When Startups Become Side Projects, What’s Left to Build Big?

Theo Browne traced model evolution from tool calling to orchestration, showing that a service which was a startup two years ago can now be a side project. He demonstrated a PR triage system reduced to a markdown file on cron, signaling plummeting operational complexity.

This collapse shifts how leaders think about build scope: the old tiers—side project, startup, too big—have moved down, leaving “too big” undefined. Small teams can now pursue broader platforms that previously required dozens of engineers, challenging incumbent SaaS with more use cases, less depth.

Browne argues the main obstacle isn’t model capability but developer identity: terminals, strict git, and language tribalism are skeuomorphic, holding us back. Letting go is necessary to exploit AI orchestration. Building a platform that lets users extend features themselves may be more viable, but carries governance, support, and reliability risks the talk doesn’t address.

The evidence is personal and anecdotal, not production data. Still, the signal is strong enough that engineering leaders should re-evaluate what internal tools and products can be built with far smaller teams, and when it’s time to abandon legacy workflows that constrain AI’s leverage.