Once AI writes code and copy faster than anyone can read it, tests, guards and reviewer agents carry the weight, and a check that quietly reports success is worse than having none. This thread is about building those checks, catching them passing when they shouldn't, and deciding which way each one should fail.
Checking the checkers
When AI checks AI, what checks the checker?
guardrails · agentic review · fail-safe design · verification
Patterns
Never let the builder mark its own homework
Each reviewer starts with a fresh context, so it hasn't absorbed the reasoning that produced the work and doesn't share its blind spots.
Decide which way every check fails
A check that crashes and reports "no errors" is worse than no check; when a check can't run, it should say "couldn't be checked", not pass.
Make every check say what it didn't check
A reviewer rewarded only for finding problems will find problems, so "nothing wrong" counts as a proper answer, and every report lists what it found sound and what it didn't look at.
Builds
- EXP-009Building
Mullo
The best page, the worse journey: when an AI tool's quality isn't worth the effort
An AI that draws my notes as an illustrated page. Against a general-purpose tool it made the better page but took far more effort, so I'm redesigning it around what only it does.
- EXP-007Building · Pilot
This lab, rebuilt
Nothing goes live unread: building a public site with AI agents, without losing control
This lab, rebuilt in three days by Claude Code from briefs I worked out with Claude. The setup that keeps an agent-built site honest, and the rules that came from mistakes.
- EXP-006Building · Pilot
Fold's review panel
Never mark your own homework: independent AI reviewers for a solo builder
Independent AI reviewers for an app I was building faster than I could read its code. They caught bugs that had passed the tests, and one finding changed what Fold is.
Notes
- Diverse by prompt, biased by role: how AI illustration passes society's stereotypes onhuman-in-the-loop · trust calibration · provenanceNote
- A recipe for honest pictures: how to avoid AI image slop without copyright troubleinference cost · unit economics · token budgetsNote