The Machines Are Getting Cheap. The Toolpaths Are Not

sliptonic.com

This is difficult, complicated software, and not just something you can vibecode over the weekend. But AI assistance is not to be ignored, and spite driven programming is a force to be reckoned with. Somewhere out there is an angry person who's been mistreated and wants to show "them" who's the boss. Maybe. So I give it a fair chance that cheap tool path software will emerge. It's niche software, and that's exactly what AI driven development enables. Jevon's paradox may apply to software engineering after all.

There are a few factors working against the idea that a vibe coded solution will pop up and be viable. The first is access to actual hardware. A vibe-coder can build something for their own use but can't test it on equipment they don't have access to. This is non-trivial when you're talking about machines with really elaborate kinematics like rotating heads and tilting beds. That means a community has to form around the software.

The other thing is the same problem that has held back CAM development until now - domain expertise. Knowing the correct thing to build means having real-world experience as a machinist. Workholding, feeds and speeds, order of operations, tool properties, gcode quirks, postprocessing, etc, etc, etc. Modern CAM systems aren't just fancy toolpath algorithms. They manage the whole process of articulating the steps to manufacture the part. Unfortunately, the people that really understand that seldom know enough about software development to even vibe code it effectively. Maybe a team of those people could do it collectively but then you're back to building a community.

A community is what FreeCAD has now. We have about a dozen people working on CAM. That's why my post calls on the manufacturers to get involved.

Reply on news.ycombinator.com