# AI use cases for Is GPT Nerfed (kiyoakii/is-gpt-nerfed)

Repo: [kiyoakii/is-gpt-nerfed](https://github.com/kiyoakii/is-gpt-nerfed) | Use cases: 3 | Page: https://dedupedai.com/use-cases/kiyoakii-is-gpt-nerfed/

A macOS menu bar app and Codex plugin that checks whether the model you selected in Codex is the one actually answering, and flags silent downgrades.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Catch a silent model downgrade in Codex: https://dedupedai.com/use-cases/kiyoakii-is-gpt-nerfed/detect-silent-model-downgrade/
- Probe a brand-new session before a long task: https://dedupedai.com/use-cases/kiyoakii-is-gpt-nerfed/check-a-fresh-session-before-real-work/
- Stop tool use when a session shows a mismatch: https://dedupedai.com/use-cases/kiyoakii-is-gpt-nerfed/halt-tools-after-a-mismatch/

## How to catch a silent model downgrade in Codex with Is GPT Nerfed

Repo: [kiyoakii/is-gpt-nerfed](https://github.com/kiyoakii/is-gpt-nerfed) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README

Install the menu bar checker and get alerted when a Codex session answers with a different model.

**Who it is for:** A Codex user on a Mac who selected a specific model and suspects the answers come from a cheaper or different one.

**The problem:** Codex can serve a different model than the one you picked, and nothing in the interface tells you.

**The outcome:** The menu bar panel shows a verdict for each recent session, and a mismatch reaches you as a macOS notification, a message in the session, and a red face in the menu bar.

### Steps

1. Open Terminal and run `curl -fsSL https://raw.githubusercontent.com/kiyoakii/is-gpt-nerfed/main/install-app.sh | sh`.
2. If macOS blocks the first launch, allow IsGPTNerfed under System Settings → Privacy & Security.
3. Click the face in the menu bar, press **Install**, and say yes when asked to trust the hooks.
4. Work in Codex as usual. Each session is probed in the background after 30 minutes of activity.
5. Open the menu bar panel, or run `nerfed report`, and read the last verdict for each session.

### Prompt to paste into your agent

```text
In the kiyoakii/is-gpt-nerfed repo on macOS, install the app with the install-app.sh line from the README, then run `nerfed report`. Return the last verdict for each session listed, and check that every session shown in the menu bar panel appears in the output.
```

### Caveat

If the server swaps the weights but keeps the model name, only the fingerprint or a smaller context window can reveal it, and the bank only covers models it already knows.

Risk, remote script: The install line pipes a script from GitHub straight into sh, so read it first if you do not trust the source.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/kiyoakii-is-gpt-nerfed/detect-silent-model-downgrade/

## How to probe a brand-new session before a long task with Is GPT Nerfed

Repo: [kiyoakii/is-gpt-nerfed](https://github.com/kiyoakii/is-gpt-nerfed) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Run the Fresh session probe to see which model a new Codex session gets right now.

**Who it is for:** A developer about to start a long or costly Codex task who wants to confirm the model before committing to it.

**The problem:** A session that is already running may have been served differently earlier, and you want to know what a new session gets at this moment.

**The outcome:** A fingerprint verdict for a brand-new session that tells you whether the model you selected is answering now.

### Steps

1. Install the app with `curl -fsSL https://raw.githubusercontent.com/kiyoakii/is-gpt-nerfed/main/install-app.sh | sh` if it is not already installed.
2. Open the menu bar panel and find the Fresh session row.
3. Click Probe on the Fresh session row to probe a brand-new session.
4. Read the verdict. Match means the model you selected answered.
5. Start the long task only after the verdict is Match, or after you have decided how to handle a mismatch.

### Prompt to paste into your agent

```text
In the kiyoakii/is-gpt-nerfed repo, trigger a probe of the Fresh session from the menu bar app or with `nerfed probe now`, then report the verdict and the model you selected. Verify that the verdict is one of Match, Suspicious, Downgrade, Upgrade, Rerouted, Downgraded, Upgraded, Unlisted, or Invalid.
```

### Caveat

A probe costs three short answers on your account, and a model outside the fingerprint bank is only mapped to its nearest look-alike.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/kiyoakii-is-gpt-nerfed/check-a-fresh-session-before-real-work/

## How to stop tool use when a session shows a mismatch with Is GPT Nerfed

Repo: [kiyoakii/is-gpt-nerfed](https://github.com/kiyoakii/is-gpt-nerfed) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Turn on halt_on_mismatch so Codex blocks tools until you say resume after a mismatch.

**Who it is for:** A team lead or careful solo developer who would rather pause an agent than let it keep working on a possibly wrong model.

**The problem:** An agent running tools keeps going after the model has changed, and the work done after that point may not be what you asked for.

**The outcome:** After a mismatch, tools stay blocked in that session until you say resume.

### Steps

1. Install the app with `curl -fsSL https://raw.githubusercontent.com/kiyoakii/is-gpt-nerfed/main/install-app.sh | sh` if it is not already installed.
2. Run `nerfed config set halt_on_mismatch true`, or change the setting to on in the app's Settings. (inferred, check the README)
3. Work in Codex as usual and let the scheduled probes run.
4. When a mismatch appears, read the evidence in the session report with `nerfed explain <session>`. (inferred, check the README)
5. Say resume in the session only after you have confirmed the model is acceptable.

### Prompt to paste into your agent

```text
In the kiyoakii/is-gpt-nerfed repo, set halt_on_mismatch to true, then confirm the setting is saved with `nerfed config` output or the app Settings. Report the current value and whether a mismatch has blocked tools in any session.
```

### Caveat

The exact syntax for setting values from the terminal is not shown in the README, so confirm it with the app Settings before relying on it.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/kiyoakii-is-gpt-nerfed/halt-tools-after-a-mismatch/