Hello world.

After more than 25 years in the software industry, and an unexpected exit from Big Tech, I found myself in an unfamiliar situation: time. Lots of it. The kind of time engineers rarely have, the kind that invites curiosity, experimentation, and occasionally, questionable ideas.

This is the story of one such idea.

A Simple (and Slightly Dangerous) Thought Experiment

What if you could build a system that:

* Discovered profitable market niches automatically

* Generated digital wall art tailored to those niches

* Published the results directly to online marketplaces

* And quietly earned income in the background

In other words: could software design, create, and sell creative products without human involvement?

It sounded like the perfect retirement-side experiment—equal parts engineering challenge and philosophical mischief.

Designing With an AI Co-Architect

To move from idea to implementation, I worked alongside OpenAI’s ChatGPT-5, using it less like a tool and more like a collaborator.

The planning phase felt like an architectural dialogue:

* The AI proposed system designs

* I interrogated assumptions

* We refined constraints together

* Repeat… until the structure made sense

It didn’t feel like architecting. It felt like conducting an orchestra, one where every musician already knew every instrument ever invented.

At times, it was like whispering into the ear of a sleeping leviathan.

The Architecture: Automation All the Way Down

The final system ran as an automated workflow in n8n, orchestrating a sequence of Node.js tasks:

* Identify promising keywords

* Analyze successful products

* Generate new creative concepts

* Produce artwork using generative models

* Build marketing assets

* Syndicate listings through Printify

For the initial MVP, distribution targeted Etsy and Amazon.

From there, I generated a full Product Requirements Document using AI and handed implementation to a coding agent powered by Anthropic’s Claude Haiku model.

Then I did something that would have been unthinkable earlier in my career.

I went for coffee.

When I came back, most of the system was written.

The Surreal New Development Loop

My role shifted dramatically:

* Review AI-generated code

* Run integration tests

* Fix edge cases

* Deploy

No sprint planning.No backlog grooming.No weeks of incremental implementation.

By the end of Day One, the pipeline was live in my home cloud.

Naturally, I assumed I had built a fully automated passive-income machine.

Naturally, I was wrong.

Reality: The Humans Are Still Needed

The very first production run exposed the messy parts of the real world.

1. AI Image Imperfections

Roughly 1 in 15 generated images contained subtle visual defects, things only a human eye could catch. There is still no reliable automated test for “this looks weird.”

2. Intellectual Property Landmines

Some outputs drifted dangerously close to recognizable styles or copyrighted material. Text can be screened. Visual similarity? Much harder.

3. Marketplace Integration Chaos

The Print-on-Demand pipeline was far from seamless:

* Listings failed unpredictably

* Variants split into separate products

* Metadata required manual correction

Instead of passive income, I found myself babysitting automation.

After a full day of supervision, I had published about 100 products.

Automation, it turns out, still needs adult supervision.

The Moment of Existential Debugging

By Day Two, I stepped back and asked a different kind of question:

Was this even a good idea?

If perfectly automated, the system could flood marketplaces with thousands of derivative works. But that doesn’t create new value, it just redistributes attention within an existing market.

In economic terms, it wasn’t innovation.It was acceleration without direction.

And that led to one of the great luxuries of early retirement:

If you’re working on something pointless, you can just stop.

So I walked away.

The Twist: It Made Money Anyway

A month later, curiosity got the better of me. I checked the accounts.

Those ~200 listings I had published before quitting?

They had quietly generated revenue:

* About $600 on Etsy

* Around $70 on Amazon

After costs, fees, and fulfillment, the profit was roughly $200.

Not life-changing.But not zero either.

An abandoned experiment had paid for a nice dinner and perhaps a bottle of soju.

What This Experiment Really Taught Me

This project wasn’t about wall art. It was about understanding how AI is reshaping the act of building.

We are entering a phase where:

* Designing systems may matter more than coding them

* Iteration happens conversationally, not procedurally

* Engineers supervise intelligence rather than implement logic

* The bottleneck is no longer syntax, it’s judgment

AI can build astonishing things quickly.

But deciding what is worth building remains stubbornly human.

What’s Next?

There are far more interesting problems to explore than teaching machines to sell wall décor. So I’ve moved on to projects that feel less like automation experiments and more like genuine discovery.

Still, this strange little system remains one of the most educational things I’ve built.

Not because it succeeded.

But because it showed me where the real questions now live.

If you enjoy thoughtful experiments at the intersection of software, AI, and life after Big Tech, feel free to follow along. There’s plenty more tinkering ahead.

Get full access to AsianDadEnergy's Newsletter at asiandadenergy.substack.com/subscribe