@_catwu It feels like this has been promised before. Introducing hooks... hooks break. Sometimes claude can fix it's own hooks. How can I trust the old stuff when the new stuff doesn't work. AI as a product will ultimately be about trust
@Shyvadi0 do you mind explaining what approach you used? I got about rank 400 with a paced feasibility strategy and couldn't get shadow pricing / duals to land.
@karpathy This is probably dumb, but it's been a few years now, if we have memory, why aren't we rating conversations so the AI can implicitly learn what we like and how we like to learn?
@BenjaminDEKR@ibab@joannejang The company might be inept w/ protocols; someone might be lying about how that prompt change got there; some people might be willfully ignorant / naΓ―ve.
@ibab@joannejang What a L response.
Models will win/lose on trust.
People can distrust because the model's owner lies
People can distrust because the team is incompetent.
You're saying that somehow, one, single person was able to flummox Grok's prompt w/o oversight from team.
Sorry ... you're video seems to have changed. I don't see the + button. I see the [[ ]] for notes but I suppose... can I simply direct the bot to examine an entire folder of files at once? Is that part of the paid feature?
@logancyang hello. I just added copilot to obsidian. Added my key. Not sure how to use. I have a repo in a folder named 'gg' in a vault 'coding' but when I @coding, the bot can't find it.
"It seems there are no search results from the @coding tool related to your query.."
@sama OpenAI will probably release many great products over the years. Thinking about unified accounts now (e.g., Sora | ChatGPT) might save headaches later.