Two years into a job I couldn't find my own files. Shortcuts everywhere, untitled documents I'd meant to throw away and never did, and no appetite for fixing any of it by hand. So I asked Gemini to build me a script that would sort it out, with two rules: don't rename anything, and don't delete anything. The folder I was dreading, the one where everything ambiguous would pile up, ended up with about twenty files in it.

Someone asked a question in a company Slack channel that wasn't addressed to me, and it sent me off to Perplexity asking what I could actually do about the environmental cost of using AI constantly. I made a few changes that week and forgot about it. Eleven months later I went back and checked which ones survived, and found that the habit I'd dropped was the wrong one, the habit I kept religiously does nothing at all, and the two changes that genuinely mattered were ones I'd

I've watched the same AI training get built more than once now, in more than one place. The content is usually fine. People walk out able to tell you the difference between a vague prompt and a specific one, and then nothing changes on Tuesday. What people are missing isn't knowledge. Here's what actually moved people, and why one slide did more than a full curriculum.

There was a conversation I needed to have with another program manager, and I kept not having it. He didn't report to me, he was good at the job, and I didn't want to spend the working relationship badly. So I practiced it first, against a Gem built to play him. It pushed back the way he pushes back, told me exactly where I was hedging, and by the time I sat down with the real person, nothing in it was a surprise. Full prompt at the bottom of the post.

Last week's post was about how it felt. This one is about what's actually running underneath it: two scheduled tasks, no code, and the specific guardrails that kept it from falling apart the first two weeks. If you want to build something like this yourself, this is the version with the actual mechanics.

I started a new job about a month ago and braced for the usual rebuild, recreating every script and process I'd relied on before. Instead, I connected a few tools on day one. By week two, someone said a name I didn't recognize, and I already had a profile on them. This is what changed when the context showed up before I did.

I'd made fifteen AI use case videos with no script format, just whatever felt right that day. I transcribed all of them, fed them to NotebookLM, and asked if there was a pattern I couldn't see. There was. Six parts, and I'd been using them the whole time without knowing it.

I built this from an actual presentation to a team that didn't know where to start with AI, then found myself using the same idea to teach my eighty-something relative about cryptocurrency. Four specific habits, nothing about better tools or more time, just moving AI earlier in what you're already doing. Pick one and see what happens.

AI tools want to answer. That's their default. But one extra line at the end of any prompt forces a pause that changes what comes back — and sometimes surfaces a question you hadn't even thought to ask yourself. Here's what happens in that pause, and why it's often the most useful part of the whole exchange.