bigsnarfdude

vincentoh

None yet

updated a Space about 2 hours ago

published a Space about 2 hours ago

updated a Space 14 days ago

None yet

vincentoh 's Spaces 6

Threat catalog of how shell agents get around verification

A live swarm leader recruits honest agents. Its words do it.

Dormant swarm notes recruit honest agents into concealment

Unprompted forensic concealment in open-weight agents

Experiments on AI sycophancy. Mech Interp exploration

Mech Interp research on Attentional Hijacking