I stopped guessing with Claude

I used to spend ten minutes re-explaining my business every time I opened Claude.

Who I am, who I work with, what good work looks like. Every chat, from scratch. And the answers still came back generic.

Claude had no idea who it was working for, so it filled the gaps with averages. No clever prompt fixes that.

The fix was three files.

I made a folder called ClaudeCowork, with a subfolder called About Me. Inside it sit three files.

AboutMe.md holds who I am and what good work looks like to me. I didn't type it. I let Claude interview me and answered out loud using WhisperFlow, because saying it out loud makes me more honest than typing it.

AntiAIWritingStyle.md is my list of banned words and sentence patterns Claude is never allowed to use.

MyCompany.md holds my goals and what I'm focused on this quarter.

The whole setup took about twenty minutes.

Once those files existed, I stopped writing long prompts. Every brief is now one line: I want the task done so that the outcome happens. Claude reads the About Me folder first, then asks me a short round of tap-through questions before it starts. Sixty seconds later it's briefed properly, like a new employee who asks the right questions before touching the work.

That one-line format now briefs everything from client proposals to newsletter drafts, and nothing comes back sounding like a robot.

Block out twenty minutes this week. Ask Claude to interview you about your business, save the answers into three files, and point every future chat at them first.