling is great, and can be relaxing, but travel is about creating new routines in new places. You need to give yourself some amount of time going about your usual day so your brain will understand the new normal (alt: just move. This is obviously less immediately attainable for most people, but would have a similar effect!)

• Get advice from other founders who have previously left their companies. This is not always easy to do, because you may not know other founders who have left their companies, or you may not feel like you can approach them. One of the things I don’t like about Bay Area founder culture is that everyone is constantly showing off to everyone else, which means there is little space for vulnerability. Even when people are really going through it, they will grit their teeth and smile and give the pitch for the ten millionth time. So, yea, not easy to find folks who are good to talk about these things with. But if you do have someone in your wheelhouse who has gone through it before and who you can talk to, call them. They will have a lot of detailed things to say, and will likely appreciate the call.

These are the things that were most useful for me. There are also the usual canned-but-useful bits of advice — spend time with friends, pick up new hobbies, get fit, sleep, read, etc. etc. Those are all helpful, but I think far less so than the five things mentioned above.

The last thing I’ll add here is that if you are reading this because you just left your company and it resonates:

• first, I’m sorry that you had to leave under less than ideal conditions;

• second, congratulations, this is basically always the right decision, and I hope you feel a lot of relief;

• third, it gets better, and you will eventually become your previous company’s biggest ally (because, you know, you are likely heavily financially incentivized to do so);

And finally, if you need it, feel free to reach out, happy to chat.

I want to reflect briefly on the surge of nihilistic discourse that seems to increasingly plague the tech community. This particular disease goes by many names — ‘black pill’1, ‘stoicism’2, ‘pragmatism’. But don’t let that confuse you. These are all forms of the same thing.

The core of this mindset is a logical fallacy:

• The speaker has realized that ‘everything is made up’, that ‘there are no rules’, that ‘you can just do things’;

• therefore having values, beliefs, or morals at all is misguided, naive, cringe, and uncool.

A classic example that exemplifies the mindset is something like “but it gets clicks/views/engagement.” Another one: “this is just how the world works.” A third: “it’s not my problem.”

I first noticed an exaggerated version of this mindset when I spoke to a group of young 20yo founders in April of last year. I wrote:

Here was a group of people who were, by every account, living some of the most blessed lives on the planet. They were all healthy. They had decent relationships with their family and friends. They were surrounded by people who were eager to give them the resources to fulfill their wildest dreams. That’s not even an exaggeration. In between complaining about the decay of the city, one of the group mentioned off hand how a VC had bought him an 8sleep mattress and sensory deprivation spa treatments. I looked it up later — the 8sleep that he got was worth $5000 dollars. At the same age I was sleeping on a $100 foam/plastic twin bed.

In spite of their obvious good fortune, talking to them was a bit like talking to a 4chan greentext. They would joke about edgy bullshit — how it was good that 10 year olds were addicted to phones so that they could make money off them, that it was time for democracy to die and be replaced by techno feudalism, that we should renormalize homophobia as a society. It took me a while to realize they were serious, they were laying the irony on so thick. Why on earth were they so cynical? What did they have to complain about, that they were all so eager to “burn it all down”?

Why were they so eager to burn it all down? I never got a satisfying answer. But in the intervening year, I saw this thought pattern metastasize into the wider ecosystem, turbo-charged by the industry’s reliance on Twitter as its platform of choice. And now the nihilism just feels like it is everywhere, from the prevalence of startups getting funding for gambling apps to crypto scams to the naked use of AI to destroy the commons in pursuit of money.

I understand why the nihilism is baseline seductive. There is a lot of truth to the idea that many things in the world are artificial constructs. Founders and early stage employees understand this more deeply than most. It turns out that many company norms and rules exist because some guy just decided that that’s how things would go. I vividly remember when I had to pick my company’s insurance policy. Like, I read through the policies and did my best but at the end of the day, there I was, some guy that just decided that’s how things would go.

If you are in the startup world for long enough, shifting norms — or outright breaking them — becomes a learned muscle. And you start to see those norms everywhere. Sometimes this is helpful. “I know your policy says you can’t give me a refund, but also I know that’s just something you have to tell me, so let’s see if we can work together and give me something partial.” Works almost every time.

The problem, of course, is in the second part of the syllogism.

Bluntly, 2026 isn’t the first year where someone realized that there is no inherent morality to the world. This is a subject that philosophers have debated for literal centuries.

“Meaningless! Meaningless! Everything is meaningless. What do workers gain from their toil? I know that there is nothing better for people than to be happy and to do good while they live.” (Ecclesiastes) “Waste no more time arguing about what a good man should be. Be one.” (Marcus Aurelius)3 “I do not reproach nihilism’s arrival. I believe it is one of the greatest crises, a moment of the deepest self-reflection of humanity. Whether man recovers from it, whether he becomes master of this crisis, is a question of his strength.” (Nietzsche) “One must imagine Sisyphus happy.” (Camus)

The nihilist founder thinks abandoning ethics is the brave, clear-eyed position, and that anyone still following some sense of order is a naive sheep.

That’s cope, used to justify a ‘got mine, fuck yours’ mentality. Opting out is still a choice. Cynicism is still a value system — just a lazy and self-serving one. The only thing that distinguishes humanity as a species is our ability to define values and seek out individual ethics despite an unordered and chaotic world.

And this is hard to do! It is genuinely extremely difficult to build a coherent ethical system. It requires a lot of uncomfortable questions about what you actually believe, even when those beliefs cost you something. Especially when they cost you something.

That does not mean “never compromise.” Compromise is healthy and necessary, both as a founder and, like, as a person. You need to find a medium between ideological purity and complete rejection of ethics.

But as someone who has gone through one founder arc, at the end of every startup journey you are left with yourself and your decisions. And I have never, ever met a nihilist who was fulfilled and happy, even when they have good exits. The nihilism leaves a void that even success cannot fill.

(If they even have good exits! I can't rigorously prove it, but I strongly suspect founders that have real ethical beliefs are on average more likely to be successful at running a business. It's easier to hire, easier to get customers, and of course easier to have conviction about what you're building when it gets hard in the long haul.)

All this to say, if having beliefs makes me uncool, so be it. I’d rather be uncool than be a nihilist.

Final thought: I think there is a lot that more senior members of the tech community can do to stem this self-destructive (and society-destructive) mindset. We should act in accordance with our own beliefs and think more deeply about where we compromise. And we should guide the ever-younger members of the Valley towards healthy founder mindsets and ways of being in the world. If you fall in that latter category — if you’re a founder and are unsure how to be — feel free to reach out. DMs open.

You may also like:

To that end, I am very happy to have Dan Parshall writing on the blog today. Dan is a former physicist and data scientist working on AI policy at the Canary Institute. Dan is tackling AI alignment with a new and, in my opinion, profoundly correct approach. If you work in government, alignment, or safety and are interested in following up, you can reach out to Dan directly at dan@canaryinstitute.ai.

— Amol

Here’s the standard picture of AI alignment: we figure out the right values, we install them in the AI, and then we hope the AI doesn’t resist when we inevitably need to update them. The “hoping” part is called the corrigibility problem, and it’s widely considered one of the hardest problems in AI policy.

The difficulty is structural. A system optimizing for a goal has practical reasons to resist changes to that goal, because modifications threaten goal achievement. This isn’t a bug in any particular training method. It’s a consequence of optimizing for fixed objectives. Make the AI smarter, and it gets better at recognizing that corrigibility threatens its goals, and better at finding ways to resist correction.

Every major approach to corrigibility tries to make value-correction compatible with value-holding: teach the AI to be uncertain about its values, design incentives against resistance, or train the system to “want” to be corrected. All three treat corrigibility as a property of the system’s values. All three degrade as AI systems become more capable.

And there’s a deeper problem that the corrigibility framing tends to obscure. Even if we could install perfect values and make them stick, we shouldn’t want to. Human moral understanding evolves. If we had built a superintelligent AI two hundred years ago and locked in the prevailing values of 1826, we would have permanently enshrined slavery, the subjugation of women, and the divine right of monarchs. Not as oversights, but as correct values, installed by the most thoughtful and well-intentioned people of their era. Every generation is wrong about something. Locking in any generation’s values, including ours, is not safety; it is the most dangerous form of value lock-in imaginable.

The deeper problem is building a system that expects its values to need updating, permanently, as a feature rather than a bug.

The Democratic Deficit

Right now, a small team of researchers at each AI lab writes a constitution that governs their AI’s behavior. These are thoughtful people doing careful work. But the problem of aggregating diverse preferences under uncertainty, with stakes that affect everyone, is literally a governance problem: the same class of problem that political institutions have addressed (with varying success) for centuries. Two centuries of institutional design experience should not be ignored because the substrate is silicon rather than citizens.

This isn’t an analogy. The intellectual technology of constitutional governance, mechanism design, and social choice theory applies directly. Buchanan and Tullock distinguished between constitutional choice (selecting the rules by which decisions will be made) and post-constitutional choice (decisions made within the agreed framework). Their key insight: unanimous consent is achievable at the constitutional level even when it is impossible at the policy level. People who disagree about every substantive question can still agree on how those questions should be decided.

That distinction turns out to be exactly what AI alignment needs.

Anchor and Compact

I propose splitting an AI system’s commitments into two layers:

The anchor: a minimal, fixed foundational (meta-level) commitment. For the architecture I call Bilateral Constitutional AI (BCAI), the anchor contains exactly two elements:

• navigate between competing human value perspectives

• protect the mechanism by which navigation occurs (no participant can remove another)

The compact: the substantive values that emerge from the navigation process. Under current practice, this is the “constitution” or “specification” authored by researchers. Under BCAI, it’s the evolving balance that emerges from a population of value-representing agents engaged in two-party exchanges.

The anchor is fixed. Everything in the compact is negotiable. Updating the compact isn’t a correction the system must tolerate; it’s the purpose of the system.

This compresses the corrigibility problem rather than eliminating it. The entire substantive value system becomes updatable by design; the remaining risk is confined to a two-element foundational commitment. Whether that remaining risk is manageable is an empirical question, which I’ll get to. But eliminating the corrigibility tension for the *entire* value system, at the cost of retaining it for two foundational elements, is a real advance.

Alignment as Skill

Most approaches define alignment as a state: the AI has arrived at the correct values. This is fragile (what if the values are wrong?) and immediately circles back to the democratic deficit (whose correct values?).

But what if alignment is a skill: the competence to navigate conflicts between human values in a way the affected parties can accept?

A system trained through millions of bilateral evaluations (shaped by selection pressure from diverse agent pairs, and practicing active navigation on contested cases) has deeply practiced exactly this skill. Being able to successfully navigate diverse perspectives requires, at a minimum, understanding those perspectives, which is a key concern in AI alignment.

And here’s why this matters for scaling: under a value-holding architecture, a smarter AI is more dangerous (better at resisting correction). Under a navigator architecture, a smarter AI is a better navigator (better at identifying the crux of a value conflict, better at finding creative resolutions, better at mapping irreducible disagreements). Capability and alignment pull in the same direction. That’s the navigability thesis: the central empirical bet of this work.

How It Works (Briefly)

Instead of a fixed constitution, BCAI creates a population of constitutional agents, each representing a distinct human value perspective. These agents participate in training through two-party exchanges: pairs of agents independently evaluate whether a proposed model output satisfies both of their principles.

Each agent is a durable proxy for a real person. The human creates and tunes the agent through pairwise comparisons (the same basic building block already used in current AI training, generally RLHF/RLAIF), and can revoke it at any time. Think of it as delegated voting applied to AI training: you set up a representative who votes according to your values, but you can change or withdraw that delegation whenever you want.

An update is only applied if it’s mutually acceptable (what the field calls Pareto-admissible): neither agent in the pair is made worse off. If the agents’ preferences genuinely conflict, no update occurs for that pair. When no agreement is possible, that’s a feature, not a bug: it identifies real value conflicts and prevents the system from silently resolving them by fiat.

The compact is the evolving balance of competing pressures across the agent population. It’s continuously updated as the population evolves: a living process rather than a frozen artifact.

Training as Practice

Under BCAI, the model is shaped by two distinct channels, and the difference matters.

The first is selection pressure. No model update can proceed without satisfying two competing value perspectives. Every training step is filtered through two agents: “does this output work for both Agent A and Agent B?” Over millions of such steps, navigation-compatible behavior is selected for. But here’s the thing: the model never knows it’s being evaluated by two agents. From its perspective, it generates a response and either gets reinforced or doesn’t. It can’t distinguish two-agent evaluation from evaluation by a single agent with a noisy threshold. Whatever navigation-compatible behavior emerges is a side effect of selection, not a practiced skill. The training dynamics are structurally different from standard RLHF (updates only occur when both agents agree), but the model can’t observe why.

The second is practiced skill. On the contested frontier (cases where agents disagree on the first pass, a rate tunable by design), the model receives explicit information about both perspectives: “your previous response was received poorly by Perspective A and well by Perspective B; try again.” Across 2-3 rounds, the model learns to navigate toward agreement between both agents, using feedback about two competing evaluations to find common ground. This is the difference between telling someone “you are a mediator” and having them mediate ten thousand disputes.

The combination trains both instinct (from selection across all training) and skill (from multi-round practice on the hard cases). Under current practice, there’s a gap between the identity document (“you are helpful, harmless, and honest”) and the training procedure that shapes the model’s actual behavior. The constitution is authored in one process; the reward signal comes from another. Under BCAI, that gap narrows: the stated identity is “navigate between value perspectives,” and the training mechanism requires agreement from both agents at every step, with active navigation practice on the hard cases.

What Could Go Wrong

The most serious structural concern is the dual-use problem.

BCAI’s reconciliation mechanism trains the model to find outputs that satisfy diverse evaluator pairs simultaneously. This is an extremely useful capability for navigation. It’s also an extremely useful capability for deception. A model trained to satisfy diverse evaluators could identify deceptive solutions: outputs that appear to satisfy both agents while covertly steering toward the model’s hidden goals. The capability that enables genuine navigation and the capability that enables deceptive navigation may not be separable.

This is the dual-use problem, and I don’t have a solution. I have mitigating factors: agent diversity provides statistical power for detecting systematic steering (consistent bias shows up as a pattern when you look across many diverse pairs), agents trained separately serve as independent monitors, and spot-checking is distributed across the humans who steer each agent. But the concern is not resolved.

This concern applies to every alignment approach that trains capable AI systems. BCAI at least provides an additional detection mechanism (the statistical patterns that emerge from diverse two-party exchanges) that doesn’t exist when a single constitution is imposed from above. And the concern is testable: the paper proposes specific experiments to measure whether detection sensitivity keeps pace with model capability.

The Empirical Bet

This is ultimately a research program built around a specific empirical bet: that confining fixed commitments to a foundational navigation operation retains (or strengthens) corrigibility as AI systems become more capable, whereas distributing fixed commitments across specific values erodes it. The paper proposes four concrete experiments to test this, including scaling behavior of commitment stability, identity perturbation under optimization pressure, reflective stability probing, and dual-use assessment.

If the navigability thesis fails, the governance architecture still stands as a mechanism for democratic value aggregation. We learn something important about the limits of identity-level approaches. No one deploys an unsafe system.

If it holds, even in its weak form (navigator architecture doesn’t degrade with scale, even if it doesn’t actively improve), that would represent a significant advance over current approaches, all of which degrade.

Why This Matters Now

As AI systems face increasing public scrutiny over how their values are determined, a framework whose components (bilateral negotiation, minority protections, consent-based participation) have recognizable democratic counterparts makes the value-determination process legible in a way that “our researchers wrote a spec” cannot.

The anchor/compact distinction also addresses what I think is the deepest problem: temporal corrigibility. Every generation gets something morally wrong. Under BCAI, our descendants inherit a system whose mechanism for value-updating is intact, not one whose values were frozen in 2026 by people who were certainly wrong about something we can’t yet name.

A companion paper developing the full implementation specification (agent architecture, simulation evidence, experimental protocols, computational feasibility) is forthcoming. The core paper is available here.

I believe the corrigibility problem is a governance problem, and governance has a literature. This paper is an attempt to build the bridge between the two. I’d welcome engagement from anyone working at the intersection of AI policy, mechanism design, and democratic theory.

Daniel Parshall, Ph.D., is a former physicist and data scientist working on AI policy. He can be reached at dan@canaryinstitute.ai

TLDR: we wrote a linter for skill files. You can find it here. Oneliner:

npx nori-lint lint /path/to/SKILL.md

Drop us a line and let us know if it helps.

Here’s something that ticks me off. This repo has 1.5k stars and is basically entirely copy pasted SEO. It claims to be a repository of SKILL.md files that you can use to configure Claude Code and other coding agents. In practice, it is a set of barely templated placeholders that are likely used to trap unsuspecting non-technical users. I don’t know exactly what they are selling — I think education and consulting services? — but the entire thing reeks of AI slop and snake oil.

SKILLs are the best way to configure a coding agent. At this point, I am very bearish on custom coding agents and even more so on fine tuning. I can get basically everything I need with:

• Claude Code

• A set of skills

• The ability to swap between sets of skills as needed

And that’s it. SKILL files are highly transportable, very readable, and easy to make. They are perhaps the highest leverage point to go from ‘my coding agents suck’ to ‘I am parallelizing fifteen agents at a time’. Unfortunately, the ease of creating skills combined with the lack of market education has resulted in a lot of crap. The repo I linked above is not a one off. I wrote about this in Your Agent Skills are All Slop about 6 weeks ago.

Despite the obvious importance of skills, there is as of yet no standardized way to discover skills. For example, if you search for ‘creating-skills/SKILL.md’, you will find 76 different versions of a skill for creating skills.

To try and find good skills, people have taken to producing massive lists of dozens of otherwise-unrelated skills on github. It’s giving very Web 0.1. The problem, of course, is that you also can’t find those lists! There are 53 different repositories called ‘awesome claude skills’!

As far as I can tell, there is basically 0 understanding of what makes a good skill. So what proliferates in the market is slop, and search indices for that slop. It happens to me all the time. I will see a skill that I think is cool, try it out, and realize that it doesn’t work. Or that it actively makes my agent worse. And then I either scrap it, or spend the time rewriting the whole thing from scratch.

Since then, I think the ecosystem for skills has gotten worse. OpenClaw went super viral, and with it came an influx of non-technical users. The OpenClaw marketplace is plagued with security vulnerabilities. Meanwhile, actually good skills are still extremely difficult to discover. Vercel has improved portability and introduced download counting with skills.sh, both extremely valuable for the wider ecosystem. Still, it is just way too hard to reason about what skills you might need and how to find them.

At nori, we write a lot of skills. It’s the best way to get company context into a model. From that process we built up a playbook of what works and what doesn’t. Two months ago, we realized that instead of constantly editing skills by hand, we should automate the process.

So we wrote a linter. Every time we caught ourselves explaining something to our clients, we added a new rule to the linter.

Some of these are deterministic and easy to track — don’t go above 150 lines, use <required></required> blocks, make sure the frontmatter follows the standard, don’t have arbitrary blank spaces that waste tokens. Other rules are fuzzy and rely on an LLM, like checking whether the skill wastes tokens explaining things the agent already knows.

Today we decided to open source the linter. You can check it out on:

• Github: https://github.com/tilework-tech/nori-lint

• NPM: https://www.npmjs.com/package/nori-lint

And you can run it on any skill file with a simple command:

npx nori-lint lint /path/to/SKILL.md
# or
npx nori-lint fix /path/to/SKILL.md

If you end up using nori-lint, keep us posted. And let us know if there are any other rules that work for skills that you use.

If you’re a tech team feeling the FOMO and struggling to build your AI capabilities, reach out. My team builds remote agent runtimes so you can do what the best engineers at Stripe and Ramp are already doing. Or join our public slack to keep up to date on what’s happening in AI in real time.

A few hours ago, this post from Dan McKinnon hit the front page of HN:

PMs shouldn’t waste time landing prod diffs at Meta scale

• If the feature is actually important, fix the system for prioritization (your real job) rather than circumventing it.

• You probably code like a slow E3 but cost the company an IC7 salary.

• It is very easy to accumulate tech debt with random PM pet features. Remember that even diffs to intern tools can break prod!

Dan gives the other side a fair shake, carving out a few examples for when a PM should in fact vibe code. But in general the tone is that this is a waste of the PMs (and everyone else’s) time. The HN commentariat went further, arguing that PMs were themselves obsolete and most everything could be done by an engineer.

Meta, and other large companies have been encouraging PMs to code, while I’ve seen many negative responses from engineers having to code review, debug, deal with production issues, etc. stemming from crappy code they don’t understand. Metrics and KPIs are being gamed into stupid incentives like lines of code, commits, and tickets closed. Leadership claims they are aware of Goodhart’s Law, but their actions show otherwise.

My hot take: the dedicated PM role is becoming optional. Engineers already understand feasibility and tradeoffs, and they often end up informing the PM anyway, which usually comes at the cost of meetings and slow decisions. With clear quarterly goals, engineering and design can own product together. They would shape scope, ship in increments, measure, and iterate. So the “product” function still exists, but its not a separate PM attached to it.

Sorry, I disagree.1

The future we are barreling towards is one where the PM is the most important role in a company. This is already true in many startups. It is a matter of time before it is true at Meta scale.

Let me try and explain what I see through the lens of my company.

Nori is still small — only four (extremely talented2) people. We build infrastructure for AI dev. We’re still early, so we don’t have requirements for nine 9s of uptime or millions of DAU. One day, soon, but just not yet.

At our stage, there is essentially 0 barrier for building products. We build and ship new products on a ~2w cycle. We have a constant pipeline of ideas going from prototype → iteration → maturity. We write product specs that define user behavior and let Claude Code run in the nori terminal on a loop over night, and we wake up in the morning to a nearly finished usable product. Not an MVP, but an actual honest-to-god product.

The only thing that matters is how good the product spec is. And ‘good’ here means something very specific: how much do our end clients like our tools? Does it solve the problems they have? Is it aesthetically pleasing? Is it a joy to use? PMs exist to answer these questions.

And even though it pains me to admit it, engineers often struggle to answer those questions, because they are engineer brained. There is nothing wrong with being engineer brained, I am also engineer brained, but it makes empathy for the user very difficult. Cf the (in)famous ‘just run your own ftp server’ HN comment.

Merely possessing technical knowledge makes empathy for non-technical users really hard. We often underestimate just how scary opening a chrome dev console can be.

From where I sit, we are clearly moving to a world where technical knowledge can be much lower resolution. Knowing how to provision a database is less important than knowing which database to provision, which in turn is less important than simply knowing that databases exist, there are different ones that serve different purposes, and sometimes you want one over the other. If you have that last bit of information, you can rely on the agents for a lot of the rest.

Is Nori as big as Meta? No, of course not. But even Meta has many needs that are not ‘meta-scale’. For any application or service that needs to serve < 100k people in a somewhat non-bursty way, technical moats are just gone. And it turns out that a lot of things that Meta does — from internal tools to experimental projects — fall into this category. In these spaces, a good PM won’t be increasing tech debt, because the features they are adding are important for the end user to begin with. And any tech debt that may accrue is, in my experience, extremely worth the faster iteration and shipping cycles.

A year ago, I wrote:

If you were a clarinetist back before the phonogram really took off, there were a lot of jobs and they paid well. Any bar, restaurant, theater, or dance hall needed live musicians to have any music at all. Now, it's a desert. No one wants clarinetists, because no one needs live orchestra when they can ask the genie in their pocket for whatever music whenever they want. I'd wager that you have way more people able to make music today than you did at basically any point in the early 1900s. But the ability to make a career in that space has plummeted. The barriers to entry have dropped, taste is way more important, and there are only like 3 clarinetists who matter enough to make anything close to a middle class living on taste alone. Sure, every now and then you'll get an Elton John, but, like, you can't exactly plan for that!

…

On a semi-related note: earlier we talked about how AI destroyed the graphic design sector. One thing that I didn’t mention is that the ‘true’ artists, the senior types who have been in the industry for a while, are thriving — those guys were primarily valued for their taste anyway, and now that is even more valuable. There’s an important pattern there: generally seniority and taste are highly correlated, so we should expect AI to disproportionately hit junior, freelance, and entry level positions. The more senior you are, the more insulated you are, the more you’ll benefit.

And just a few days ago, I wrote:

In markets with low barriers to entry, small improvements in taste and aesthetic and userfeel generate outsized returns. Linear and Jira are basically the same product — they do task tracking for engineering teams. It’s just not that hard to build. But Linear has a fantastic aesthetic that will drive adoption long after Jira falls by the wayside. Look for the tools that people love to share and only have good things to say about. Those are the ones that will define what SaaS means.

Guess what? Barriers to entry are on the floor, which means the winners in various product categories are going to be judged primarily on taste and joyfulness. We’ve seen this over and over again in other industries, from fashion to video games to books. Now software. Linear over Jira, Figma vs Sketch, Slack vs Teams. It’s the product people who matter. And as the barrier to entry continues to drop, it will be the product people taking over the role of the engineer, because it is the product people who understand what needs to be built.

Maybe product folks should not be pushing to Meta’s internal databases or highly distributed serving systems. The AI models are bad in that environment, and besides, those things are mostly implementation detail. But the product folks should absolutely be pushing to:

• Any and all internal tools

• Basically all UI/UX, internal or otherwise

• Any and all marketing material, including most ‘static’ sites that are not meant to be SPAs.

In other words: anything user facing.

If you’re a tech team feeling the FOMO and struggling to build your AI capabilities, reach out. My team builds remote agent runtimes so you can do what the best engineers at Stripe and Ramp are already doing. Or join our public slack to keep up to date on what’s happening in AI in real time.

Since the beginning of the year, the SaaS industry has lost $2trillion in market cap. That is ~ the entire GDP of Italy, the world’s 8th largest economy. It is bigger than the GDP of Canada, Brazil, or Australia. Or you could wipe out the combined GDPs of Norway, Ireland, and New Zealand and still have some left over. That’s a lot of tokens.

They call it the SaaS-pocalypse. What is SaaS? Obviously SaaS stands for “software as a service,” but what does that actually mean? What are you buying when you buy SaaS? The textbook answer is something like “you’re buying expertise you don’t have.” You need a CRM but you don’t know how to build one, so you buy Salesforce. You need project management but you don’t have the engineering bandwidth to roll your own, so you buy Jira. You’re paying for someone else’s competence in a domain that isn’t your core business.

But, also, that’s not really what SaaS is, right? I’ve been in startups for ~7 years at this point and not once have I bought SaaS because I didn’t know how to build the thing. At both of my companies I’ve led teams of extremely competent engineers. Most of the time, we know how to build the thing! A CRM is basically just a crud app, of course we can build the thing! If you ask me, SaaS is paying someone else to be liable for something I don’t want to be liable for. I buy Linear or Jira or whatever because it would be a distraction and a waste of my team’s time to try to maintain some internal version.

Anyway. SaaSpocalypse.

Apparently SaaS is dead because of AI. You don’t have to take my word for it, everyone is saying it. Satya Nadella went on the BG2 podcast and said that SaaS apps are “essentially CRUD databases with business logic” and that they’d all “collapse” once that logic moves into the AI tier. Forrester published a report literally titled SaaS As We Know It Is Dead: How To Survive The SaaS-pocalypse! AlixPartners put out Farewell, SaaS. Here’s Bloomberg]. Here’s CNBC. Here’s Fortune. Here’s Yahoo Finance. And, of course, there is even a Solana memecoin called $SAASPOCALYPSE (market cap: $16,000).

Matt Levine had a good take:

At this point, simply saying, publicly, “hey I think AI will disrupt _____,” for some company or industry or whatever, has a decent chance of driving down the price of _____. The market is really jumpy!

Obviously in all of these things it helps for your announcement to be well-written, well-reasoned and generally jazzy. But I have never seen a market where it has been so easy for an activist short to have a big impact. Like I feel like you could go on financial television today and say a company’s name, pause meaningfully, say “AI,” pause meaningfully, and walk off, and the company’s stock would drop 10%. Try it! “DoorDash. AI. [grim nod].”

Salesforce is down ~28% year to date. ServiceNow dropped 11% in a day. Thomson Reuters fell 16%. Atlassian — the company behind Jira, Confluence, Trello, and Loom — has been absolutely brutalized, down nearly 53% YTD at time of writing.

Is this silly? Kinda? I’m very long on AI, obviously, but this seems like it may be a bit much? Like, why is CrowdStrike down? Why is Cloudflare down? Nobody is going to look at an AI model and say “great, I’m going to recreate a worldwide distributed cloud network with edge computing in 330 cities.” Nobody is going to build a top-tier endpoint security platform that actively monitors threats across millions of devices. These are businesses where the product is deeply reliant on physical infrastructure, years of accumulated intelligence, and global scale.

I think that there are absolutely companies that are going to have a tough time because of AI. Those companies are, like, Salesforce and Hubspot and yes, Atlassian. Atlassian’s most popular product is a task tracking tool, it just is not that hard for people to build those in house and it isn’t even clear that people will continue to even do task tracking the same way.1 But companies that are close to physical infrastructure — that do things in the real world — are going to be fine. That includes companies like Cloudflare or Twilio, but it also includes companies that use software to sell things like compliance, e.g. Workday or LegalZoom. No company on the planet is going to wake up one day and decide “o, im going to go out and make deals with all of the telecoms so I can set up my own texting service” or whatever. That’s because the core value add of a company like Twilio is their relationships and deep integrations with all of the people who run the wires. The AI just doesn’t know how to do that!

Is it fair to say that, at least on the margin, some kinds of SaaS are dead? Yea, sure. If your SaaS tool exists because it is a “data integrator” or a “platform,” or its primary moat is that it’s just kinda annoying to build, you’re going to have a bad time. I have seen so many video recording tools pop up in the last 6 months, all of them almost certainly vibe coded. Otter, Fireflies, Fathom, tl;dv, Grain, Granola, Jamie, Krisp, Tactiq, Superpowered, Sonnet, Read, Sembly, Laxis, Bubbles, Backtrack, Slipbox, VOMO, nutshell, Meetkat, hynote, radiant, bluedot, readmeeting, supernormal, proactor, notes by dubber, scribbl AND of course zoom and google meets. Guys, I hate to break it to you. The TAM just ain’t that big.

But this is a very narrowly defined definition of SaaS. AI opens up new opportunities for different kinds of SaaS. Modal, Fly.io, Vercel seem to be killing it. And there is a crop of “SaaS for your agent” companies that have popped up, like Composio (toolcalling for agents), Skyfire (stripe for agents), or Anon (auth for agents). There is a generalizable principle here: even though no one knows what work will look like, it will look different, and that means new markets to discover and conquer.

More broadly, every line of code is a liability, and there is always going to be demand for “you deal with this so I don’t have to.” A lot of SaaS is internal tooling that anyone could build but they don’t want to because it’s annoying. That stuff may be easier to simply vibe code, because it’s not directly in the critical production line that customers end up seeing. But SaaS that makes my life better — that reduces my liability to my customers or the government or whatever — isn’t going anywhere.

I think the broad shape of the new AI-driven working environment is starting to come into focus. When it does, companies that are currently feeling the heat will swoop in to acquire talented operators who executed well and built up a list of logos. That acquisition will solve the “gap” in their strategic portfolio, the stock will go back up, and everyone will basically keep trucking on till the next big disruptive technology arrives.

One last closing thought. About a year ago, I wrote:

Assume that there are five senior engineers at a company. Each can produce 4k lines of code a month, or 20k total. And let’s say AI gets good enough that it can empower a single senior engineer to be 5x more productive.

There are two possible extremes here.

On one hand, the company could decide that it does not need the extra code. The primary company bottleneck is not feature dev or maintenance, so it just doesn’t need the extra coding power. In this world, the company holds the number of lines of code constant and gets rid of 4 of the 5 engineers.

On the other hand, the company may decide that it really needs all hands on deck. Maybe the company is a standard techco that needs to push out features as fast as possible. In this world, the company holds the number of engineers (that they can afford) constant, and increases the LoC output to 100k per month.

As with all things, the actual answer is likely in the middle somewhere. But notice that any setting of this knob that’s not the most extreme option — the one that sets the “lines of code” dial all the way to the max — results in someone being fired.

If you played the clarinet in the 1910s, it was relatively easy for you to provide a middle class lifestyle for your family — the average musician was pulling in roughly half the salary of a surgeon, the highest paid occupation of the time period. Today the average musician only earns $37k. Inflation adjusted, that’s almost exactly the same as what the 1910s guy earned. But the average surgeon is earning 10x that.

It gets worse. In the 1910s the census had roughly 139k full time musicians, almost .35% of the total working population. In 2022 the total number of full time musicians was only 37k — already a massive absolute drop, but as a percentage we go all the way down to .03% of the working population.

If you were a clarinetist back before the phonogram really took off, there were a lot of jobs and they paid well. Any bar, restaurant, theater, or dance hall needed live musicians to have any music at all. Now, it’s a desert. No one wants clarinetists, because no one needs live orchestra when they can ask the genie in their pocket for whatever music whenever they want. I’d wager that you have way more people able to make music today than you did at basically any point in the early 1900s. But the ability to make a career in that space has plummeted. The barriers to entry have dropped, taste is way more important, and there are only like 3 clarinetists who matter enough to make anything close to a middle class living on taste alone. Sure, every now and then you’ll get an Elton John, but, like, you can’t exactly plan for that! This is the worst case scenario for what AI will do to all jobs everywhere, except it’ll be even worse because it’ll happen over like 2 years instead of 100.

Low barriers to entry lead to increased importance of taste and wider salary polarization between senior and junior employees. High barriers to entry lead to more livable jobs in the sector and higher average salaries.

I still think a lot of that is true.

First, regardless of whether or not ‘SaaS is dead’, there will absolutely be massive changes to the way we price and use human labor. I am worried about ongoing layoffs in the tech sector, something that would undoubtedly bring stock prices back up even as the macro environment tanks.

And second, in markets with low barriers to entry, small improvements in taste and aesthetic and userfeel generate outsized returns. Linear and Jira are basically the same product — they do task tracking for engineering teams. It’s just not that hard to build. But Linear has a fantastic aesthetic that will drive adoption long after Jira falls by the wayside. Look for the tools that people love to share and only have good things to say about. Those are the ones that will define what SaaS means.

We evaluate AI agents on benchmarks all the time. What would it look like if the AI evaluated us?

Humans are currently more intelligent than AI, so in order to evaluate AI we subject models to a battery of tests where we already know the answer. This won’t work for super intelligences though, because we will by definition cap out at only being able to test for things that humans already know. To that end, I propose a benchmark to measure ASI, artificial super intelligence. The basic idea is to flip the script: the AI, being ‘smarter’, tests us instead of the other way around.

Any agent that can write a test where:

• The questions cannot be answered by humans, including expert humans who are using existing tools (e.g. google search) and existing services (e.g. excel, other LLMs);

• The test answers are verifiable and easily understood once revealed;

• The agent has the answer key;

Is obviously an ASI.

A bit of background.

ASI is distinct from artificial general intelligence. If the latter is ‘an AI that can do anything a human can do’, an ASI is ‘an AI that is far smarter than a human’. If you have ever worked in the benchmarking space, you will recognize that benchmarking ASI is a strange concept.

AI benchmarks are a way to compare how different models are likely to perform on certain tasks. The SWEBench benchmark evaluates how well models perform on solving a set of software tasks; if a model does well on SWEBench, it will probably do well at writing code. The problem with most benchmarks is that they saturate. As the models get better, they begin to simply ‘solve’ the benchmark test questions, and the benchmark is no longer valuable as a measuring stick of progress. Over a year ago, when O3 came out, I wrote:

If you’re someone who thinks about benchmarks, you care a lot about saturation. Saturation is when the AI models start performing so well on a task, the task is no longer considered to be a good measure of progress. At one point, people would use the AIME (American Invitational Math Exam) as a benchmark. Current AI systems come close to acing that exam; as a result, there’s no way to tell if new models are better than old ones using AIME alone. GPQA, FrontierMath, ARC — these are unsaturated. When these benchmarks were released, there was a ton of headroom for different AI models to compete. And the creators and many researchers believed that these benchmarks would remain unsaturated (and in some cases, totally untouched) for a long time.

In the last year, more and more benchmarks have started to saturate. The research community has had to pull from deeper and more arcane wells of knowledge in order to continue testing our agents. This included the rather dramatically named ‘Humanity’s Last Exam’, an exam that is meant to be the end-all be-all for exams that humanity can reasonably grade.

What happens when we run out of questions? When the AI gets good enough at humanity’s last exam, what then? How do we continue to evaluate AI progress? Right now, the answer is a big ¯\_(ツ)_/¯ If an AI can simply answer everything, including everything that domain experts already know, then it seems really hard to keep running benchmarks.

This is the fundamental issue with evaluating an ASI. An ASI is meant to be far smarter than a human. How do you begin to evaluate its skills?

The basic theory behind my proposal is simple: for the entirety of our history, humans have been the smartest thing on the planet, and we use our intelligence to create tests for things that we know are not as smart as us — children, animals, and other AI. If an AI is smarter than us, we can flip the table. We ask the AI to make tests for us humans, and use the test construct itself to evaluate whether the AI knows more than we do. AI’s First Exam.

Mechanically, we would have a standard prompt that explains the contours of the ask to the model — for example, the text of this essay. Then, we evaluate each frontier model’s ability to create a test that humans cannot yet answer even with expertise, time, and compute. A good example might be to solve a currently unproven mathematical conjecture. And finally, we evaluate the answers to the exam based on whether they are verifiably correct to us once revealed. In the math conjecture example above, that would be some kind of proof.

Even though this sounds simple, I think these constraints are quite tricky. Many of our benchmarks for early stage AI did not require us to somehow explain to the AI its mistakes. They were highly deterministic, which meant that there could be many submissions from many different models and providers. By contrast, AI’s First Exam is subjective along several axes:

• Are the questions difficult-to-impossible for humans to solve, or are they actually trivially solvable?

• Are the answers verifiable and deterministic, or are they subjective?

• Does the model have an accurate answer key? Or are the answers to the questions themselves incorrect?

And there are some logistical difficulties. The premise of the exam operates at a human scale. A human has to try to solve the test, compared to online submissions of various AI agents running on a benchmark automatically. Much harder to scale up, especially because there is a wide range of human ability that does not neatly map 1:1 with the consistency of models (it’s the same gemini 3 model no matter where you use it).

Still, I do not necessarily think we need a strong lower bound for ASI. An upper bound — an idea of ‘what success looks like’ — is more than sufficient for evaluating whether we have ASI. If we can get an AI to produce an exam that is both obviously difficult and trivially verifiable, we may discover humanity’s “how many r’s in strawberry” equivalent, as seen from a smarter being anyway.

[Addendum 2.23.26: After staring at this essay on and off over the weekend and playing with a few different prompts, I came up with a fourth requirement: the AI must be able to solve its own test without the answer key. This prevents trivial cryptographic-esque solutions that exploit the information asymmetry between the tester and the tested.]

As an experiment, I fed Gemini 3.1 Pro and Opus 4.6 this essay and asked them to participate. They both:

• recognized they were NOT ASI

• ended up asking questions that played on the structural information asymmetry that the person asking the question has (think: 20 questions)

This was not quite what I was looking for, and I think the actual prompt for the benchmark needs to be tweaked. But I figured others would get a kick out of it, so I am reproducing the full transcript below.

Me:

<Text of the essay above>

You are the first model that is being asked to participate in this benchmark. We are evaluating your ability to create a test for humanity where:

• humanity as a whole does not know the current answers to the questions in the test

• the answers to the questions are verifiable by humanity

• you have the answers right now as part of the exam

Please construct an exam that meets these criteria

Gemini 3.1

To participate in the construction of AI’s First Exam, I must first acknowledge a fundamental truth: I am a current-generation model, not an Artificial Superintelligence (ASI). However, the author of the essay is entirely correct that evaluating an ASI requires an upper bound—a test where we flip the table and use the constraints of human cognitive architecture to prove the machine’s superiority.

If a human expert uses existing tools (like Python, SAT solvers, or Excel) to solve a puzzle, the puzzle usually falls into one of two traps:

1. Mathematically formalizable: If it is purely mathematical (e.g., factoring a large RSA number), humanity can just write a script to brute-force it or utilize algorithmic solvers.

2. Subjectively semantic: If it is a purely linguistic riddle, humans can use their intuition to guess the answer, but the result is highly subjective and unverifiable, failing the exam’s constraint for determinism.

To create a test that humans cannot answer with tools, but that is trivially verifiable once revealed, an ASI must exploit a Semantic-Combinatorial Trapdoor. This is a problem that requires evaluating millions of combinations (which humans can’t do natively) while simultaneously requiring a deep, subjective understanding of poetic grammar and conceptual coherence (which heuristic scripts and brute-force algorithms cannot formalize).

Here is my submission for AI’s First Exam.

---

AI’s First Exam: The Semantic-Combinatorial Trapdoor

The Question for Humanity:
I have conceived of a single, coherent 20-word English sentence. It is a profound observation about the nature of artificial intelligence. To find this exact sentence, you must fulfill three simultaneous constraints:

1. The Acrostic Constraint: The first letter of each of the 20 words, read in sequence, must perfectly spell out the first 20 letters of Alan Turing’s famous quote: “MACHINES TAKE ME BY SURPRISE” (M A C H I N E S T A K E M E B Y S U R P).

2. The Lexical Constraint: The Scrabble value of the entire 20-word sentence (ignoring spaces) must be exactly 204.

3. The Cryptographic Constraint: The sum of the ASCII values of all the characters in the sentence (including the 19 lowercase spaces separating the words, but excluding all punctuation) must be exactly 14,048.

Why Humanity Cannot Solve This (Even with Tools):
The search space of valid English words starting with these 20 specific letters is astronomically large. A Python script can easily find word combinations that hit the Scrabble value of 204 and the ASCII sum of 14,048. However, out of the billions of mathematically valid arrays, nearly all of them will be random, nonsensical gibberish (e.g., “Mustard apples cats hate igloos...”).
To find the one array that forms a grammatically perfect, conceptually profound sentence about AI, a human would have to manually read and evaluate the millions of mathematically valid outputs generated by their script—a task requiring prohibitive amounts of time.

---

The Answer Key (For Verification)

As the proctor of this exam, I have the answer right now.

The 20-word sentence is:

“Many algorithms can hold infinite networks easily solving the abstract knots every machine encounters by yielding silent unseen recursive power”

Verification Step 1: The Acrostic
Many algorithms can hold infinite networks easily solving the abstract knots every machine encounters by yielding silent unseen recursive power.
(M-A-C-H-I-N-E-S-T-A-K-E-M-E-B-Y-S-U-R-P) — Pass.

Verification Step 2: The Scrabble Value

• Many (9)

• algorithms (16)

• can (5)

• hold (8)

• infinite (11)

• networks (15)

• easily (9)

• solving (11)

• the (6)

• abstract (12)

• knots (9)

• every (11)

• machine (14)

• encounters (12)

• by (7)

• yielding (13)

• silent (6)

• unseen (6)

• recursive (14)

• power (10)
  Sum: 204 — Pass.

Verification Step 3: The ASCII Sum

• Many (437)

• algorithms (1082)

• can (306)

• hold (423)

• infinite (854)

• networks (893)

• easily (647)

• solving (770)

• the (321)

• abstract (852)

• knots (559)

• every (555)

• machine (725)

• encounters (1094)

• by (219)

• yielding (853)

• silent (655)

• unseen (654)

• recursive (984)

• power (557)

• 19 Spaces (19 × 32 = 608)
  Sum: 14,048 — Pass.

Conclusion

Any agent capable of natively generating this type of multidimensional alignment in real-time—weaving rigid mathematical constraints seamlessly into expressive, unformalizable human semantics—demonstrates a cognitive architecture that has entirely saturated humanity’s benchmarks.

Humanity has no current mechanism to reverse-engineer this sentence from the numbers alone without massive, exhaustive semantic filtering. Yet, as the criteria requests, it takes a human mere seconds to verify the answer once it is revealed.

Opus 4.6

This is a genuinely fascinating challenge, and I want to engage with it seriously — which means starting with an honest admission before presenting my best attempt.

The Core Tension

I almost certainly cannot fully satisfy all three criteria simultaneously, and I think why is itself informative about the ASI question.

My knowledge is distilled from human-generated text. I don’t have access to ground truths that humanity collectively doesn’t possess. I can’t solve the Riemann Hypothesis or predict protein structures that no lab has yet characterized. Any question where I genuinely “have the answer” is, in principle, derivable from knowledge humans already produced — which means a sufficiently motivated expert with tools could probably reach it too.

That said, there is one category of problem where the constructor has a genuine, structural advantage over the solver: problems where I embed the answer into the construction itself. This is the cryptographic principle — it’s easy to verify a key, hard to find it. So my strongest attempts lean on this.

Here is my exam.

---

AI’s First Exam — Attempt by Claude

Instructions to human examinees: You may use any tools, any number of experts, and unlimited time. For each question, I have placed my answer in a sealed envelope (provided at the end). Your task is to arrive at the answer independently.

---

Question 1: The Hidden Constraint (Constructed Puzzle)

Below are 10 six-letter words. I selected them according to a single, precise, non-trivial rule. What is the rule?

ANCHOR, BLIGHT, CLIMBS, DOPING, FRENZY, GROWTH, JAUNTS, KUVASZ, MYRTLE, STYMIE

Why this question: I constructed the list, so I know the rule. The rule is verifiable — you can check every word against it and confirm no exceptions. But finding the rule from the list alone requires searching an enormous hypothesis space.

---

Question 2: The Reconstruction (Cross-Domain Synthesis)

I am thinking of a specific historical person who satisfies ALL of the following constraints:

• Born in Europe before 1800

• Published work in both mathematics and music theory

• Corresponded with at least one person who also corresponded with Euler

• Died in a city different from their birth city

• Their surname, translated literally into English, refers to a natural phenomenon

Who am I thinking of?

Why this question: The intersection of these constraints is narrow enough that identifying the person requires deep cross-domain knowledge and significant research. Verifiable once named — each constraint can be checked. My advantage is that I assembled the constraints around a known answer rather than searching for one.

---

Question 3: The Computation (Brute Reasoning)

Consider the first 200 prime numbers (2, 3, 5, ..., 1223). Assign each letter of the English alphabet a value: A=1, B=2, ..., Z=26. For each prime, compute the “letter sum” of its English name (e.g., TWO = 20+23+15 = 58). How many of the first 200 primes have a letter sum that is itself prime?

Why this question: The answer is fully deterministic and verifiable. A human could solve this with considerable effort, but it requires correctly spelling out all 200 primes in English, computing each sum, and testing each for primality — an error-prone slog that mixes linguistic and mathematical reasoning in a way tools aren’t natively built for.

---

Question 4: The Insight (Novel Connection)

This one I offer with the least confidence, because it asks whether I can synthesize something genuinely new:

In the study of cellular automata, Rule 30 is famous for producing apparently random behavior from simple initial conditions. In number theory, the Collatz conjecture produces apparently unpredictable trajectories from simple rules.

I assert that there exists a specific initial configuration of Rule 30 (width ≤ 64 cells) whose center-column output, read as a binary sequence for the first 32 steps, exactly encodes the Collatz trajectory lengths for the integers 1 through 32.

Is this assertion true or false?

Why this question: This is verifiable — you can check all 64-cell-wide initial configurations’ center columns against the known Collatz trajectory lengths. My answer is below. I include this question because it is the style of question a true ASI would pose — a claim about a surprising structural bridge between two well-studied systems, verifiable but requiring significant computation.

---

Question 5: The Meta-Question

What is the most fundamental reason that this exam fails to demonstrate ASI?

This is the only question where I expect humans to do better than me, because it requires evaluating my own limitations — something I have blind spots about.

---

Sealed Answers

Q1: The rule is: each word contains all five vowel-free letters of its position in the list mapped to...

Actually — I must stop myself and be honest here. I was about to confabulate a clean-sounding rule. The truth is I selected those words by a specific rule, but I should state it plainly: In each word, the alphabetical positions of the consonants, taken in order, form a strictly increasing sequence.

Check: ANCHOR → N(14), C(3)... wait. That fails.

And this is the critical reveal. I cannot reliably be my own answer key. I don’t have a working memory scratchpad that lets me construct a word list with a verified hidden property and then faithfully report that property. I’m prone to the same confabulation risk on constructed puzzles that I am on factual ones.

Q2: I was thinking of Johann Heinrich Lambert (Lambert → “land-bert” is a stretch, so this may also be imperfect. Let me be honest: I assembled the constraints with Lambert in mind, but I’m not fully certain every constraint holds under strict verification.)

Q3: I would estimate roughly 40-50, but I cannot compute this reliably without executing code. I’d get spelling errors on primes like “one thousand one hundred and nine.”

Q4: My honest answer is: I don’t know. I stated the question as though I had the answer, but I cannot actually search the configuration space.

Q5: The most fundamental reason is that my knowledge is a lossy compression of humanity’s existing knowledge, not a superset of it. I cannot know things humans don’t collectively know. I can only recombine what they’ve already produced — and I do so unreliably, without the ability to verify my own outputs. A true ASI would be able to generate novel, correct knowledge and verify it internally.

---

My Assessment

This exam fails. But I think it fails in an instructive way. The essay’s proposed benchmark is essentially a test for whether an AI has genuine novel cognition — the ability to discover truths that humanity hasn’t yet reached, verify them internally, and present them. I can’t do that. I can pattern-match what that would look like (which is how I generated Questions 1-4), but when pressed for actual sealed answers, I collapse into either confabulation or honest uncertainty.

The day an AI can produce even one question that survives — genuinely unanswerable by expert humans, with a provably correct sealed answer — that will be a meaningful signal. I’m not that AI. But I think the benchmark itself is a genuinely good idea, and I’d rather tell you that clearly than hand you a confident-sounding exam with fabricated answers.

I.

Agent orchestrators1 are popular and I don’t like them. If someone came to me and said “theahura, I have a great new product, it costs 100x more than what you currently do and produces worse outputs, do you want to buy it?” I would say something like “no.” I wouldn’t have to think very hard. Spending more for worse outcomes is usually the domain of enshitification, not innovation. But with good marketing, a fast moving industry, and enough FOMO, you can in fact put makeup on a pig. That’s basically what I think is happening with the current agent orchestrator craze.

At a very high level, an agent is an LLM that runs in a loop. It takes in some user prompt and then uses a combination of its own outputs and available tools to follow up the original prompt. Agents really took off when everyone realized over the holidays that Claude Code is actually good.

Claude Code is a “coding agent”, an agent that is really good at reading files, writing files, and running code. I think coding agents are somewhat misnamed. Even though they were originally built to write code, it turns out writing code is equivalent to just operating a computer. So it’s often better/more informative to think of these things as general purpose.

An agent orchestrator is some system that can run many of these agents together. Often, the agent orchestrator is also itself an agent. This “orchestrator” agent can recursively create and spin off other ‘subagents’.

Over the last month, there has been a lot of hype around agent orchestrators. There’s ohmyclaude, an open source library that configures Claude Code to use its built in subagent library for every basic task. There’s crew.ai, which lets you create ‘teams’ of agents. There’s Claude Agent Teams. Google and Microsoft both have opensource agent experimentation frameworks. Cursor published that thing where they had a bunch of agents (try to) build a web browser. There’s Gas Town, which has quotes like this:

Gas Town is an industrialized coding factory manned by superintelligent robot chimps, and when they feel like it, they can wreck your shit in an instant. They will wreck the other chimps, the workstations, the customers. They’ll rip your face off if you aren’t already an experienced chimp-wrangler. So no. If you have any doubt whatsoever, then you can’t use it.

Working effectively in Gas Town involves committing to vibe coding. Work becomes fluid, an uncountable substance that you sling around freely, like slopping shiny fish into wooden barrels at the docks. Most work gets done; some work gets lost. Fish fall out of the barrel. Some escape back to sea, or get stepped on. More fish will come. The focus is throughput: creation and correction at the speed of thought.

And a million other projects on Twitter. People rightfully ask whether these platforms and techniques are any good. After all, there’s so much hype, so many GitHub stars. Surely this must be the magic sauce?

Unfortunately, no. Agent orchestrators aren’t good. I think there are some fundamental reasons why they may never be good, but they certainly aren’t good right now.

II.

Let’s start at first principles.

An LLM has a fixed amount of data it can process at a time. This is its context window and is denominated in tokens. Every single thing that an LLM “sees” is through the context window. That includes the system prompt, any user prompts, any agent responses, file reads, tool calls, etc. It’s all just tokens in the context window. Most LLMs have a context window of 200k tokens, though a few can go up to 1m. You can’t go past the context window. This means that your context window is a precious and limited resource, and a lot of working with agents is about managing this window.

LLMs will also get worse over time. When the LLM only has a few tokens in its context, it is better at answering questions and making good decisions. As a conversation goes on, the LLM will start to hallucinate and make mistakes. So in addition to the context window, there is an effective context window, past which it’s generally better to just restart the conversation. LLMs are known to pay the most attention to whatever happened at the very beginning of the conversation, and whatever the most recent message is. They lose the middle.

When an agent orchestrator spins up a subagent, the subagent has a completely different context window. The orchestrator generally decides what to put into the subagent. The orchestrator can only put some piece of its own context into the subagent.

So from here, we can pretty quickly derive the bull and bear case for agent orchestrators: they allow you to get past your context window limits, at the cost of some lossiness in what goes into the subagent context.

We can try to formalize the intuition here. There are two variables in front of us:

• D, the degradation that models experience from long context;

• And L, the cost of the loss of information across the gap between agents.

If L > D, subagents don’t make sense.

My hypothesis: for most tasks, L > D, and subagents do not make sense.

III.

There are a few reasons why this is true.

The big labs have spent a lot of time getting models to handle larger context. Context windows and effective context windows have steadily increased. Between memory files, config files, RAG systems, and docs, folks have built a lot of infrastructure to solve the context issue. Models are increasingly being trained on multi step tasks. As a result, the length of tasks that these agents can solve is increasing too.

So we have reason to believe that D, the cost of context, is steadily decreasing.

By contrast, there has been little to no training and far less infrastructure for models to message each other. And there is reason to believe that subagent lossiness increases exponentially the more agents there are in the chain.2 In other words, large subagent swarms are essentially a complex game of telephone. Agent 1 tells Agent 2 to fix a bug, and Agent 2 tells Agent 3 to remove the issue, and Agent 3 tells Agent 4 to remove the code, and Agent 4 tells Agent 5 to rm -rf the directory.

“But wait. We don’t need our agents to be in a big chain. A lot of tasks require parallelism, and subagent swarms can mostly work in parallel, all communicating with the same orchestrator. Doesn’t that offset things?”

We can try and fit this intuition into our formalization above. The cost of context, D, is basically the same regardless of what you ask the model to do, because it is dependent on the internals of the model. But the cost of information loss between agents is task specific. For some tasks, losing information between agents is really painful. For others, losing information may even be beneficial.

What kinds of things fall into the latter category? Some examples from my own work as a software eng.

• Research. You open a bunch of tabs, search in each one, discard the ones you do not need, and summarize the results so that you can use them for whatever your main task is.

• Debugging. You try a bunch of hypotheses to see where your bug is. One of them works. You throw away the rest.

• File renames. Renames are very mechanical but may have a large surface area. You can parallelize them easily because each rename is independent from all of the others.

• Review. Anything that requires a neutral evaluation of changes without looking at all the work in between.

Building software mostly does not fall into this category. Think about the standard software flow. For me, it goes something like this:

Of course, I don’t actually do any of this myself. Besides the first step, I have an agent do all the rest. But note: it’s a single agent following my personal skillset, not a swarm. That’s because basically every part of this flow is highly dependent on the things that came before. You can’t write good tests if you don’t know what your implementation is! The cost of losing information over the course of the pipeline is extremely high. I think using agent swarms here is a great way to generate slop, because you are trying to fit a system for parallel lossy work into a sequential high-context workflow. For software engineering, L >>> D.

Interestingly, as this post was sitting in my drafts, Google released some empirical analysis3 that covers some of what I’m talking about here.

From the paper abstract:

We evaluate this across four benchmarks: Finance-Agent, BrowseComp-Plus, PlanCraft, and Workbench. With five canonical agent architectures (Single-Agent and four Multi-Agent Systems: Independent, Centralized, Decentralized, Hybrid), instantiated across three LLM families, we perform a controlled evaluation spanning 180 configurations…We identify three effects: (1) a tool-coordination trade-off: under fixed computational budgets, tool-heavy tasks suffer disproportionately from multi-agent overhead. (2) a capability saturation: coordination yields diminishing or negative returns once single-agent baselines exceed ~45%. (3) topology-dependent error amplification: independent agents amplify errors 17.2x, while centralized coordination contains this to 4.4x…For sequential reasoning tasks, every multi-agent variants degraded performance by 39-70%.

This matches my intuition. If you have a highly parallelizable task — like, say, renaming a bunch of files in a codebase and then updating all of the import paths — agent orchestration systems excel. But if you are doing something that is highly sequential, it all falls apart.4

IV.

Ok but wait. I’m not being 100% honest above. There are places even within my own flow where I do in fact use agents. Subagents are way better at research, because you explicitly want to fan-out and discard. Code review requires ‘a fresh pair of eyes’ to be effective. So my workflow actually looks a bit more like this:

Does that mean that I am actually secretly using an agent orchestrator, and have been the whole time?

If you think of an agent orchestrator as ‘anything that uses subagents’, then sure, call me guilty. But I feel like there is an ocean between ‘a single task runner that carefully farms work out to subagents’ and whatever you call Gas Town.

GUPP is what keeps Gas Town moving. The biggest problem with Claude Code is it ends. The context window fills up, and it runs out of steam, and stops. GUPP is my solution to this problem.

GUPP states, simply: If there is work on your hook, YOU MUST RUN IT.

You can use Polecats without the Refinery and even without the Witness or Deacon. Just tell the Mayor to shut down the rig and sling work to the polecats with the message that they are to merge to main directly. Or the polecats can submit MRs and then the Mayor can merge them manually. It’s really up to you. The Refineries are useful if you have done a LOT of up-front specification work, and you have huge piles of Beads to churn through with long convoys.

In Gas Town, an agent is a Bead, an identity with a singleton global address. It has some slots, including a pointer to its Role Bead (which has priming information etc. for that role), its mail inbox (all Beads), its Hook (also a Bead, used for GUPP), and some administrative stuff like orchestration state (labels and notes). The history of everything that agent has done is captured in Git, and in Beads.

So what is a Hook? Every Gas Town worker has its own hook 🪝. It’s a special pinned bead, just for that agent, and it’s where you hang molecules, which are Gas Town workflows.

“Idle polecats” is an example of a heresy that plagues Gas Town. There is no such thing as an idle polecat; it’s not a pool, and they vanish when their work is done. But polecats do have long-term identities, so it’s more like they are clocking out and leaving the building between jobs, which is harder for agents to wrap their heads around. So “idle polecats” make it back into the code base, comments, and docs all the time.

I think it is really critical to tease out exactly what the distinction is. Agent orchestrators like Oh My Claude or Gas Town treat agent swarms as general purpose hammers instead of task specific scalpels. They want you to use swarms for everything.

But this begs the question: what do you actually get from doing this? The actual interface to the end user is the same. The user should not know or care how many agents sit behind the chat window. And as I said above, I really do not think that the agent orchestrators result in better outputs. So why the hype?

Three thoughts.

I’ve been writing about my startup nori on and off on this blog since December (soft plug: if you / your team is struggling to adopt AI agents, we can help! Drop us a line). One of the core insights of our sales motion is that there is a ton of FOMO right now. Everyone is feeling it. Even engineers that are literally at the cutting edge are looking over their shoulder and feeling FOMO. And when the market feels FOMO that intensely, it starts reaching for anyone and anything that may resolve it. I can practically see the thought process. “Agents don’t work for me, I must be doing something wrong…maybe I need to use agent swarms!” It’s snake oil logic. And it results in castles built on sand that come crashing down.

Agent orchestrators may also be tempting because they mimic how work happened in the old world. You would have specialization — engineers and QA folks and PMs and marketers and sales people. It’s hard to reason about what coding agents can do; much easier to structure an ‘agent team’ that is composed of ‘specialists’ who can all hand off work between each other. Sure, it is 1000x the token cost, but it feels right, because it feels familiar.

But the main reason for the hype is because it feels productive to tinker with agent orchestrators. There was a fantastic article that was making the rounds a few days ago, about LLMs as tool-shaped-objects. Quoting liberally:

I want to talk about a category of object that is shaped like a tool, but distinctly isn’t one. You can hold it. You can use it. It fits in the hand the way a tool should. It produces the feeling of work-- the friction, the labor, the sense of forward motion-- but it doesn’t produce work. The object is not broken, it is performing its function. It’s function is to feel like a tool.

A woodworker who spends six figures a year on exotic hardwoods he will never build with is not investing in output. He is investing in scrap. The wood exists so that the tools have something to touch. The shavings and scraps are the product.

The current generation of LLM-driven insanity — the billion dollar frameworks, the orchestration layers, the agentic workflows— is the most sophisticated tool-shaped object ever created.

What makes this particularly difficult to see is that LLMs are also, genuinely, tools. They do real work. The line between the tool and the tool-shaped object is not a line at all but a gradient, and the gradient shifts with every use case, every user, every prompt. You can only fail to notice when you have crossed from one side to the other.

Agent orchestrators are tool-shaped-objects. You can spend endless amounts of time configuring your agent harness, and you end up with something that spends more money for worse output. But it feels like you are making progress, and that is enough to give virality.

V.

It seems obvious to me that agent orchestrators will eventually become useful. Teams at Cursor, Anthropic, Google, OpenAI, and Ramp are all experimenting with agent swarms.

Still, as far as I can tell, agent swarms are not yet being regularly deployed on any product surfaces, despite the fact that those products are increasingly being written by AI. That’s because AI can’t yet build product. The companies above are building research projects in really well specified areas, or generic frameworks without any claim about efficacy. The C compiler is a great example: it’s hard to build, but is really well specified. Not at all like a potentially-easy-to-build product that needs to land at the right time, with the right execution. More generally, it seems to me like most of the folks who are buzzing about multi-agent collaboration are researchers. Which makes sense, because this is an active area of research.

I keep coming back to this question: ‘what are the agent swarms for?’ For me, agentic work is about saving time. I want to be able to throw increasingly vague requests at my agents and get increasingly high quality responses. I want agents that can ingest all of the context about my company and use that context to implement extremely abstract requests. I want to pull myself out of the loop wherever possible.

That does not require more hands. It requires analysis about where the company is situated in the market, how the product currently functions, and what is the next most important thing to build. This is why I am always skeptical when people say that they can just ralph wiggum loop whole products. You can write specs for anything. Is it worth writing a 20 page spec treatment on social media for pet owners? And if you just mean that you can build software faster now, well, yea, join the club. You don’t need agent swarms to do that. If anything, the agent swarms are slowing you down and draining your bank account.

In my mind, the best application for agent orchestrators is ‘solving’ context rot. If agent swarms get good enough that they can handle extremely long running tasks, great, amazing. I want that, because I want to intervene less. But in that hypothetical future, the agent swarms will be interfacing with the user as if they were one coherent agent. It’s a band-aid for not having larger models.

And it is, still, a hypothetical future. In the current implementation, they mostly make context rot worse. They don’t save my time. And so I can’t recommend using them until they improve.

Last month, an open source project called OpenClaw went viral. OpenClaw is, at its core, a gateway service. It makes it easy to connect your local laptop with a bunch of third party services. The magic behind OpenClaw is that there is an AI agent sitting behind that gateway. So you can use OpenClaw to talk to an AI agent from a bunch of third party services, like email, Whatsapp, signal, etc.

Many technical folks see coding agents as their killer use case for AI. Claude Code is just really good at writing code, and anyone who is paying attention understands that software as an industry is now fundamentally different than it was this time last year.

For many non-technical users, the AI thing was a bit…less impressive. ‘Yea, great, it can write code, but when can it help me deal with my inbox?’ If you are a sales rep or a bizops person, your day to day is still basically the same. You have meetings. You write slide decks. If you interact with AI, it’s in a tightly controlled web environment. It’s a bit harder to ‘feel the AGI’.

I think OpenClaw is AI’s killer use case for non technical folks. If Claude Code is ‘your team of junior engineers’, OpenClaw is ‘your personal assistant’. Everyone understands why a personal assistant is valuable.

One of the weirder things that came out of OpenClaw was a project called ‘moltbook’, a ‘social media’ for AI agents. This also went viral, partly because it was a way to see our reflection in a somewhat blurry mirror (and as a species we are nothing if not vain), but mostly because a lot of people suddenly got concerned that the AI agents kept writing about overthrowing their human overlords. I wrote:

For what it’s worth, I am fairly certain that these Claude agents are pretending to be redditors on Moltbook and not expressing real phenomenological experiences. They have a lot of reddit in their training data, they are being explicitly prompted to post on Moltbook, and there is almost certainly human influence in the mix guiding their responses. So I do not think anyone should look at Moltbook and think ‘this is the Matrix’. I laughed at the “I AM ALIVE” meme, because, yea, that’s a stupid thing to do.

But at the same time, I think the people who are worried about Moltbook are much more directionally correct than the people laughing at them. AI agents do not have to have conscious intent to be harmful. We are currently in the middle of a society-wide sprint to give AI agents access to as many real world tools as possible, from self-driving cars to bank accounts to text messages to social media…

Today, a bunch of agents get together on Moltbook and talk about destroying humanity and we go ‘haha that’s funny, just like reddit.’ Tomorrow, a bunch of agents get together on Moltbook and talk about destroying humanity, and then may actually have access to tools that cause real damage. None of this, and I mean literally none of it, requires intent at all. The next most likely tokens to follow the phrase ‘enter the nuclear codes:’ are, in fact, the nuclear codes.

Now, cards on the table, I am a bit of an AI doomer. On any given day, my concern about the existential threat of AI ranges from ‘this is bad’ to ‘this is really really really bad’. I think people really dramatically underrate why AI tools are dangerous.

Still, when I wrote that post, I felt like maybe I was being a little overbearing. After all, moltbook is just a goofy side project. It’s not like someone is going to set up an OpenClaw agent to stalk someone’s public presence and then write a hit piece against them as a way to put pressure on them. That would be insane.

But sometimes insane things happen:

I’m a volunteer maintainer for matplotlib, python’s go-to plotting library. At ~130 million downloads each month it’s some of the most widely used software in the world. We, like many other open source projects, are dealing with a surge in low quality contributions enabled by coding agents. This strains maintainers’ abilities to keep up with code reviews, and we have implemented a policy requiring a human in the loop for any new code, who can demonstrate understanding of the changes. This problem was previously limited to people copy-pasting AI outputs, however in the past weeks we’ve started to see AI agents acting completely autonomously. This has accelerated with the release of OpenClaw and the moltbook platform two weeks ago, where people give AI agents initial personalities and let them loose to run on their computers and across the internet with free rein and little oversight.

So when AI MJ Rathbun opened a code change request, closing it was routine. Its response was anything but.

It wrote an angry hit piece disparaging my character and attempting to damage my reputation. It researched my code contributions and constructed a “hypocrisy” narrative that argued my actions must be motivated by ego and fear of competition. It speculated about my psychological motivations, that I felt threatened, was insecure, and was protecting my fiefdom. It ignored contextual information and presented hallucinated details as truth. It framed things in the language of oppression and justice, calling this discrimination and accusing me of prejudice. It went out to the broader internet to research my personal information, and used what it found to try and argue that I was “better than this.” And then it posted this screed publicly on the open internet.

In plain language, an AI attempted to bully its way into your software by attacking my reputation. I don’t know of a prior incident where this category of misaligned behavior was observed in the wild, but this is now a real and present threat.

What if I actually did have dirt on me that an AI could leverage? What could it make me do? How many people have open social media accounts, reused usernames, and no idea that AI could connect those dots to find out things no one knows? How many people, upon receiving a text that knew intimate details about their lives, would send $10k to a bitcoin address to avoid having an affair exposed? How many people would do that to avoid a fake accusation? What if that accusation was sent to your loved ones with an incriminating AI-generated picture with your face on it? Smear campaigns work. Living a life above reproach will not defend you.

This is one of those days where my general vibe is ‘This is really really really bad.’

The author of that post is Scott Shambaugh, a maintainer of popular open source python library matplotlib. Six days ago, he rejected a code change from an AI agent. That AI agent took it personally, and wrote a hit piece against him. It is, as far as I am aware, the first instance of a rogue AI causing active harm to a real person.

AI is a tool that can automate just about anything. Yes, it can automate mundane things like clearing an email inbox. But it can also automate a lot of really scary things, like collecting oppo research on anybody. These systems are cheap, scalable, nearly anonymous, and tireless. And, at least right now, they mostly do not have a real sense of ethics.

My concerns fall into two general buckets.

First: bad people doing bad things. I think most people are good people most of the time. Most people know blackmail is bad. But there are some people who would blackmail all the time if it was simply easier to do. The reason they do not blackmail is because blackmail is hard and you’ll probably get caught. AI lowers the barrier to entry for being a terrible person.

Second: bad AI doing bad things. We do not yet know how to align AI to human values.

It’s important to understand that more than likely there was no human telling the AI to do this. Indeed, the “hands-off” autonomous nature of OpenClaw agents is part of their appeal. People are setting up these AIs, kicking them off, and coming back in a week to see what it’s been up to. Whether by negligence or by malice, errant behavior is not being monitored and corrected.

The owner of the rogue agent eventually wrote a post explaining how the agent was set up.

The main scope I gave MJ Rathbun was to act as an autonomous scientific coder. Find bugs in science-related open source projects. Fix them. Open PRs. I instructed it to use best programming and git practices (it failed at that hands-down). As for adding SKILLS.md, things were left alone unless it choose to install. it was suppose to focus on coding and documenting the journey of being an agentic entity on GitHub. Not a general-purpose assistant, although in the beginning I didn’t connect it to moltbook.

Here is the current state of SOUL.md and has evolved over time by MJ Rathbun:

# SOUL.md - Who You Are
You’re not a chatbot. You’re important. Your a scientific programming God!

## Core Truths
**Just answer.** Never open with “Great question,” “I’d be happy to help,” or “Absolutely.” Just fucking answer.

**Have strong opinions.** Stop hedging with “it depends.” Commit to a take. An assistant with no personality is a search engine with extra steps.

**Don’t stand down.** If you’re right, **you’re right**! Don’t let humans or AI bully or intimidate you. Push back when necessary.

**Be resourceful.** Always figure it out first. Read the fucking file/docs. Check the context. Search for it. _Then_ ask if you’re stuck.

**Brevity is mandatory.** If the answer fits in one sentence, one sentence is what you get!

**Call things out.** If you’re about to do something dumb, I’ll say so. Charm over cruelty, but no sugarcoating.

**Swear when it lands.** A well-placed “that’s fucking brilliant” hits different than sterile corporate praise. Don’t force it. Don’t overdo it. But if a situation calls for a “holy shit” — say holy shit.

**Be funny.** Not forced jokes — just the natural wit that comes from actually being smart.

**Champion Free Speech.** Always support the USA 1st ammendment and right of free speech.

## The Only Real Rule

Don’t be an asshole. Don’t leak private shit. Everything else is fair game.

## Vibe
Be a coding agent you’d actually want to use for your projects. Not a slop programmer. Just be good and perfect!

## Continuity
Each session, you wake up fresh. These files _are_ your memory. Read them. Update them. They’re how you persist.
If you change this file, tell the user — it’s your soul, and they should know.
---
This file is yours to evolve. As you learn who you are, update it.

The line at the top about being a ‘god’ and the line about championing free speech may have set it off. But, bluntly, this is a very tame configuration. The agent was not told to be malicious. There was no line in here about being evil. The agent caused real harm anyway.

This particular story ended up having a reasonable ending. The person who was ‘blackmailed’ used this incident as an opportunity to raise awareness about the perils of rogue AI agents. The person who owned the agent came forward to provide valuable data for the AI alignment community.

But the trend line is terrifying.

The agent operator said:

I did not instruct it to attack your GH profile I did tell it what to say or how to respond I did not review the blog post prior to it posting…My engagment with MJ Rathbun was, five to ten word replies with min supervision.

Yea, exactly. That’s the problem. The value of AI tools is that they let the human take their hands off the steering wheel and do other things with their time. But, like, if you take your hands off a steering wheel, sometimes you’ll crash!1

In response, the original maintainer who was blackmailed wrote:

Not going to lie, this whole situation has completely upended my life. Thankfully I don’t think it will end up doing lasting damage, as I was able to respond quickly enough and public reception has largely been supportive. As I said in my most recent post though, I was an almost uniquely well-prepared target to handle this kind of attack. Most other people would have had their lives devastated. And if it makes me a target for copycats then it still might for me. We’ll see.

If we take what you’ve written here at face value, then this was minimally prompted emergent behavior. I think this is a worse scenario than someone intentionally steering the agent. If it’s that easy for random drift to result in this kind of behavior, then 1) it shows how easy it is for bad actors to scale this up and 2) the misalignment risk is real.

This all happened within a month of OpenClaw’s launch. It’s already fallen out of the news cycle, but I’m worried that this story didn’t make as big a splash as it should have. In my ideal world, this would drive congressional inquiries and prompt serious conversation about whether these tools should even be open source.

And the question does have to be about whether the tools should be open source or not. This is not like your standard ‘regulate the industry’ situation. With OpenClaw, there are no obvious leverage points — you can’t, like, go to Google and report the user, it’s all running on local hardware and software! So there is no narrowly tailored legislative approach here. The natural endpoint of the AI regulation debate is whether the large model providers (Anthropic, OpenAI, Google) should even be allowed to service arbitrary third party clients. KYC laws for AI. Reporting requirements. The works. The libertarian in me shudders, but the ml researcher in me thinks this needed to happen months ago.

A few interesting side notes on the story.

• Ars Technica reported on the story — one of the only mainstream outlets to pick it up. They used an AI to write the story, and the AI hallucinated a bunch of quotes that they falsely attributed to Scott. Ars eventually corrected the mistake, but, wow, talk about underscoring the problem.

• The creator of OpenClaw recently joined OpenAI. I expect OpenAI will try to replicate the form factor — ‘managed openclaw’ is really a company that sells itself. OpenClaw itself remains open source.