Building Software Factories with AI Agents

Building a Software Factory: Week 1, Zero to Product

TLDR: Ona spent a week building a fully autonomous software factory, documenting the entire process in public. They took an empty GitHub repo and shipped a working Notion-like app called Memo with 130+ PRs merged, 12,202 lines of code, and 14 automations running the factory. The key insight: speed is not the bottleneck anymore, the quality layer is what separates "agents writing code" from a true production line.

Building a software factory: Week 1, zero to product

Veto: Blocking Executables by What They Are, Not Where They Live

TLDR: Veto, a tool from Ona, now uses bare executable names in its deny list instead of specific paths. A BPF discovery agent scans the devcontainer and all nested Docker-in-Docker containers, finds every executable matching that name, hashes each one, and blocks them automatically. Same enforcement layer, but now it covers the full filesystem hierarchy without manual path hunting.

Veto finds the executables. You just name them.