Synthetic by design
Every displayed measurement comes from deterministic FieldRook fixtures, not someone else's workstation. Inference is explicitly simulated.
OPERATION PROVE IT / PUBLIC EVIDENCE LAB
Real FieldRook investigation logic. Synthetic Windows scenarios. Every conclusion tied to evidence — including the ones that cannot be confirmed.
Offline fixtures · Simulated inference · No customer or production telemetry
Pick an investigation, follow the chain of observations, and inspect the underlying fixture data.
Read-only demonstration.
No controls operate a real computer.
This is an auditable demonstration of specific code paths. It is not a claim that FieldRook can solve any Windows problem or that a language model is always correct.
Every displayed measurement comes from deterministic FieldRook fixtures, not someone else's workstation. Inference is explicitly simulated.
Facts and hypotheses cite exported evidence IDs. The UI lets you inspect the underlying bounded tool readings and known gaps.
Recovery cases record a manual intervention claim. Fresh re-measurement determines VERIFIED_RESOLVED or NOT_RESOLVED. No automated recovery runs here.
Check out the referenced commit, run the offline gauntlet, and use the published export script to regenerate the evidence. No cloud API key needed.
cd backend && python -m eval.runAudit the code, fixtures, and methodology in the public repository. The three curated cases and source-backed results are intentionally small enough for independent inspection.
View reproducibility scripts ↗