Theo describes a two-week deep dive into Claude Code that changed how he builds software: he used multiple parallel agents, avoided opening an IDE, shipped a new image-studio app, converted it into a monorepo with a mobile app, and built a Twitter-lockout extension and other automation projects. His core view is that Claude Code is now good enough to materially expand what experienced developers choose to build, especially when paired with good context, Convex, and a strong harness, but it is still not reliable enough to fully trust without review.
Watch on YouTube ›Get the market thesis, key claims, assets, contradictions, and follow-up questions from any financial video — then unlock a version personalized to your portfolio, watchlist, and favorite speakers.
Theo’s central thesis is that Claude Code—especially with Anthropic’s recent rate-limit increase and Opus 4.5—has crossed a threshold where it is no longer just an editor assistant or tab-complete tool, but a system that can meaningfully change how an experienced engineer works day to day. He says he ran multiple Claude Code instances in parallel, often in “allow dangerously” mode, and built more in a holiday break than he normally would have because the friction of starting and iterating on projects fell so much. The most important part of his argument is not that AI made him capable of things he could not do before; it made a lot more of his ideas economically worth doing. He supports that view with a series of concrete demos. First, he walks through a new image studio app for T3 Chat, explaining that what began as a mockup is now a real application. …
Near term, the setup is simple: Theo is likely to keep pushing Claude Code for fast, parallel background work, but the tradeoff is recurring risk from flaky history, resumability, and unsafe commands. The immediate upside is faster iteration on side projects; the immediate downside is that reliability issues could still break the workflow at inconvenient moments.
Over the next few months, his base case is that Claude Code stays his go-to for greenfield builds, large refactors, and ops-style tasks, while he keeps a second tool for direct code inspection. The key confirmation is whether longer multi-hour sessions keep producing usable output without heavy babysitting or whether the rough edges force him back toward more manual tooling.
Structurally, he thinks software work is moving toward agent orchestration, where the human defines intent and reviews outputs rather than typing every change. If that regime holds, the durable winners are tools with strong repo-native harnesses and tight integration with a developer’s actual workflow, not just better autocomplete.
Claude Code can complete a large monorepo conversion that adds a React Native mobile app and shares the Convex backend.
The speaker says the model wrote a thorough plan, executed it, fixed errors, and completed the monorepo conversion with the mobile app and shared backend.
Claude Code and Anthropic's higher rate limits significantly increased the speaker's productivity and changed how they write code.
The speaker says the doubled rate limits let them run multiple instances in parallel, build more than ever, and that Claude Code fundamentally changed their coding workflow.
Claude Code can effectively carry out large parts of a multi-package app build and repeated UI overhauls without the developer opening an IDE.
The speaker describes using Claude Code to build a web and mobile app, perform multiple UI redesigns, and handle tasks across packages while not opening an IDE.
What changed about your workflow when Anthropic doubled your Cloud Code rate limits over the holiday break?
The speaker says they started running multiple Cloud Code instances in parallel, stopped opening an IDE for days, and ended up building more than ever. They describe the experience as fundamentally changing how they write code and making them feel 'Claude codepilled.'
What did you build during the holiday break using Claude Code?
They built two full projects from scratch, including a web and mobile app for one of them. They also overhauled those projects, shipped new features in T3 Chat, and configured their entire operating system with Cloud Code.
What was the new image studio app meant to do?
It was built as a prototyping playground to figure out the image generation experience for T3 Chat. The speaker emphasizes that the earlier image-gen mock demo is now a real app.
Unlock the full claims, asset map, scores, related transcripts, follow-up questions, and AI chat — shaped around your portfolio, watchlist, favorite speakers, and risks.
