My thesis is clearer now. I'm short on "complex reasoning and agents" because it is often a scapegoat for poorly described problem spaces.
— jason liu (@jxnlco) July 28, 2024
My thoughts on capabilities are about figuring out the 80/20 and baking "complex reasoning" into specific tools, so you put fewer…
This point resonates with me. The more time I spend prompting models, the more it’s becoming clear that the clarity of the instructions are what matter most. Writing clear, unambiguous instructions is not easy. Decrease scope and you have a chance of doing it well.
Protobuf messages/schema in prompt instructions seem to be able to provide similar output structure to Pydantic, as instructor and other libraries do.
I’ve experimented with this approach in the past but there is a bit more that could be done to structure the messages so that they could serve the dual-purpose of prompt instructions and as RPC schemas.