Intel

Privacy Terms Submit Map API for Agents

The core reward-propagation problem for code quality — Intel — VibeLeaderboard

← All Intel

Clip / AI Agents

The core reward-propagation problem for code quality

From Harness Engineering is not Enough: Why Software Factories Fail — Dex Horthy, HumanLayer · ≈13:16

“Verifying code quality and maintainability is orders of magnitude harder than the code runs and the test pass.”

“Because the cost function of bad architecture is measured in months and years.”

“If you have a coding episode and then you only find out months later that like somebody vibed this a little bit too hard, it's really hard to propagate that reward signal back across the gap.”

What’s in it

Explains why training AI coding models on code quality is so hard
Argues test-passing is a weak proxy for real code maintainability
Breaks down why reward signals fail across long feedback delays

Watch this moment ↗

Clip transcript

maintainability of the code, it gets way harder to train on this stuff. Um so you remember this picture? Verifying code quality and maintainability is orders of magnitude harder than the code runs and the test pass. Because the cost function of bad architecture is measured in months and years. If you have a coding episode and then you only find out months later that like somebody vibed this a little bit too hard, it's really hard to propagate that reward signal back across the gap.