← Back to context

Comment by klysm

2 days ago

I don’t buy that it would be very good at reliably finding problems in schematics either. There’s no big dataset on the internet to train on

5 comments

klysm

Reply
wrs  2 days ago

Some anecdata: This weekend as a lark I asked Claude Code to design a (fairly simple) analog circuit and simulate it in LTSpice to verify. It did three edit-simulate-fix cycles and to my surprise ended up with something that seemed pretty sane.

That said, schematics (as opposed to netlists) don't seem to be a practical I/O format yet. It did generate a KiCad schematic file when asked, but it was pretty bad (penguin on a bicycle level).

Anyway, somehow there does seem to be some electronic tools training happening, becuase I tried this maybe a year ago and it was pretty hopeless.

  • wafflesfreak  2 days ago

    This is exactly why the first version of our tool worked with netlists only. We've since evolved to parsing the full KiCad project and generating a netlist from it so we can also extract schematic-specific metadata that doesn't make it into the netlist (designer notes/annotations, component positions, etc.)

    • wrs  2 days ago

      This looks great, but I want to know when LLMs will be useful for generating schematics rather than just checking them! It’s such a letdown right now to jump from doing firmware with Claude Code to drawing schematics manually like it’s 2022. :) When does KiCad get its little assistant pane on the right?

      1 reply →