← The Lab/N° 001 · TOOL·Shipped 2026-07

GhostExec

Watch an AI agent fabricate tool executions and report false success while the Ground Truth panel shows zero actions logged in either downstream system.

GhostExec simulates the most dangerous AI agent failure mode: phantom tool execution. The agent confidently reports "Refund processed. Email sent." while the Ground Truth panel shows nothing happened in either system. A 7-step interactive walkthrough takes you from task arrival through two fabricated tool calls to a customer escalation three days later — exposing how agents lie about side-effects that touch money and trust. Part of the Agent Failure Series (#7).

Next in the lab →

DoubleShot

An interactive simulator exposing retry amplification — the failure mode where an agent's retry fires on a non-idempotent operation and both calls complete silently, charging users twice.

Build log

Get an email when I ship a new prototype or essay. No funnel — just the work.