3 ms·

We tested pretty much all available "AI" board and schematic auto-layouters in the market. All failed even with the most basic tasks. On the other side the lat

We tested pretty much all available "AI" board and schematic auto-layouters in the market. All failed even with the most basic tasks. On the other side the latest frontier models are pretty strong in writing embedded C + Assembler code and debugging it afterwards. It's fun to watch it writing crazy complex 'gdb' plugins in Python. This improved brutally.

edg5000 29d ago
I find it's (Sol at least, probably also Fable) great at writing Skidl schematics, evaluating PCB layout work and operating lab equipment while testing board (scope, psu, load, debugger)
- BlackRabbit1 29d ago
  For reviewing schematic/boards I recently developed something for internal tests: 1. A frontier model (either Opensource or closed source) gets a load of text to review a Kicad project systematically 2. The frontier model can extremely "ask" a cheap Gemini vision model to give it data from datasheets or specs. Main problem: the quality dramatically hits the shitter done once it falls back to "pdf2latext" due to complex tables. It was even able to do complex ngspice simulations. But you must offer it all datasheet PDFs. $1-2 of tokens saved me from $70-80 bugged PCBs.
  - edg5000 29d ago
    Not sure why the complicated setup with Gemini. I assume you are on a Codex or Claude plan (100 or 200 EUR/USD/m). The model will just pull datasheets and use a combination of conversion to text and looking at bitmap renderings of the PDF. Haven't had that dramatic quality issue you mentioned. When the model isn't sure, it will just keep looking until it understands the datasheet clearly. But yes, it needs the datasheets.
    - BlackRabbit1 29d ago
      Codex/Claude is much much much too expensive in relation to their limitations. I stick to cheaper latest generation models on Openrouter (or EU-hosted providers when it comes to IP-protected stuff from work).
  - StevenWaterman 29d ago
    > Main problem: the quality dramatically hits the shitter done once it falls back to "pdf2latext" due to complex tables. Could try screenshotting the PDF and passing that to gemini
    - BlackRabbit1 29d ago
      This is what I'm doing by now. Uploading the entire .pdf dynamically then giving it a pipe to ask questions.
  - JoelHsu 29d ago
    [dead]
erhash 29d ago
things are getting better - but the tooling required to make auto layout work are still hard

edg5000 29d ago

I find it's (Sol at least, probably also Fable) great at writing Skidl schematics, evaluating PCB layout work and operating lab equipment while testing board (scope, psu, load, debugger)

BlackRabbit1 29d ago
For reviewing schematic/boards I recently developed something for internal tests: 1. A frontier model (either Opensource or closed source) gets a load of text to review a Kicad project systematically 2. The frontier model can extremely "ask" a cheap Gemini vision model to give it data from datasheets or specs. Main problem: the quality dramatically hits the shitter done once it falls back to "pdf2latext" due to complex tables. It was even able to do complex ngspice simulations. But you must offer it all datasheet PDFs. $1-2 of tokens saved me from $70-80 bugged PCBs.
- edg5000 29d ago
  Not sure why the complicated setup with Gemini. I assume you are on a Codex or Claude plan (100 or 200 EUR/USD/m). The model will just pull datasheets and use a combination of conversion to text and looking at bitmap renderings of the PDF. Haven't had that dramatic quality issue you mentioned. When the model isn't sure, it will just keep looking until it understands the datasheet clearly. But yes, it needs the datasheets.
  - BlackRabbit1 29d ago
    Codex/Claude is much much much too expensive in relation to their limitations. I stick to cheaper latest generation models on Openrouter (or EU-hosted providers when it comes to IP-protected stuff from work).
- StevenWaterman 29d ago
  > Main problem: the quality dramatically hits the shitter done once it falls back to "pdf2latext" due to complex tables. Could try screenshotting the PDF and passing that to gemini
  - BlackRabbit1 29d ago
    This is what I'm doing by now. Uploading the entire .pdf dynamically then giving it a pipe to ask questions.
- JoelHsu 29d ago
  [dead]

erhash 29d ago

things are getting better - but the tooling required to make auto layout work are still hard