> and I suspect creating agent harness might be the same thing like creating new programming language. Which most engineers would love to do but no one want it.

> and I suspect creating agent harness might be the same thing like creating new programming language. Which most engineers would love to do but no one want it. Correct. You have studied Paul Graham well. In this case, we do not need more agent harnesses and it seems to be a recurring hype that little to no-one wants. Unless they own the model and close down the agent harness, the majority of the open source harnesses won't make any money.