What If? Garage

How we use AI

Built with AI.
Proven in the real world.

People have strong opinions about AI, and a one-line mention leaves too much to guesswork. So here's exactly how it's used, what it doesn't do, and the rules we hold to.

AI is a tool, a collaborator, and what makes one-off builds on this scale possible at all.

It speeds up research enormously. It lays out the options and the reasons one might be better than another, until I make the final decision. And it lets me focus on learning the skills it can't replace, while it handles the skills it excels at, like coding, image rendering, or turning a few specs into a Fusion script that models a whole part, instead of me spending years becoming proficient in those first.

Ryan, founder of What If? Garage

That's why the site says “we.” The ideas and the hands-on work happen here, but it's anything but a one-person effort, and it's turned what would have been years of work into weeks.

What AI does

Research and comparing options

Reading datasheets, manuals and specs, and laying out the options with the reasons one might beat another. The Camry's engine went through a shortlist of nine before the 2.3 EcoBoost won. AI makes that kind of comparison fast; the call is still ours.

Writing software

CAMRY-OS, the Camry's infotainment system, is written with AI. Every feature gets run and tested on the real bench computer, with real GPS, screens and controls, before it counts.

Drafting CAD and circuit boards

Turning a handful of measurements into a Fusion 360 script that models a whole part, or a schematic and board layout from a parts list. Those drafts get printed, test-fitted, measured and corrected by hand. Circuit boards stay labeled as drafts until they're reviewed.

Documentation and this site

Build documents, wiring architecture, parts lists, READMEs and the words on this website are written with AI from our decisions and notes. It's a big part of why the builds can be documented well enough for anyone to follow.

What we do

  • Come up with the ideas. Every “what if” starts here.
  • Make every decision. AI offers options and reasons; it never picks the part.
  • Learn the hands-on skills AI can't replace: welding, fabrication, machining, wiring, assembly and testing.
  • Do the real-world work: buy the parts, print and fit them, build it, and find out whether it actually works.

Firm lines

The rules we hold to

01

Every decision is ours

AI suggests, compares and drafts. Every part choice and design call is made by a person, and the build documents say who decided what and when.

02

Professionals check anything safety-critical

A structural engineer stamps the shop. A professional tuner checks the engine map before it ever runs. Cage geometry goes past NHRA, and an ASE Master Technician signs off on the assembled car. Circuit boards get reviewed before they're ordered.

03

AI images are always labeled

Concept renders on this site are AI images, and they say so. They get replaced with real photos as each build comes together.

04

Nothing is done until it works

Statuses on this site reflect the real world. “Designed” means drawn, not built. “Chosen” means picked, not bought. “Working” means it's been tested on real hardware.

Straight answers

Isn't this just AI slop?+

Slop is output nobody checked. Everything here is checked against something real: code runs on the actual hardware, printed parts get test-fitted and reprinted until they fit, and designs that haven't been built yet are labeled that way. The site tells you which parts are proven and which aren't.

Can you trust AI with engineering?+

Not on its own, and we don't. AI is fast at research and drafting, and it can be confidently wrong. That's why every decision is made by a person, everything gets tested on real hardware, and anything that can hurt someone (structure, cage, engine tune, high-current electronics) goes past a qualified professional before it's trusted.

Isn't it cheating? Did you actually learn anything?+

The goal is to learn the skills that matter in a garage: welding, fabrication, machining, wiring and diagnosis. AI handles the skills it's genuinely good at, like coding, image rendering and turning specs into CAD scripts, so the time goes into the hands-on work instead of years of becoming fluent in everything else first.

Some background, from Ryan

I didn't wake up one day after twenty years at a desk and decide to build custom cars. The passion for cars was there young, whether I realized it or not, and so was my inability to leave things the way they came.

My first build was a Power Wheels Corvette. I took it apart, painted it black, swapped the motors and ran different batteries, a good two decades before that became a thing to do. My dad helped a lot, but I was the one who wanted a faster black Corvette, and that wasn't something you could just buy. Some years later I did the same mods to my younger brother's Jeep Wrangler. It wasn't even mine to tinker with, but I couldn't leave it alone. He was happy in the end.

My first real car build ended with a ruined car, because the end goal was far beyond what I could actually make happen at the time. Even with that epic failure, the desire to build custom cars stayed. It took a pause when life got in the way, but it was never a hobby I gave up on.

I was never much of an academic student. Not because math and science were a problem, but because I'm not one to sit in a classroom and listen to a lecture. I thrived in shop class. In high school I spent most of my time in wood shop designing and building desks, cabinets and tables. I always wanted to take metal shop, but my dad was the full-time instructor for both the conventional and CNC machining programs at a local community college, so from about 15 I had better access to machines, tools and materials than a high school metal shop could offer.

From 16, as a high school sophomore, I competed in several WESTEC Manufacturing Challenge competitions with the Pierce College team. Every team I was on finished on the podium, and on many of them I did the vast majority of the CAD/CAM work and machined the finished parts. In between, I designed my own parts: some never made it off the screen, like a completely new trigger frame for a paintball gun that's now obsolete. I did make and sell aftermarket parts for several paintball guns, and made parts for my own RC cars, though I never managed to scale those into a product.

I earned my AS in NC (numerical control) programming and operation around 2008. I'd finished the degree classes well before that; the general education side took longer than it should have.

Since then I've worked in a few shops: as a temp at Haas, making the parts that go into the same machines I use every day; a few years making tooling; and now making motorcycle wheels.

I'm a jack of all trades, but my strong suit is machining and mechanical design. Design, not engineering, even if that's what it ends up being. At the end of the day I'm a machinist. I'm not a software engineer, a programmer, a mechanical engineer or a slew of other professional titles, though I have a base understanding of much of what those fields require. That's where AI comes in: it covers the specialties I was never trained in, so my time goes into the hands-on work I've been doing my whole life.

Doesn't AI take people's jobs?+

Builds like these wouldn't exist at all without it. One person could never write an infotainment system, design custom control boards and fabricate a tube chassis in a reasonable lifetime. The work that needs a qualified professional still goes to one: the structural engineer, the concrete crew, the tuner, the inspectors.

Are the pictures real?+

Concept renders are AI images and are labeled as concepts. Screenshots of CAMRY-OS are real, taken from the software running on the bench. Real build photos go up as the work happens.

So could I do this too?+

Yes. The parts lists, write-ups and build repos are public so anyone can follow them. But a copy of one of these builds wouldn't be a What If build. What makes it one is doing it first. The real point is inspiration: take what you learn here and use it to come up with your own one-off, something nobody has built yet. Whether you build it yourself or bring the idea to the shop, the goal is to expand on what's been done here, not treat it like a recipe for copies.

See it in practice: the Camry's infotainment page shows what's working, what's bought and what's still a draft.