You Don't Need Full Control in 3D Design. You Need the Right Result

There's a belief deeply embedded in engineering culture: the more "Manual" control you have over a 3D model; every radius, every edge, every connection point; the better the outcome. More control equals better design.

We think this is wrong. And here's why.

Why Engineers Want Full Control

The instinct makes sense. Designing physical structures is an intricate process. A flange needs to mate with a specific surface. A load path runs through a particular edge. A tolerance stack-up demands that every dimension be exactly right. Miss any of it and the part fails; or worse, the assembly does.

So designers demand tools that let them touch everything. And for good reason: those detailed requirements, boundary conditions, specifications, and initial conditions have to come from somewhere. They come from sizing tools, from analysis, and from the hard-won experience of the engineer doing the work.

The designer is, in essence, a translator. They take a mission; build this structure, carry this load, fit in this envelope, survive this environment;and convert it into a detailed geometric description that a machine can manufacture. That translation is difficult, expert-driven work, and it requires tools that match that complexity.

So full control isn't a luxury. It's a necessity; when the user is a human being doing that translation manually.

But What If the User Isn't?

Here's where the assumption breaks down.

The complexity of a design tool should be defined by the complexity of the user's task; not by tradition. If a human needs 200 parameters to specify a part correctly, a human-facing tool needs 200 controls or desgrees of freedom. That's fine.

But what if the detailed requirements, boundary conditions, and specifications could be transmitted to the design system through another method; one that doesn't require a human to manually set every parameter? What if the system could interpret the mission and generate the geometry directly?

Suddenly, the case for 200 manual controls gets weaker.

The 95% Rule

There's a fair objection here, and it's worth taking seriously: even if a system can generate most of a design automatically, the engineer still needs to review it. They need to verify it meets the intent. They need to catch what the system missed. And they need to make final adjustments before anything gets manufactured.

That's completely valid. A design tool that removes the engineer from the process entirely isn't a better tool; it's a liability.

But here's the question worth asking: what if a system could handle 95% of the design work; generating geometry that meets the requirements, respects the constraints, and produces a watertight, manufacturable file; and then export it to a native CAD format so the engineer can do the final 5%?

Most engineers would take that deal immediately.

And that's exactly the direction Kyrall is moving. Generate the geometry from the intent. Export to STEP or Native CAD formats so the engineer can finish the job. Keep the human in the loop for the part that actually requires human judgment.

The last 5% of manual work won't be manual forever, either. But that's a problem for another day.

The Tractor Analogy

Think about plowing a field.

A farmer with a hand plow has total control. Every furrow is exactly where they put it. Every depth is exactly what they chose. Maximum control.

A farmer with a tractor gives up some of that granular control. The tractor doesn't respond to every micro-adjustment the way a hand tool does. But the farmer still steers. They still decide where the rows go. They still make the judgment calls.

And they plow the field in a fraction of the time.

The goal was never to manually control every clod of dirt. The goal was a plowed field. The tractor doesn't compromise the goal; it serves it more efficiently.

Design tools are no different. The goal is a correct, manufacturable part that meets the requirements. Full manual control over every geometric parameter is one way to achieve that. It isn't the only way. And as AI-assisted design matures, it increasingly won't be the most efficient way.

This Has Happened Before

When CAD software first appeared, the reaction from professional draftsmen was skeptical at best. The tools were slow. They were clunky. They couldn't replicate the nuance and precision that an experienced drafter could achieve with pencil, paper, and decades of accumulated technique. Manual drafting gave you total control over every line weight, every annotation, every geometric relationship. CAD, by comparison, felt like a step backward.

We know how that story ended.

Today nobody seriously argues that manual drafting is more efficient than CAD for professional engineering work. The transition seems obvious in hindsight. At the time, it wasn't obvious at all; it required a generation of designers to unlearn one intuition and build a new one.

We believe the same dynamic is playing out right now with AI-assisted design.

The first generation of tools will be slow, imprecise, and frustrating. They won't do exactly what the designer wants. They'll require workarounds. Early adopters will complain; fairly; that the outputs don't meet the bar that experienced engineers can hit manually. These criticisms will all be correct, and they will all be temporary.

Because the direction of travel is clear. Imagine a design workflow where:

• 100 detailed variations of a 3D assembly are generated in minutes, each one parametrically consistent and manufacturing-ready
• Modifying a single part automatically propagates through the entire assembly, respecting every geometric constraint downstream
• 3D modeling tools connect directly to sizing and analysis tools, so the design updates when the physics does
• The pipeline from requirements engineering through design, simulation, and manufacturing runs in a single unified format; no translation, no data loss, no version conflicts

Once that pipeline exists and runs reliably, the question of whether you can manually adjust every radius and every edge length becomes much less interesting. It will still be possible. It just won't be the point.

The goal was never to have more controls in the toolbar. The goal was always the correct 3D model, built fast enough to matter.

What This Means

Kyrall is built on this idea. Not that control doesn't matter; it does; but that control over the outcome matters more than control over every input during the design process.

The designers who will build the best products in the next decade won't be the ones who spent the most time moving and adjusting geometry manually. They'll be the ones who spend their time setting up the pipeline to generate 3D models faster and more efficiently.