Jev gates what four agent dots can do autonomously
A demo of an agent brain where Jev handles permission gating: the dots think freely, but running actions, spending money and talking to people all require user approval, with a full thought log attached.
AN 18-YEAR-OLD STANFORD DROPOUT BUILT AN AI BRAIN THAT'S 93% SURE AND STILL ASKS FOR PERMISSION. INVESTORS CALLED IT A BUG. IT'S THE WHOLE PRODUCT 4 dots do the work, jev decides what they're allowed to do alone. here's its thought log from one run: 03:05 pick · path A 03:05 plan · 4 steps 03:06 tool · read logs, run tests 03:07 cost · $0.04 03:08 assign · coder dot, review dot 03:09 doubt · 1 flag 03:09 test · dry run 03:10 verify · 12 pass 03:13 wait · you 03:13 ok · you said yes 03:14 act · steps 1 to 4 03:15 learn · +1 memory 03:15 done · logged the brain is free to think as much as it wants. in one run that was 72 thoughts across 9 branches, with 3 thrown away before you ever saw them running the action, spending money and talking to people are locked behind "you". the last line of its code says it plainly: # you decide ↓
This post quotes another post: x.com/tekari0/status/2109228819228065864