The Lab

Good enough to be dangerous.

Not an AI expert. A very keen power user.

AI didn't make me a developer. It made me able to build the things I've spent years wishing existed. I see a problem or get an idea, and now there's nothing stopping me having a crack at it.

This is where I play and cook. The builds, the failures, and exactly how each one got made.

Steve mid-jump, mid-build, slightly unhinged.

What this actually is

Rock. Do. Suck. has a middle setting. This is it.

I don't rock at AI. I'm not an expert, I've got no intention of pretending to be one, and there are people who'd run rings around me. I'm dangerous enough to be useful, and it turns out that's plenty.

That's the Do move: get AI to carry the weight, and know enough to not get played. The Lab is me doing both in public so you can see what it actually looks like.

I build with Claudio and Chatty G.

I don't drag the Learn Awesome team into this. They'd lose their minds, and they've got real client work on. The Lab is me and the robots.

How a build happens

Idea, validate, build, ship, and see if anyone cares.

01

A gripe or an idea

What if there was something that could… or I was at a cafe and wished… or I've always wondered why nobody…

And often enough: I hate what that company charges for this, so I'm building my own version of their tool. Those ones tend to be the most fun.

02

Pressure-test it with the crew

Research it, poke holes in it, work out whether it's daft or whether someone's already done it better. A lot die here, which is the point.

03

Build the MVP

An actual working thing, not a mockup or a Figma file. If it can't be used, it hasn't been built.

04

Put it out and watch

Into the world to see if anybody actually wants it. Most won't get traction and I'll say so. That's more useful than another success story.

05

If it works, bring in the grown-ups

I can get something working. Making it bulletproof is someone else's rock. That's when the real developers come in and we take it to production, in public.

Who's in it

Two robots and a short list of people who are properly good.

The AI crew

Claudio & Chatty G

Research, arguing with me about whether an idea is any good, and doing the parts I can't. Tools, not colleagues. But they're the reason any of this is possible at all.

The actual developers

When it needs to be real

If an MVP gets traction it needs people who properly rock at this. These are mine.

Jay · InventientKal · InventientMatt Mal · 99fold

On the bench

What's been built and what's next.

A few of these already exist. I just hadn't been filming them. That starts now, so this list will fill up fast.

BuiltEXP-001

NFC review tags for cafes

How a cafe adds a hundred reviews this month with a sticker and a tap. Costs almost nothing. Nobody's selling this properly.

DocumentingEXP-002

Placeholder — a real one

Swap this for one you've already made. Problem, what the big players charge, what it took you.

DocumentingEXP-003

Placeholder — a real one

Same again. These are the ones sitting on your drive that nobody's seen yet.

IdeaEXP-004

Placeholder — an idea

Something you've wondered about but haven't started. Ideas count. Half the fun is watching one die.

IdeaEXP-005

Placeholder — an idea

Another one for the bench.

YoursEXP-006

Got one?

If you've got a gripe that sounds like "why does nobody just…", send it. Some of the best ones weren't mine.

Why I bother

Keeping the big end of town honest.

I grew up before the internet and learnt it as it arrived, so I remember what finding anything used to take. I've also spent years around real developers, so I know what software used to cost to build.

Big companies charged a fortune for things that weren't special, because they could and because you had no way of knowing. Now you do. You can build the thing you actually want, without the feature bloat, for close to nothing.

To be fair to them, some of what you're paying for is reliability, and that's worth real money. You just deserve to know which one you're buying.

Know enough to not get played.

One of the five Do moves, and the whole point of this page. I can't stand watching people get taken advantage of because nobody told them there was another way. I wrote a whole talk about it called Every Town Needs a Batman.

Where it lands

Short version first. Full build if you want it.

Want one of these for your team?

A hack day, not a theory workshop.

A room, a real problem of yours, and something actually working by the end of it. No slides about the future of AI, no capability matrix. Your people build a thing and leave knowing they can do it again on Monday.

Nick everything.

The whole point is that you can do this too. Every build gets documented so you can copy it, improve it, or tell me where I went wrong.

See what's cooking