Show HN: Spice simulation → oscilloscope → verification with Claude Code

In a compact demo that feels a bit like CI/CD for circuits, it has been reported that Lucas Gerads wired Claude Code into a SPICE simulator and a LeCroy oscilloscope to close the feedback loop between simulation and measurement. He tried the usual natural-language approach first — tell Claude what you want, have it sketch a circuit — and found it fine for trivial designs but brittle for anything more complex. The pivot was simple and clever: give the model immediate, instrumented feedback. The result? Faster validation, fewer guesswork moments, and far less eyeballing of oscilloscope traces.

The demo and why it matters

The published demo is deliberately tiny: a trivial RC circuit and a microcontroller, proof-of-concept rather than a finished product. But small things scale. Gerads reports that Claude-assisted workflows sped up SPICE model validation, embedded programming chores, and the dreary business of aligning and normalizing oscilloscope data — tasks that used to mean lots of manual fiddling. Is this the start of truly interactive hardware tooling for LLMs? Maybe. It’s exciting because it’s practical, not just theoretical.

Lessons learned

A few blunt rules emerged. Claude doesn't see your physical setup — don’t let it guess. Keep measurement data fresh and don’t dump raw blobs into the model’s context; hand over files and let the tool interact with them indirectly. For microcontrollers: give an explicit pinout/pinmux map and supply a Makefile with build, flash, ping, and erase targets so Claude can rely on fixed primitives instead of inventing shell commands. In short: provide structure and boundaries. Trust the model, but verify every step.

Repos and takeaway

Gerads has open-source pieces to support the workflow: lecroy-mcp (MCP server for LeCroy scopes), spicelib-mcp (MCP wrapper for spicelib), and rc-filter-demo-files (the demo artifacts). The setup isn’t magic. It’s a pragmatic bridge between simulation, measurement, and conversational coding — a small, useful nudge toward making AI a partner in hands-on hardware work. If you care about moving AI out of the notebook and into the lab bench, this is worth a look.

Sources: lucasgerads.com, Hacker News