Agent responsibly vercel.com
Matthew Binshtok, writing up an internal Vercel talk, on the gap between code that looks like an experienced engineer wrote it and code that knows anything about your production environment. Agent output clears review because it reads well and the tests are green, but it has no view of load, of what fails at 3am, or of the assumptions your infrastructure quietly depends on. His compression of the problem:
Green CI is no longer proof of safety.
The line he draws is between leveraging agents and relying on them, and it rests on ownership: putting your name on a pull request means “I have read this and I understand what it does.” The remedy is not more review discipline but infrastructure that makes the safe path the default one. Three safeguards: deployments that watch themselves and roll back without being asked, continuous validation through load testing and chaos experiments rather than a one-off pre-release check, and executable guardrails, meaning operational knowledge encoded as tools an agent can actually run instead of documentation it will not read.
The idea to keep is his read on where the bottleneck went: the scarce resource is no longer writing code, it’s the judgment of what is safe to ship. No figures in the piece, and it is Vercel describing practices that map onto Vercel’s platform. Constrain the agent, not the user and Pave The Road both argue the same structural point from different angles.