Person/Theo Browne/YouTube/26/07/What do we build now?

• What do we build now? — Theo Browne, @t3dotgg

  • Summary

    • AI models now understand code deeply enough to plan and organize work themselves, not merely autocomplete it.
    • The argument: push these tools further to build bigger, bolder projects instead of letting old habits and tooling cap ambition.
    • A call to rethink workflows and compete at a higher level of what software can be.

  • My Notes

    • Talk by Person/Theo Browne at the AI/Engineer/World's Fair (2026)

  • Video

    • • 00:17 Opening: A Journey Through the Model Eras

        • Frames the whole talk as a shared case of "AI psychosis" — the disorienting pace of recent model jumps.

        • Sonnet 3.5 — the tool-call era

          • "Sonnet 3.5 was a big moment for me because it felt like these models could suddenly complete much more end-to-end tasks."
          • Not the first model to call tools, "rather it was the first one that did them consistently and reliably enough in context of a code base."

        • Opus 4.5 — long-running tasks

          • "A model that could test the work and actually get it into a good state and complete tasks that take hours instead of minutes."
          • Where his "psychosis started in November and December of last year."

        • Mythos & Fable — the jump to orchestration

          • "The first model that doesn't just understand your code base, but it understands itself."
          • "It knows how to spawn additional models and break up work... You don't need some custom system, some fancy software factory. You just need to prompt it to go a little further."

      • 03:03 We Have to Go Bigger

        • "You're not going to see the benefits going forward if you're not pushing the model further, you're not pushing yourself further with what you're building."
        • "The models are getting better faster than we are. So we can't necessarily get better. So instead we have to go bigger."
        • Concedes he was wrong to claim models were "hitting a wall."

      • 03:44 Getting Over Ourselves: The Skeuomorphic Phase

        • Developers carry strong opinions ingrained from tools like tmux, GNU screen, and SSH — learned "before I even wrote code."

        • The iOS 6 → iOS 7 analogy

          • Skeuomorphic iOS apps had to look like the physical objects they replaced (a compass, a bookshelf) — designed "to convince you to use them, not to be useful."
          • "iOS 7 represents the shift to not focusing on convincing anymore... stop convincing and start embracing" — uglier, but far more useful.

        • Devs are still skeuomorphic

          • "We're currently in our Skeuomorphism phase as software developers... We're pretending our terminals are the ultimate interface when they're not even good interfaces."
          • "Natural language has no place in a terminal, but we pretend it does because the terminal's familiar" — said as someone who loves the terminal (and the Vim phase) deeply.

      • 06:47 The Identities We Cling To

        • "Why can't we commit our environment files?... It's just how Git was built... and we aren't letting go of that."
        • "Why do we qualify ourselves by the languages we know?" — an identity trap for juniors and seniors alike.
        • "Why are we so scared of deleting code?... we have such a bad sunk cost mindset." Names the "guilt merged PR" — merging weak work to dodge a hard conversation.
        • "The things we care about are not necessarily the things that matter anymore."

      • 09:29 Three Things I've Built, and the Shifting Tiers

        • Walks three projects bottom-up: a Reddit meme scraper (side project); Ping, a "Zoom for streamers" Y Combinator startup built around OBS; and a full-stack cloud — "Vercel, but it goes further each direction."
        • Old tiers: side project / startup / "too big."
        • "Now that the models are bigger, the tiers have shifted. Everything is now one tier lower" — "what used to be a startup is now a side project."

      • 11:28 Executable Markdown: The "Galaxy-Brain" Tier

        • Below "side project" is a new tier: "It's a Markdown file."
        • "The fact that you can now execute markdown by just piping it to Codex or Claude is unbelievable."
        • His PR-triage service is now a markdown file on a cron: "every morning at 9:00 a.m. this runs... my markdown file generates me my work for the day."

      • 12:30 What Does "Too Big" Even Mean Now?

        • The full-stack cloud is his own next bet ("Don't do it... I wouldn't want to compete with me on this one").
        • Above it sits a gap he can't name: "I don't know what too big means anymore. Is it training your own model from scratch?... building your own operating system?"
        • "That's scary, but it's also exciting. It means I need to keep pushing myself to go bigger than makes sense in order to find these limits."

      • 13:12 Think Wider, Not Bigger: Breadth vs Depth

        • Reframes "bigger" as wider — the breadth (range of things covered) versus depth (features per area) of software.
        • Vercel never matches AWS's range but wins on depth in front-end-leaning full-stack — "so much so that even the agents prefer it."
        • What changed: "that range is viable in a way that it never was before... you can build a database platform into your product in a day or two of work with enough prompting."
        • Architect so users extend it themselves: Slack "is the right shape for people to build the features they want" and accidentally became the platform people run their agents in.

      • 15:14 If Your Idea Doesn't Feel Stupid, It's Not Big Enough

        • "It's time to compete with Slack. It's time to build your own AWS. It's time to challenge Salesforce directly."
        • "If your idea doesn't feel stupid, it's cuz your idea's not big enough."