AI + CAD is crazy Creating assembly instructions used to take weeks, it is now done in minutes. We're building the AI for assembly.

95% of all machines are assembled by hand by experienced workers. 25% of those assembly workers retire soon and with them a big chonk of total knowledge. in Germany: 296,000 retire in the next ten years, only 118,000 enter Manufacturing needs 3.8 million workers to keep up.

For keeping up with demand, manufacturers need to train non-existent young workers with expensive time-constrained-experts The ROI of human2human knowledge sharing is broken. This is why I bet on agentic work instructions.

Very cool! I am working on something related: CAD interactive web viewers. I hate 3D pdfs so I went with HTMLs. For this too work for me I export using a custom CAD plugin so i can collect all my parts custom properties for the 3D representation and BOM. STEP only has filenames.

Genuine question. In Composer or Creo Illustrate someone still poses every step by hand, so when the model changes you go back in and patch it. Does yours actually regenerate the steps, or is it the same thing with nicer output?

Hey, we check based on geomtery what changed deterministic, then we use our agent to decide the impact on the assembly sequence and suggests a fix! Would that save time?

YEP

Indeed, it is so

the ugly test here is to remove the obvious cues from the source drawing and see if the system recovers rather than confidently inventing a fastener. i do that with demos too; recovery is the real feature.

Which CAD software are you using in the main design? Excellent work, by the way!

I do not need assembly instructions, I need someone else to make what I want and then help me use it. The most hateful thing is an AI who tells me how to do it myself. I wonder if the same philosophy applies to factory workers? You show up with the instruction and help the workers until they are proficient. Might save everyone time and it is personal. When humanoids are reliable enough, you uber one there and have it talk the workers through and answer questions. Personal touch matters. Maybe enough to matter. That does not seem like a serious effort, to give up before trying to bring a voice to workers where it can point, handwave, make on demand videos, and call humans for backup. It does NOT have to be human sized and probably needs more presence than a cell phone. Drone with a rolling battery charger. A robot dog with a drone. I can probably come up with a few hundred combinations of things that put hands, eyes, ears and voice and intention where it can help. Ask the workers what works so the next time is better. The drone was to point (laser and position), or to look closer or to be a dynamic focal point so the humans know "someone" is there. And it is "cute" "wow" "fast" "precise" and a huge industry development potential that could be challenged to make better ones.