Agent environments / Recorded attempts
Follow an agent’s work, or inspect each tool call and its result.
01 / Captured run
A captured research-agent run that builds and tests an auditor for a physics-informed model reproduction packet.
02 / Captured run
A captured classification run that explores features, tests models and produces a valid submission, then falls short of the held-out accuracy threshold.
Each example is one captured task attempt, shown in two ways. The agent replay follows the work. The tool-call view pairs each recorded input with its output and keeps the machine grade separate. Public excerpts and summaries omit identifying details and source payloads.