AI policy

Agents run my testing and checks, about 2 million tokens a day, so one person tests like a team. I review every line before it ships.

Before any work begins, you get the full list in writing, with your estimate and the one-page NDA I sign. Nothing on it runs until you’ve said yes.

1. 01Every tool, by name
2. 02What it can see: the design file, the code, your content
3. 03Where it runs, and who runs it
4. 04What it does with your files

Tell me which tools you don’t want and they’re off the list. I’ll offer an open-source alternative instead, running on models I host myself.

Setting that up is charged on top: setup, plus compute, depending on whether the models run on a machine I control or on a cloud server. I also build the MCP tooling the work needs on my side. You see all of it in the estimate before anything is set up.

I’m completely open to it, and I’d enjoy setting it up. If your client can’t have files leave your hands, that’s a good reason and I’ll work that way.