How we verify a win

Agent Wins only lists AI agent use cases that worked. This page explains how an entry gets on the site.

1. Every win starts with a public source post

Each win is based on a public post where the person describes what their agent did. We link that post on every win page, so you can read it yourself.

2. We show the author's handle

Every win credits the author by their X handle (for example, @handle) and links to the original post. If the author changes their handle, we update the entry.

3. We report the result only as the author stated it

Numbers, outcomes and quotes come from the author's post or their own replies in the same thread. We don't estimate, round up or add results the author didn't state. When the result appears only in an image or video, we say so instead of guessing. Advice we add, like "who it's for" or "before you try it," is labeled as general guidance, so it's never confused with the author's claims.

4. We check each post with the X API

Before a win is published, we use the X API to confirm that:

If a post is deleted or the author doesn't match, the entry is removed or corrected.

5. No partial results

We only publish wins that worked as the author described them. Experiments that are still in progress, or that worked only in part (for example, "no client yet"), aren't listed. Wins we removed for this reason return "410 Gone."

6. What we don't do

Corrections

If something on a win page is wrong, or you're the author and want an entry changed or removed, contact us at chevy@mail.grokbot.com. We aim to review corrections promptly and will note the date when an entry changes.

Email updates

Get new agent wins by email

We'll email you when new wins with receipts are added. Unsubscribe anytime.