Hi,

AI coding agents now write code faster than most teams can review it. Faros analyzed two years of data from 22,000 developers and found 54% more bugs per developer, PR reviews taking about five times longer, and more than three times as many incidents per PR when AI adoption was high. That's a correlation, not proof that AI caused it, but it matches what a lot of teams are seeing.

My answer is automated software verification: use the same agents and harnesses that write the code to check it, at every stage from pre-commit to production.

In the new post:

• The AI reviewers available today: Superpowers, RoboRev, Claude Code Review, Codex, Copilot and Cursor Bugbot, and why one AI reviewer isn't enough.
• A full verification pipeline: from linting and tests to secret scanning, migration checks and agentic review against the spec, with an example GitHub Actions workflow.
• What to run when: cheap checks early, heavier ones at the PR and pre-production stages, scaled to how risky the change is.
• How big tech does it: how Uber, Shopify, Stripe and others run AI software factories.
• My own workflow: one feature from spec to production, with a second model reviewing every commit.

Read the full post →

The short version: if AI makes writing code cheap, verification has to become a first-class part of how we build software.

If you'd like to see the lightweight harness I use for this workflow, just reply and I'll share it. I read every reply.

Until next time, may your tests pass and your context windows never run out.

~Mihai