Try AI with Stan

Why bother

I’m not an AI expert. I’m a guy who builds things in a workshop and who has finally figured out, after a lot of fumbling, how to use AI in a way that makes me sharper instead of lazier. That’s the difference that matters, and it’s mostly about how you use the thing.

Here’s the framing that did it for me:

AI is the apprentice you finally got - willing, fast, sometimes wrong, needs supervision.

That’s the whole model. Not a search engine. Not an oracle. Not a magic friend. An apprentice. Treat it like one and you’ll get more out of fifteen minutes with it than most people get in a month.

Where most people get stuck

Two friction points trip up nearly everyone I’ve helped. They tripped me up first. Both are easy to fix once you can see them.

Pick one AI to try

You don’t need to compare them. They’re all good enough for what we’re going to do. Just pick one and go.

My honest recommendation for someone starting today: Claude.ai, free tier. Sign up with an email. You’ll get enough free use per day to do everything on this page and then some. If you already have a ChatGPT or Gemini account and prefer those, that’s fine - the project below works the same way on any of them.

No paid tier required for any of this. If you find yourself using it daily after a few weeks, then consider whether the $20/month tier is worth it. Not before.

Do one project with me

Here’s the project we’re going to do together: have AI help you organize and label something in your house. A workshop drawer, a kitchen pantry, a junk closet, the cables behind your TV. Pick something small and real.

I picked my workshop bench drawers for this. There are eight of them, and I’ve been losing track of what’s in which one for years.

1. Tell it who you are and what you’re doing

Open Claude (or whichever tool you picked) and start a new conversation. Type something like this, in your own words:

I’m organizing the eight drawers under my workshop bench. They’re full of stuff I’ve been throwing in there for years. I’m going to dump each one out and tell you what’s in it, and I want you to help me think about how to group things and what to label each drawer. Ask me questions if you need to.

Notice what just happened: you told it who you are, what you’re trying to do, and you invited it to push back. Most people skip those three things and wonder why they get useless answers.

2. Dump out a drawer and list what’s in it

Don’t worry about being neat. Just type it out: “a roll of electrical tape, two screwdrivers I can’t identify, a tape measure, a box of mismatched screws, a flashlight that doesn’t work, three rubber bands.”

It will start proposing groupings. Don’t take the first answer. Tell it what you don’t like about its suggestion. Tell it which drawers are closer to your bench and should hold the things you reach for most. Tell it you don’t want to throw the broken flashlight away yet.

3. Argue with it

This is the part most people miss. The first answer an AI gives you is rarely the best one. The second or third answer - after you’ve told it what you actually meant - is where the value lives.

For my drawers, the first system it proposed was “by tool type.” I told it that wasn’t how I work. We landed on “by how often I reach for it,” with the most-used stuff in the top two drawers and the long-term storage in the bottom two. That’s a better system because it knows something about me now, not just about drawers in general.

4. Print the labels

Ask it to write the label text for each drawer in short, plain words. “Sharp tools.” “Tape and adhesives.” “Don’t open without thinking.” (Mine has a drawer like that.) Print them on whatever you’ve got - masking tape and a Sharpie is fine.

5. Notice what just happened

You used an AI to do something it’s actually good at: thinking through a small problem with a real person, in context, with pushback. You weren’t looking up an answer. You were having a working conversation with an apprentice who knows a lot but doesn’t know you. By the end, it knew you a little.

That’s the model. Apply it to anything else you’d like another set of eyes on - planning a trip, drafting a hard email, picking a new tool for a job, talking through a decision you’ve been putting off.

A few honest notes

AI gets things wrong sometimes. Confidently wrong. If it tells you something that matters - a medication dose, a legal claim, a wiring spec - check it against a real source. The apprentice is willing and fast; it is not the master.

There are no stupid questions when you’re talking to an AI. It can’t judge you. Take advantage of that. Ask the thing you’ve always been a little embarrassed not to know.

And don’t fall into the habit of letting it write things for you that should be in your own voice. It’s a great sounding board. It’s a poor ghostwriter.

Two things we keep out of the shop

The Workshop’s AI is here for building, learning, and getting real things done. Two subjects sit outside that on purpose: politics and religion. There are plenty of good forums for those, and this isn’t meant to be one of them.

So the assistant gently declines anything heading that way, with a quick heads-up rather than a lecture. Reframe around something else and it’s right back to work. No debate, no drama, just a workshop that stays a workshop.

Tell me how it went

This is the first attempt at this kind of write-up. The thing I most want to know is whether the voice and the framing actually worked for you - that’s the signal that tells me whether to keep doing these.

Short feedback form below. Five quick questions. Tone first, everything else second.