I just needed an API, and ended up building a whole Saas
I built @socialrobotio during nights, weekends, and between my daughters' naps. MRR is 0, only friends are testing it, but i'm just getting started
Automatic skill updates should be auditable, I think this could benefit agents like @openclaw or @NousResearch hermes agents from continuously adding slop.
A skill update should come with a commit paired with a date and a reason for the update. Problem -> Solution
Self-learning agents should have skill immutability.
Or at least a review-based approach to skill updates, one success after a change doesnt guarantee correctness.
It's very easy for agents to slop the sh1t out of skills and call it a day.
I love hermes-agent @NousResearch ! Im trying to find the best model to deploy it on multi-user environments without risking data leaks.
What's people doing in this sense? Should I sandbox it and dupe profiles per user?
Separate containers per agent seems way too much