You probably came from the newsletter or the poll, looking for somewhere to actually run the move instead of bookmarking it. This is that place.
One line for the whole room:
One useful AI move each week. Try it on real work. Post what happened. Keep what works.
You get better at AI by running moves on your own work and keeping the ones that hold up. So that’s what we do here.
The promise
Every week, one move. Small enough to try in 10 minutes, on the tool you already pay for.
You bring a real task. You post what happened, including the parts that broke. Other people do the same. The results stack up where everyone can see them, so the next person doesn’t start from zero.
Everything here is free
You don’t need to pay for anything to get your first result. The whole workbench is free: every thread and every tested move.
Later there are paid live sessions and a cohort, if you ever want them. They sell help and speed. Access stays free. If the free room ever turns into a sales pitch for paid content, it stops being worth your time. That won’t happen.
The five moves
The room is organized by what you’re working on this week. Find the phase that matches it.
- Reset. Get one usable draft in 10 minutes.
- Repeat. Run the same prompt three times, then put it behind a shortcut so it’s one keystroke away.
- Check. Learn where the model lies, and catch it in 10 seconds before it reaches a client.
- System. Build a context profile so the model stops starting from scratch every chat.
- Share. Hand a working setup to someone else. The best ones become permanent here.
Same room, same weekly thread. Someone writing their first draft and someone building a template use the same bench.
The four spaces
Start Here. This page. Come back whenever you want the map again.
The Workbench. The weekly ritual. Each Monday there’s a move. You reply with your result. The thread does the teaching, the replies do the proving.
Check Before I Use It. For any AI output you don’t trust. Post the task, the output (paraphrased), and what feels off. The room helps you catch it.
The Vault. Tested moves with honest breakdowns. Every item tells you when to use it, where it breaks, and what to check first. Nothing goes in until someone actually used it in the Workbench.
The three rules
1. Post what actually happened. Real tasks, real results, real failures. “It gave me garbage” is a useful post. Theory belongs in articles; this is the bench.
2. One fix per reply. When you help someone, give them one improvement they can act on right now. One clear next step does more than a list of twelve.
3. Paraphrase your work. Don’t paste raw output. No client names, no internal numbers, no contracts or employee records in a public room. Describe the task, summarize the result, keep the private details out.
Your first move
You probably landed in a thread already, with a prompt pinned at the top and other people’s results under it. Go back to it. Run the move once on something real. Reply with what happened.
That first reply is also your introduction. Your name, your task, what the model did. No separate intro post needed.
One thing, so you know who’s at the bench with you. I still organize my project files by hand, even though AI could do it for me. We’ve all got a habit like that. This room is where we catch them, one at a time.
Then you’ve started. Welcome to the bench.