@rothken You don't even need clickstream, the conversations logs will be sufficient. Its also going to be relevant in timekeeping disputes. Lawyers working on separate projects while waiting on AI (what else would they do) will find it hard to challenge disputes when clearly multi-conv.
@willchen500@gabepereyra I think Hargora are both revelling in the agentic driven stock pumping that Opus has precipitated. Much of legal AI shouldn't need the uncertainty of a ralph loop and wasted tokens spent in rabbit holes. That uncertainty is hopefully enough to keep current customers happy in FUD
@curlycloud__@mattpocockuk A technique I've been working with is this, get your AI to create a html navigation for the markdown it produces, ask it to allow you to comment on the markdown, storing comments, paragraph and filename in local storage and download all to feed all the comments back to the AI
hey surprise - you can just launch interactive in tmux and then tail the jsonl - shipped a small wrapper...ralph loop iterating to full parity rn https://t.co/3N4klSSEwd
Thank you to all those who contributed to the Mike repo: https://t.co/xCa3B6uu0m
After a whirlwind week I have finally found a quiet moment to work through some of the PRs and issues.
Highlights:
- security hardening measures
- OpenAI api key support
- project page UI fixes
Thanks:
@sidprobstein@kveton@aaronjmars@thegrif
And many others who I canโt find / arenโt on X