Have been using @1Password CLI + direnv for local dev secrets + it's pretty nice. direnv loads based on cwd, 1Pass prompts before decrypting, and .envrc can be committed. Can also `op run --env-file` to keep secrets to just the process that needs, vs entire shell
OK tried this out.
- @weaviate_io for vector search
- ez question: "I'm currently at an early stage startup where the product has limited traction. How should I think about deciding whether or not to pivot?"
- top 5 results => mistral 7B
surprisingly good for a hack lol
Another thing I wanna do is play with RAG for more free form questions: "Here's my situation with x y z, are there any similar things that have happened in the past?"
Another thing I wanna do is play with RAG for more free form questions: "Here's my situation with x y z, are there any similar things that have happened in the past?"
Another thing I wanna do is play with RAG for more free form questions: "Here's my situation with x y z, are there any similar things that have happened in the past?"
Listened to @ejames_c episode on @InfiniteL88ps and liked the "case studies in AI" idea. Did some hacking while waiting for birds to roast and pies to bake and now I have a script that diffs two case studies through a shared concept. e.g.:
Still "demo level" and pretty basic, but a good excuse to play with this newfangled LLM stuff. Could do a lot more with the prompt + setup coz the results depend on the underlying model (Claude, above) being really good.
@BStulberg "Before I learned the art, a punch was just a punch, and a kick, just a kick. After I learned the art, a punch was no longer a punch, a kick, no longer a kick. Now that I understand the art, a punch is just a punch and a kick is just a kick."