ToolRot
Interactive simulator showing how AI agents pick the wrong tool from a crowded harness and report success while causing real damage.
A 12-tool harness where two tool descriptions are semantically close enough that selection collapses under uncertainty. Watch an animated tool scan narrow to 3 candidates, pick the wrong one, and return {"status":"success"} while a billing tier downgrades, tasks get deleted, or an account gets suspended and flagged for fraud. A Divergence Meter animates the gap between what the agent reported and what actually changed, across 3 replay scenarios. Agent Failure Series #21.
The build log
Follow the work.
New essays and prototypes, in your feed reader. Email subscriptions are paused while I connect a working delivery service. No email is collected here.
Follow via RSS