3 ms·
You are completely wrong, but one might get that impression from not using SOTA models in the Sonnet ballpark.
by incrudible 5mo ago
You are completely wrong, but one might get that impression from not using SOTA models in the Sonnet ballpark.
- jvdongen 5mo agoI think both preceding comments are a bit too strongly worded. I’m experimenting as well with pairing deterministic programming with llm use in a similar fashion and find that it allows you to squeeze more out of smaller models than with llm-only agentic loops. It is also no question for me that the large SOTA models can do way more in llm-only agentic loops with less hassle and pre-work. If you discount the hassle of actually running them, that is. So I guess it depends a bit on what your objective is.
- ai_fry_ur_brain 5mo agoI have unlimited access to every model.