> ## Content Index
> Fetch the complete content index at: https://www.liamdarmody.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# AI Experiment 013: AI Built It. Another AI Caught the Mistakes.
- URL: https://www.liamdarmody.ai/ai-experiment-013/
- Published: 2026-11-10T16:59:33.000Z
- Updated: 2026-11-10T16:59:33.000Z
- Author: Liam Darmody
- Tags: AI Lab

When I started using AI agents, I started an experiment: could I make my first $100 from a brand-new online business, something outside my expertise, built as passively as possible? The winner from a scored list of 25 ideas was not what I would have expected: Woodworking plans. I am not a woodworker. In fact, my wife got me a woodworking kit some years ago and it's still collecting dust in the garage.

The test was simple. Build one digital product, list it where buyers already search, and let the market vote. Kill bar set in advance: five sales in two weeks, or it dies.

## What did AI build?

The first version of The Weekend Workshop, four beginner furniture plans with cut lists, was generated with code in an afternoon. It was also, in my exact words a month later, bland and janky. Then the whole thing sat parked for weeks because Etsy flagged the AI's browser as a bot and refused to let it anywhere near the signup flow. The human had to do the typing: identity, taxes, banking. I'm okay with that.

When I finally looked at the product again, I sent it to Claude for a redesign. Claude did something more valuable first. It audited the engineering and found seven real errors. A step stool whose stated height was wrong. A planter box whose bottom, as drawn, physically could not be assembled. A side table held together by a screw that would have needed to pass through three and a half inches of wood. The bookshelf was fine but everything else needed fixing

One AI wrote the plans. A different AI caught why they would have failed.

## Where does it stand now?

The corrected 19-page PDF is [live on Etsy](https://www.etsy.com/listing/4592647649/4-beginner-woodworking-plans-with-exact?ref=liamdarmody.ai) right now for $12, in a shop called WeekendBuildPlans, listed with Etsy's own disclosure that it was created with an AI generator. The listing fee was 20 cents. It generated $0\. #FAIL

![](https://storage.ghost.io/c/4c/2e/4c2ec722-2bd5-4a69-ad93-121c6c1c3e73/content/images/2026/10/exp014-inline-pdf-cover.png)

![](https://storage.ghost.io/c/4c/2e/4c2ec722-2bd5-4a69-ad93-121c6c1c3e73/content/images/2026/10/exp014-inline-stool-drawings.png)

## What I Learned

Using AI to check AI is smart. I do it often. Start something in ChatGPT and then give it to Claude to review or vice versa. I'll often go back & forth with AI on things until I can synthesize into what I like best. In this case, the errors weren't just copy, they were geometry, and a fresh model with a different failure pattern found them in minutes. So, if you build anything real with these tools, budget for a second set of artificial eyes.

The marketplace is the proven channel. No audience, no launch, no funnels. Etsy already has the buyers; the listing is the experiment.

And the human's job in all of this turned out to be small but specific: judgment calls, honest labels, and the handful of screens a machine is not allowed to touch.

The Verdict? $0 generated. $.20 spent. Experiment FAILED.

## Key Takeaways

I believe that AI provides every human with the opportunity to earn passive income by monetizing what they know. The failure point in this experiment is that I was being lazy and instead of creating something that was in my area of expertise, I just did something random. That was a mistake, however, it was fascinating to learn by doing with AI by my side. I would never have even thought about creating something on Etsy myself. Just because it failed doesn't mean it wasn't worth trying!