I Nearly Built an AI Army. Then I Asked What It Was Actually For.

Bespoke digital wireframe and layout design sketching for custom websites

I thought I had a brief moment of genius the other day.

While poring through my LinkedIn feed, someone said something about someone using AI to get some kind of result. The ordinary drivel. But it sparked a thought that then erupted into an idea.
I've got a lot of old laptops lying around at home doing absolutely nothing. I've also got one really good desktop computer and one modern, competent laptop.

I thought, hang on a minute. This is magnificent.

I've already got a local AI running on the desktop using Ollama and Qwen. Could I make my own little AI army out of all this retired hardware? And secondly, and most importantly, could I pull this off without spending any more money?

After a bit of digging, the answer appeared to be yes. At least enough of a yes to make me want to try it.

Handcrafted custom space station visual design layout by Super Space Goose

The grand plan was to use the desktop as the mothership. It has a 16GB Nvidia RTX graphics card with plenty of grunt and 64GB of RAM, so it can run the bigger model. Then I could have the modern laptop running something smaller and acting as a specialist worker.

If that worked, I could start digging up the graves of the elderly laptops, put them back to work and see how ridiculous this thing could become.

But let's not start on maximum difficulty. Start with two computers. Get them talking. Make them do things. Grow the army once I've figured out what on earth I'm doing.

That was when curiosity got murdered before it even made it into the same room as the proverbial cat.

What was I actually going to have them do?

Analyse my website. Write reports about said website. Check logs. Look at traffic. Run tests. Summarise things. Make recommendations. Rah rah.

And reading that back killed the idea.

I was about to spend a Saturday building an unnecessarily complicated machine and then spend the following Sunday trying to invent enough jobs to justify its existence.

I hadn't found a problem and thought AI might solve it. I'd found AI and was now wandering around looking for a problem.

That's backwards.

And AI seems particularly good at encouraging this because the technology itself is genuinely fascinating. You can very quickly go from "Could I build this?" to "I should build this" without spending much time in the rather important middle bit where you ask, "Why?"

Bespoke digital wireframe and layout design sketching for custom websites
Handcrafted custom space station visual design layout by Super Space Goose

There are things these models can suddenly make practical that weren't particularly practical before.

Let's say I ran a clothing company with tens of thousands of reviews, returns, customer emails and live chats gathering digital dust. Somewhere in there might be a pattern showing that everybody thinks a shirt fits small, nobody understands the trouser sizing and one particular jacket keeps getting returned for the same reason.

Someone could technically sit down and read all of it.

Technically, I could count every grain of rice in a twenty-kilogram sack.

Neither is likely to happen.

If AI makes finding those patterns cheap and quick enough to be worth attempting, then it has done something genuinely useful. It hasn't been added because somebody wanted an AI-shaped object in the business. It has made something practical that probably wasn't going to happen otherwise.

That's quite different from the list of jobs I was inventing for my little computer army.

Most of the jobs I was inventing already had perfectly serviceable ways of being done. I was just making them more complicated because having several computers whispering to each other about my website sounded much more exciting.

I can easily imagine the expensive version of the same mistake. Buy the AI platform first, hold a few workshops, get some consultants involved and eventually someone produces a PowerPoint with "AI Transformation" written across the front.

Then somebody has the unfortunate job of asking what useful thing has actually changed.

Same mistake I nearly made in my spare room.

Better catering.

The annoying thing is, I still really want to build the AI army.

Which means I probably will.

Because it is an experiment fit for the Mad Science Lab.

I'll duct-tape the computers together, document how I did it and give the resulting collection of machines a series of increasingly nonsensical jobs to see what happens.

Maybe I'll spend an entire weekend building the world's most complicated way of analysing my own website.

Maybe I'll discover one or two things it is surprisingly brilliant at and have to come back here slightly embarrassed to explain why everyone needs an AI army immediately.

I genuinely don't know, but I'd be pretty chuffed if that happened.

And that's the point of an experiment.

Photographic reference research of Kensington Garden geese for Super Space Goose design concept development

The idea has already taught me something useful before I've connected a single extra computer.

I started with the technology and went looking for the problem.

When I eventually build the daft thing, at least I'll know I'm doing it because I'm a boring nerd with nothing better to do on a Sunday.

Which, frankly, is a perfectly good reason.

Please accept my apologies if it ends up taking over the world with my Windows Vista laptop.

Creative asset discovery and brand consultation blueprint sessions with the Goose

looking for some help?

If you're ready to work with someone who creates bespoke digital experiences.

Reach out, the Goose is ready to listen.