AgentDrift
A single-file dashboard that flags when an AI agent's final synthesis contradicts the evidence its own tool calls gathered.
AgentDrift detects reasoning drift — when a multi-step agent gathers consistent evidence across its tool calls, then writes an optimistic summary that ignores what those tools actually found. It scores each run on a 0–100 Drift Score and surfaces HIGH/MEDIUM contradictions in a clean diff panel that pairs every claim against the contradicting step. Pure HTML with zero dependencies and zero build step; ships with a demo scenario plus a JSON log format so you can audit your own agent's reasoning chain.
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