Our Story

I naively assumed LLMs could just do this by now: point a model at an electronic troubleshooting guide for a classic BMW and have it create KiCad files I could use to build a modern troubleshooting tool.

They could, sort of. Not in a repeatable, scalable, and reliable way.

I dove into the science and the existing research, ran experiments, and kept the work on a real customer problem instead of an academic one. That is where I found the secret sauce.

This is a very hard problem. LLMs are essential to solving it, but they are not enough.

The trick is applying them to the right parts and using other techniques for the rest.

I've decided this is worth going big after, and I am funding a new company for it: Excaliwire.

I am looking to hire a part-time developer to help get it to the next stage. If you are interested, use Send feedback below.

If you have been trying to get the same thing out of a model, the Join waitlist button will let you share that.

If you want Excaliwire, join the waitlist. If you want to join me, send feedback.