[arglebarnacle]

The models are getting better at agentic coding, so over time using complicated harnesses and precise prompt engineering to attempt to squeeze out an extra X% performance will become irrelevant as the models approach expert-level performance. The bitter lesson in miniature.

[trees101]

there will always be a difference between the general capabilities, and the particularities of your exact environment and requirements.

Closing this gap is done in the harness, either through Skills, user behaviour/prompts , Agents.md etc etc.

I think that this is an area worth investing time in, but it is indeed hard to know what the scope of this is.