• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
  • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    User Reports Astra Struggles with Complicated PCB Designs

    JLCJAK shares results from routing a 400MHz oscilloscope board using Astra.

    DF
    M🪐
    JL
    7 Sources, 26d ago, first seen 26d ago

    TLDR

    JLCJAK posted that Astra is not helpful for routing complicated PCBs. He described running an experiment on a PCB for his 400MHz $300 oscilloscope project. The effort took two hours and twenty minutes on fast mode and consumed fifteen percent of his weekly limits. Four attached images display a grid of colored PCB layer routings in red for L1 F.Cu outer routing, green for L2 In1.Cu ground reference, and orange for L3 In2.Cu.

    Combined views

    176.6K

    7 Sources, first seen 26d ago

    Combined views

    176.6K

    7 Sources, first seen 26d ago

    1.5K likes
    1.5K likes
    76 comments
    684 saves
    127 reposts
    76 comments
    684 saves
    127 reposts

    Sentiment

    Positive37.5%62.5%Negative

    Summary

    Sentiment

    Positive37.5%62.5%Negative

    Many accounts criticized LLMs and Astra for failing at complex PCB routing and spatial reasoning due to missing physics and training data, while others welcomed focused engineering efforts or expected future gains.

    Based on 25 sentiment-bearing replies from 24 accounts across 2 conversations.

    Summary

    Many accounts criticized LLMs and Astra for failing at complex PCB routing and spatial reasoning due to missing physics and training data, while others welcomed focused engineering efforts or expected future gains.

    Based on 25 sentiment-bearing replies from 24 accounts across 2 conversations.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    7 Sources

    @jlcjakAstra is *not* helpful for routing complicated PCBs 😐 I ran the BIGGEST public experiment testing its capabilities and here are the results, 2h20 on "/fast" and 15% of my 20x weekly limits later Details and files: https://github.com/jlcjak/astra_piNas Small 🧵with my thoughts
    @mattparlmerLLMs are being trained to produce engineering artifacts that look convincing but are very broken bc the training data required to actually close the gap isn’t available to the labs at appropriate scale yet, and the stuff they have isn’t even very high quality
    @DanielleFongRT @mattparlmer: LLMs are being trained to produce engineering artifacts that look convincing but are very broken bc the training data requ…
    @sdamicothe model capabilities if you were already someone who could erm ... prompt ... a chinese manufacturer to emit very competent power electronics designs ... are ... something.

    7 Sources

    @jlcjakAstra is *not* helpful for routing complicated PCBs 😐 I ran the BIGGEST public experiment testing its capabilities and here are the results, 2h20 on "/fast" and 15% of my 20x weekly limits later Details and files: https://github.com/jlcjak/astra_piNas Small 🧵with my thoughts
    @mattparlmerLLMs are being trained to produce engineering artifacts that look convincing but are very broken bc the training data required to actually close the gap isn’t available to the labs at appropriate scale yet, and the stuff they have isn’t even very high quality
    @DanielleFongRT @mattparlmer: LLMs are being trained to produce engineering artifacts that look convincing but are very broken bc the training data requ…
    @sdamicothe model capabilities if you were already someone who could erm ... prompt ... a chinese manufacturer to emit very competent power electronics designs ... are ... something.