After much fumbling around with prompts and evals, this is exactly how I am using LLMs in production, to narrowly make choices and return structured data. Any deterministic work gets pulled out of the prompt and my goal is to narrow the model output to be as clearly defined and as minimal as possible.
Jev's focus on structured I/O and confidence scores are game changing. If this does at all what it claims, I think this is going to quickly become the new standard approach for agentic systems.
Great question! Yes, this works much like the doom player. Sensor data (LIDAR, velocity, etc.) becomes the state. You use the score primitive to operate the controls ("What level of braking should be applied" 0: None, 1: just slightly slowing down, 2: there's a suspicious cat on the side of the road you don't trust, ...
99.9% of the time you are right, but in the long tails (construction works, major road damage, temporary road markings that contradict permanent signs) I could see this kind of architecture being useful.
lubujackson · · focus · HN ↗
Jev's focus on structured I/O and confidence scores are game changing. If this does at all what it claims, I think this is going to quickly become the new standard approach for agentic systems.
CompleteSkeptic · · focus · HN ↗
copperx · · focus · HN ↗
dozerly · · focus · HN ↗
bobtheborg · · focus · HN ↗
Full disclosure, I am not they :=)
wg0 · · focus · HN ↗
Decision making isn't that of a bottleneck I suppose.
sebzim4500 · · focus · HN ↗