Inventory for Rapid Prototyping

I've heard good things about apps like SuperCook, which let you list the ingredients you have in your cupboards and fridge such that they then generate lists of recipes you can cook using what you have on hand. I personally haven't felt compelled to use such apps, as I feel comfortable enough to wield herbs and spices in an experimental way, adding dashes of spice to taste, but when it comes to more finitely deterministic ingredients such as integrated circuits and other semiconductors, I tend to like a bit of proof and validation behind my work before I start powering things up. Much like cooking, electronics tend not to be as enjoyable once they've burnt.

I've recently found that Anthropic's Claude models (Sonnet 4.5, and more recently Opus 4.6) perform quite well at electrical schematic design and validation. This has been great for rapid prototyping with breadboard circuits where the overhead to go from idea to fictional prototype seems to be very low. Ill get into the details of what I've been building in a future post, but for now I wanted to share a few notes about the workflow I've been using.

For our last anniversary, my wife got me a handful of miscellaneous electrical component kits which I have been using the heck out of (1st wedding anniversaries traditionally have the theme of "paper", but I think "semiconductor" might be a good modern equivalent). From the title photo on this post it's the:

I highly recommend these kits, and I will be getting more soon.

Inventory Workflow

Using AI tools to augment my ability to iterate on ideas and prototype rapidly, I've found it helpful to maintain a list of components I have available on hand. This lets me and the AI agent work together to come up with a viable solution using the parts available. I'm sure a significant number of electronics tinkerers and DIY-ers out there will recognize the feeling of opening up a project book/guide only to realize that they have every piece needed except for neither the critical 2N3904 or exchangeable PN2222 transistor.

Within the current few months of AI, simply sharing AI workflows seems to be the most useful way to gain experience and try things out, so my workflow for electronics prototyping has been as follows:

  1. Take a picture of the packaging, or copy a website link where the components or items is listed. Ask Google Gemini to convert it into a JSON file using a format I specify with an example.
  2. Keep a folder of JSON inventory files with the combined data from the components I have on hand.
  3. Store that data in a central location. I've created https://github.com/salvius/inventory as a central repository for the components I have on hand. It can be clones to allow local agents to access the list of available parts, accessed by hosted agents in virtual environments, or simply used as a reference source by providing links to the raw files as needed (ex. https://raw.githubusercontent.com/salvius/inventory/refs/heads/main/microchips.json)
  4. Prompt the agent to use either the cloned local data or referenced to remote files.

To help manage context for LLMs, I've attempted to structure my data files in a reasonable way, currently classified by category into files such as "capacitors.json", "diodes.json", "microchips.json", "resistors.json", "transistors.json", etc. On top of that, JSON format tends to have better accuracy with LLMs than formats such as CSV. TOON seems to be promising as well, so I may switch over to it at some point.

Conclusion

At the end of the day, this workflow aimed at shortening the distance between "I have an idea" and "It actually works." Digital inventory management provides both organization, and gives your AI collaborator the visuals it needs to see what’s actually sitting on your workbench. Give this a shot with your own parts bin, and hopefully, we can all keep the "magic smoke" inside the chips where it belongs.

No comments:

Post a Comment

Copyright 2025 Gunther Cox