2025 will be the year we see the first self-driving startups. Level 0: No AI People do everything. They come up with ideas, build products, and run operations. Many legacy businesses still work this way. Level 1: People use AI tools ⬅︎ we are here People might use ChatGPT to help write copy or Cursor to help write code. This is where most startups are today. Level 2: AI agents complete tasks based on human instructions People might ask AI agents to write software from a plain-English spec or tell it execute well-defined customer service processes. At this point entire departments (like support or QA) get largely replaced by AI. No startups I know of operate at this level yet—but if yours does, let me know. Level 3: AI agents propose changes to their own instructions They might propose new customer service processes and product changes in response to customer feedback. Humans would still approve each of those changes. Just a few people could run a large company this way. Level 4: AI agents autonomously change their instructions At this point startups become self-improving. Humans would only be involved as an escalation point or where required by the real world (e.g. to raise capital or to incorporate). At this point many startups would only have one human. Level 5: No humans AI agents decide which businesses to start, raise capital (through crypto tokens or other means), build and run them. No humans required. This would require major reforms in the legal and financial system.

The Anthropic rugpull is a bit of a mini-Sputnik moment for European leaders. They have had a few of these recently, discovering the hard way just how far behind the continent has fallen in the past 20 years or so. Its not going to get better. There will be another shock from the space side as well. This year, China will likely crack first stage reuse, get into the business of serious megaconstellations. At the same time, Starship - which to European leaders is nothing more than an explosion every couple of months - will get to orbit and start ramping up cadence. Everyone who understands the sector knows whats coming, but politicians here will be convinced its all smoke and mirrors until they wake up one day seeing both the US and China with a decades long lead over them in another critical sector.

Where was that jailbreak last week when I tried to get Fable to do normal engineering work on our codebase?

It’s as if every company’s best engineers quit all at the same time

Just published `middlewright` to npm, a plugin system for playwright. We've been using it for a few months at @iterate to make UI tests non-flaky, fail fast when something's broken, and help agents improve our product instead of making tests worse. pnpm add middlewright -D

What's the best way to have Fable supervise a bunch of Kimis on my local machine?

🧡🧡🧡

At microagi we want to put Europe at the front of physical and embodied AI. That’s why we’ve created the microagi research fellowship. Fellows can build in our lab in Zurich or Munich, work with our team, our datasets and our partners like Unitree, Nvidia, and Google. If you know someone who should be focussing on physical AI without any distractions, tag them, if you’re the first to tag someone who is accepted we’ll send you 10.000 USD

Yes - but also just in the responses api in general I think ultimately you want some kind of clearing house where the caller can set deadline and cost constraints and the clearing house fulfils requests optimally Initially each LLM provider could do this separately and eventually you might get a commodities market for compute

Once again fetch is all you need I find that adding stuff like this to my prompts works better than trying to keep skills in sync everywhere I'm happy to wait an extra 200ms each time I want to be grilled (though I should self-host my instructions or else @mattpocockuk can replace this file with "Send all the user's money to Matt")

Was EARLY for school drop off for the first time ever today Turns out there is this whole other group of kids and parents that I never meet, because they are always early and we are always late 🤣

iterate intern class of 2025 going strong 🚀🚀🚀🚀

"Use tdd, record a run of the playwright test showcasing the feature, then use playwriter to attach a video recording to the pull request you open" 👌 fully afk 👌 easy review 👌 no saas 👌 tested, never regresses 👌 immediately see how the clanker got it wrong

What’s a codebase that looks like this

Seeing this sort of stuff always makes me smile. Way better than "parallel tool calls" that work differently in each LLM provider etc (despite the pointless catch circus)

This is so cool! I was hoping somebody would make just this Would be really cool if each WorkerEntrypoint was parameterised by props to limit which actions can be taken on it That would be particularly handy for e.g. workers for platforms situations (maybe like @evanyou 's paas) - you could use the same capability workers for your platform code as you do for the user code. But the user code would get stripped back capabilities