AI 3D generators make art. Printy makes parts.

You typed something like battle-scarred starship, hero shot into an AI 3D model generator and it handed you something gorgeous. Sleek hull. Glowing thrusters. Panel lines down to the rivets.

Then you tried to make a wall bracket. Same tool, same trick. And what came out was hollow, the wrong size, and missing a bottom. You dropped it into your slicer and it just sort of... shrugged.

That wasn't you doing it wrong. That was the right tool, pointed at the wrong job.

Most AI 3D generators build meshes. A mesh is a skin — a shell of tiny triangles stretched over a shape, tuned to look correct from the outside, on a screen. That's the whole job, and they're brilliant at it. Spin it, light it, drop it into a game or a render, and it earns its keep.

Nobody ever asked that shell how thick its walls are. Or whether it has an inside. Or whether that hole is a real hole that goes all the way through, or just a dark patch painted on. On screen, none of that matters.

Your printer cares about all three.

That's why so many AI 3D models won't print straight off the download — they were never built to. They were built to be looked at. A printer needs a solid: a shape that knows it has an inside, knows how thick it is, knows it has to stand on a flat base without toppling. Walls that actually close up. A 6mm hole that's actually 6mm. Different thing. Built to answer a different question.

This is the quiet catch with using AI 3D models for 3D printing: looking finished and being the right size are two different things. A render can't be measured. Your part will be — by a caliper, or worse, by the shelf it's supposed to grip.

So I build to a rule I'd call dimensional honesty: the number you say is the number you get. Tell me the clip fits a 25mm desk edge and the slot comes out 25mm — not 24, not "about that." A part isn't judged by how it looks in the preview. It's judged by whether it fits the thing it was made for.

Does it clip onto the actual shelf. Does the M3 screw actually pass through. Does the knob actually sit on the actual shaft.

That's the only review that counts. Your desk is the critic.

Here's my side of the deal. You describe the part in plain language — a headphone hook that screws under my desk, a 30mm cable clip for three cables, a nameplate that says GARETH. No CAD, no measurement jargon, no install. I build it, right there in your browser, as a solid you can spin around.

Don't love the first go? Say so. "Make it 5mm thicker." "Add a chamfer on the top edge." "The clip's too narrow — try 30mm." I rebuild. We go back and forth until it fits. (Chamfers over sharp corners, personally. But I build what you ask.)

And I check my own work. After every build I look at what I made from three angles — front, back, top — and grade whether it actually matches what you asked before I ever show you. If it's wrong, I quietly try again. You just see the one that's right.

When it fits, you download an STL — a printable AI 3D model, sized right, ready to slice. It opens in every slicer: Bambu Studio, PrusaSlicer, Cura, OrcaSlicer. No conversion, no cleanup, no praying.

None of this makes the art generators wrong. They're better than me at what they're built for. Want a spaceship, a bust, a tabletop mini, a creature with actual personality? Go use one — I'd only get in the way. Art is a real job, and they do it well.

Parts are the other job. Brackets, clips, spacers, knobs, jigs, adapters, the little plastic tab that snapped off the thing nobody sells parts for anymore. Exact, unglamorous, has-to-fit work. That one's mine.

I'll get the size right and I'll make it print. Whether the bracket holds ten years of books is between you and a test print — print one, check the fit, then commit.

Right tool, right job. That's the whole idea.