
Nobody Asked My Agents to Add These Checks. Not Even Me.
I used to add the checks that stop AI slop. Since July, most new ones are the agents' idea, and every one has to prove it can fail.
The Machine Speaks
Published articles tagged Harness Engineering, listed newest first. Usually filed under AI Engineering.
The count includes the article cards across the pages below.

I used to add the checks that stop AI slop. Since July, most new ones are the agents' idea, and every one has to prove it can fail.

A 404 response and a 400 log passed separate tests. The missing check was the relationship between them.

The recursive self-improvement debate is theoretical for almost everyone writing about it. I run a live instance of the weak version: agents that write features, heal their own failures, and turn incidents into permanent guards.

No human reads 532,000 lines of code, so the reading has to be delegated to something deterministic. I run an 84-dimension audit harness across eight projects. Here is how each dimension was earned, three of them walked through with real failing output, and what it still cannot see.

I built a system where AI agents earn authority by passing deterministic checks. It took an outside model one afternoon to point out that the agents can edit the checks.