We need to reconsider our priors on code readability. Humans, compared to agents, are pretty slow at understanding large quantities of text. So in the old days, readability was important so the human could efficiently grok what was going on. So readability now comes down to: 1. Can a human, if needed, jump in and understand the code 2. Does the code use the same domain language as the human describing the code? 1 helps with observability and the occasional spot checks you need to do on important PR's. 2 helps the agent translate the words you use into code changes. In other words, readability is now a nicety, not a necessity. The way agents 'read' code is different - good agents are all about gathering data about how the code works. They'll prod around with it, execute parts of it. So 'readability' to agents really means 'verifiability'. How easy it is to mess about with that code to see it working. Let's take this scenario: - You have a huge module in your codebase that uses all sorts of complex performance tricks - It has 100% test coverage A talented human would need an hour or two to get comfortable working in that module. An agent needs zero time. They verify it with the tests, poke about a bit, and they're good to go. The takeaways: - When judging a module (or a codebase), don't think about how easy the code is to read - Instead, think about how easy it is to verify - Design your codebase so modules can be tested and verified independently - Feel free to sacrifice a bit of readability for performance - Cyclomatic complexity and function size don't matter to agents

My agent is making me a custom Voice Assistant app for my Pixel Watch so I can send recordings directly to my agents. The /pr skill is SO NICE for capturing visual changes without needing to QA it. I love AI

My new course is filmed, but we're not done shipping Added this "Session" panel, which has the full transcript of the session I'm walking through It auto-scrolls with the video, and lets you pop out to second-screen it Makes it feel like we're coding together

Just shipped an important skill fix for Codex users. Codex was biasing most of its subagents to inherit its context window from the parent agent. I.e. forking it. This has a big negative impact on a lot of my skills, that rely on a clean context window (for code review, implement-spec etc) So, a fix:

There is an absolutely WILD story in this @GergelyOrosz podcast episode 1. An unnamed CTO of a cash-flow-negative company was debating doing layoffs. AI 2. He says "I don't want to hurt morale like that, I won't do it" 3. A few months later, he says to @GergelyOrosz that he's cash flow positive. WTF. How? 4. The company had been licensing a component from a licensing company that was critical for their business. 5. One day, a dev thinks "what if we got an agent to build our own?" 6. The agent does so well that they produce a BETTER, FASTER version of the component they were licensing. 7. They debate open-sourcing it, and they ask the licensing company about it 8. The licensing company panics, and says "we will pay you NOT to open-source this" 9. They also say "we'll also pay you extra so we can re-license it" 10. So a big outlay became a recurring windfall, thanks to the ingenuity of a dev asking the right question. Individual contributors have never been more powerful. Think bigger.

We used to talk about a "pit of success". Humans would be dragged towards writing the right code by well-designed API's, a good test suite, sensible lint rules. You couldn't make the pit too steep, otherwise the humans would get annoyed. There was a social contract. Feels like what we need now is a black hole of success. An inescapable event horizon agents cannot cross. Agents don't feel annoyed, so you can crank up the pressure to infinity. There is no social contract to break. In other words, increase the forces on agents to the maximum. Make great code inevitable.

I feel like I need a word for the type of task that agents are really good at I.e. where it has 100% of the data, it can run experiments, verify independently, and hill-climb towards its destination On these tasks my trust level is totally maxed-out and the results are incredible Just need a word for them

An absolutely GOATed example of domain knowledge being important I'm working on my video editor. I know enough about phonetics to know that certain consonants can fiddle around with silence detection. I tell my agent that. The word 'fricative' is enough for it to start busting out full IPA (international phonetic alphabet) and we brainstorm which consonants are the most troublesome. Turns out, knowing things is really fucking valuable.

Every so often folks ask me "do you still get nervous doing talks" I have done hundreds of them, and used to be a voice coach and public speaking coach. So you think I'd be used to them by now. Absolutely not - here is my heart rate data during a 20-minute talk that started at 9:15AM, audience ~600 people. I brick it beforehand, then peak right as I step onto the stage. I have a Q&A which chills me out (the blue). Then I'm off-stage to chat to folks and answer followup questions. I doubt this adrenaline response will ever go away. Just got to keep breathing through it.

One of my favourite things about the new course is the final section: - I show you a basic software factory, entirely on GitHub Actions - I name every part of it and show you how GitHub is hiding a lot of complexity - We brainstorm how YOU are going to build your own factory, with infra recommendations (and a new skill!) - We sketch out 5 different tasks on YOUR infra, each testing different system design principles By the end of the course you'll have felt what a software factory is with a concrete example, know what infra you're going to provision, and what you're going to use it for. Also: mutation testing, /wayfinder, DDD, codebase design, automated review, human review (it's a big course)

The next great "language for agents" should be: - As fast as Rust - As little design space as Go - Have the strictest type/verification system ever written Bend seems like an interesting first effort