The archive.

A video crossed my feed this week that made me stop scrolling. Someone described a bracket in plain English: needs to hold 50kg, has to fit a specific space, should be 3D printable, use as little material as possible. Then watched AI turn that sentence into a working parametric model. Not a faster way to draw. A different way to ask.

Two gates, one industry

Professional CAD has always had two gates. The first is the licence. Tools that can run into thousands of pounds a seat quietly decide which businesses can afford to design in-house and which have to pay someone else to do it. The second gate is the learning curve: years of muscle memory in menus, constraints and feature trees before a tool stops fighting you.

Open-source CAD already removed the first gate for anyone willing to climb the second. Most people never did. The licence was never really the barrier. The software itself was.

What changes when AI understands intent

What changes when AI understands intent rather than clicks is that the second gate opens too. You stop asking how do I create this feature. You start asking what does this part need to do. That's not really a productivity upgrade for existing engineers. Plenty of them will get faster at their jobs, and that matters less than it sounds.

Who gets to have the conversation

The bigger shift is who gets to have the conversation at all. A joiner sizing a bracket. A farmer designing a part for old machinery. An idea sketched on the back of an envelope. All of these used to stop at “get someone who knows CAD.” That queue is where a lot of good ideas quietly die.

Open source removed the cost of the tool. AI is removing the cost of learning it. Put those together and the group of people capable of taking a design from idea to manufacturable part gets a lot bigger. Not because they became engineers, but because engineering judgement, knowing what the part actually needs to do, was always the scarce thing.

What this means for you this week

If you run a business that occasionally needs a custom part, a jig, a bracket, a fixture, this is worth watching even without any interest in CAD software. The gap between “I know exactly what I need” and “I have a file I can send to a printer or a fabricator” is shrinking fast. It is shrinking in a direction that favours people who understand their own problem rather than people who know a particular piece of software.

The practical move this week isn't to learn a new tool. It's to notice the next time someone says “we'd need a CAD person for that,” and ask whether that's still true.

The interesting question was never whether AI would make engineers faster. It's how many more people are about to discover they were an engineer the whole time. They just never had a way to say what they meant.