Engineering guides

The CAD kernel runs as WebAssembly in the tab: you wait through honest stages, keep solids on the device during build, and still download STEP from the account job.

Dimensioned draft views show front, top, side and isometric — one solid or a combined assembly silhouette, not a manufacturing drawing with GD&T or a title block contract.

A specific, non-hyped account of what AI-assisted CAD cannot do in 2026 — engineering judgement, load cases, assemblies, standards compliance, surfacing — and what that means for how you actually use it.

Both drive the same Open CASCADE kernel and both export STEP. The real differences are the API model, how you select geometry, and how each one behaves when a script grows past a few hundred lines.

Code CAD replaces the sketch-and-click workflow with a script that builds the part. Here is the tool landscape, the CSG-vs-B-Rep split that decides whether you get a STEP file, and when scripting genuinely beats a GUI.

The difference between a vague description and a usable generated part is a specific structure: part type, datum, driving dimensions, features with positions, process and standard-part designations. With before-and-after examples.

Language models do not emit solids. They emit a program or a structured spec that a geometry kernel executes — which explains both why AI CAD works at all and exactly where it fails.

Parametric CAD models are defined by driving dimensions and relationships instead of fixed coordinates. Here is how the feature tree, the constraint solver, and design intent actually work — and where regeneration breaks.

A working walkthrough of scripted mechanical CAD in Python — environment setup, a parametric bracket, a generated part family, geometry tests in CI, and the mistakes that waste the first afternoon.

A structured look at where AI-generated CAD goes wrong — geometry, tolerancing, and manufacturability — and how a responsible pipeline should catch each failure before it reaches a machine shop.

The practical difference between boundary representation (B-Rep) solids and triangle meshes — and why the distinction determines whether a generated model can be edited, toleranced, or machined at all.

The specific ways generative CAD tools mishandle tolerance — from silently defaulting to nominal dimensions to ignoring fit relationships between mating parts — and how to catch it before a part reaches the shop.

A practical, example-driven walkthrough of geometric dimensioning and tolerancing: the 14 symbols, what datums actually do, and when GD&T is worth the extra drawing complexity versus a plain ± tolerance.

Design for manufacturing basics for engineers using text-to-CAD — wall thickness, internal radii, hole sizing, draft angles, and how to prompt with the target process in mind.

Write a clear mechanical prompt, preview the solid, check dimensions, then export STEP — plus the mistakes that produce a wrong first draft.

It depends on the part and the tool. How to tell whether a text-to-CAD file is manufacturable — and what separates a demo from a usable STEP.

How plain-language mechanical descriptions become editable CAD solids, exactly what to include so the first draft is usable, and where a human still needs to step in.

STL is great for meshes and some 3D printing. STEP is what you want for editable mechanical CAD and most machining — here's how to decide, with the specific cases that trip people up.

AI 3D mesh tools look impressive in renders. Mechanical work needs editable B-Rep solids — here's the practical difference, how to tell which one you actually have, and what happens when you pick wrong.

STEP (ISO 10303) explained for mechanical engineers: what it stores, when to use it, and how it differs from mesh formats.

Text-to-CAD turns a written part description into an editable STEP solid. How it works, what it does well, and why manufacturing sign-off stays with you.

Demo-quality geometry vs. manufacturable solids — the specific, repeatable failure modes and how to actually verify AI-generated CAD before it reaches a machine.