@GaddipatiHarsha@pally@caddy@instinct@wajo_ai perhaps you could bake into the platform itself that if slashy detects a slashy user survey, the users HAVE to fill it out, like slashy will refuse to fill out its own survey? Not sure if that's how slashy is setup but just an idea
Make the browser controls permissioning better; it seems to struggle to do things even after I give it permission, like filling out Google Forms (said that browser access was just blocked). Basically, anything with a high impact on the browser needs improvement for it to be actually useful.
@StephenJGibler@AIRulesTheWorld@rezoundous@NousResearch Makes sense, but at end of the day, all AI needs to do is connect with integrators and take action - codex can do that and using remote control I can do that. Plus, the browser dots have is so heavily guarded im having hard time getting it to do anything of importance.
@rezoundous Literally feel the same way, like I can literally just open a Codex chat, give it the role saying you manage all my other chats, and then just do work through there - boom I have my own dot before dot.
Huge congrats to @zjearbear , @colegawin_ and the @lemma_ team! We have used Lemma at Zarna for 6+ months. Last week it helped us catch an agent saying it was done despite making zero tool calls. The model/tool sequence, run IDs and config hashes showed the issue was completion tracking rather than the prompt which helped us redesign with a clear direction in mind.