Back to directory

IsGPTNerfed

A macOS menu-bar tool that checks whether Codex is actually using the model you selected, by reading Codex's own per-turn model records and periodically fingerprinting the live model with a bias test on random-number generation.

IsGPTNerfed

IsGPTNerfed watches Codex sessions for silent changes: a swapped model, a lowered reasoning effort, a hidden internal model, or a smaller context window, all read from the records Codex itself writes after each turn, at no extra token cost.

On a schedule it also forks the current session three times through the same app-server call Codex's desktop app uses for a side chat, asks each fork for about 300 "random" numbers, and compares the statistical bias of the answers against a calibrated fingerprint bank to identify which model actually answered. The verdict reports a match, an unconfident suspicious result, or a confident downgrade, upgrade or reroute.

A mismatch triggers a macOS notification, an in-session message and a menu-bar indicator. It runs entirely on the local machine, installs from a shell script or a downloaded disk image, and requires macOS 15 or newer.

More in AI evals

See all 269
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • JevBench

    benchmarkheaven.com

    A reproducible benchmark from Benchmark Heaven that ranks Jev-class typed decision models on intelligence, calibration, speed, and cost, with a published harness, scoring rules, and results.

  • A co-evolutionary adversarial red-teaming tool for AI agents that uses quality-diversity evolutionary search to autonomously breed edge-case scenarios, testing whether an agent with tool access can be made to hallucinate, deadlock or leak funds.