Free five-minute challenge

    Stop correcting the same AI mistake twice.

    Bring one AI draft you would not send. Teach your AI what is wrong. Then give it fresh work and see whether it carries the correction forward.

    No signup. No upload here. Your draft stays in the AI client you choose.

    Run the challenge

    The challenge

    One correction. Two pieces of work.

    Use the prompt below in a new Claude or ChatGPT conversation. It asks for a small agreement before it writes.

    Challenge promptAbout 5 minutes

    Start with a weak draft. The point is to make one correction reusable, not to polish a single answer.

    Bring the honest version.

    Paste an AI draft that misses the mark. Add what feels wrong, or an example of work you would actually use.

    Approve the rules, not just the rewrite.

    The AI should show three standards in plain language. Keep, change, or remove them before it repairs the draft.

    Test the same rules on fresh work.

    Give it a related task. If it applies the standards without making you repeat them, save the compact Brain Seed it produces.

    The second output is the proof

    Did it use your standards on fresh work?

    This page never sees the work. This is a simple self-check after you test it in your own AI client.

    Before you start

    A few honest answers.

    Is this a connector or an MCP?
    No. This is the smallest possible test. You copy one prompt into the AI client you already use.
    Where does my draft go?
    Only into the Claude or ChatGPT conversation you choose. This page does not receive your draft, correction, standards, output, or Brain Seed.
    What should I bring?
    A real draft you would not send: an email, a LinkedIn post, a proposal paragraph, meeting follow-up, or short decision note.