Andrew Klofas (aklofas): sourcing memo, semiconductors
Skoor 60 · n = 4 signals · confidence 0.8 [source]
What they built
Andrew Klofas owns kicad-happy, a set of AI coding agent skills for KiCad electronics design, written in Python [source]. It analyzes schematics, PCB layouts and Gerbers and covers EMC pre-compliance, SPICE simulation, datasheet download, component sourcing and fabrication prep [source]. It runs in Claude Code, OpenAI Codex, GitHub Copilot CLI, Google Antigravity and opencode, as a GitHub Action, or as standalone Python scripts [source]. It is MIT-licensed and listed in Awesome KiCad [source].
Record
Measured signals (n = 4):
Profile: the account dates from 2009-12-04 and has 54 followers [source]. Location is Sacramento, CA, and the blog is andrewklofas.com [source]. Company and bio are blank [source]. Prior employers, education and papers are not in the evidence.
Inference: 9620 contributions a year is about 26 a day. That rate may include automated or generated commits.
Inference: 54 followers against 1393 stars suggests the users are following the tool, not the person.
Why now
The repo was created in March 2026 and supports five agent environments [source].
Inference: the project depends on terminal coding agents becoming common enough for hardware engineers to adopt agent-based design review.
Inference: the work is PCB design tooling. It sits next to semiconductors rather than inside chip design.
Unknowns
- Company, funding, co-founders, revenue and paying users are not in the evidence.
- Whether the users are hobbyists or professional hardware teams is not in the evidence.
- Accuracy of the design review against a known set of board bugs is not in the evidence.
- The README example shows a "5V" buck whose feedback divider computes to Vout = 3.87V [source]. The evidence does not say whether this is a bug the tool caught or an error the tool made.
Questions for the room
- Of the 1393 stars and 128 forks, how many come from teams shipping production boards, and can you name three of them?
- In your README example, the buck labeled 5V computes to 3.87V. Is that a real bug the tool caught on that board? What is the tool's false-positive rate across a labeled set of designs?
- How many of your 9620 commits last year were written by you, and how many were generated by agents or automation?
What would move the Skoor
It would rise with named professional users, a measured bug-catch rate or a company formation. It would fall if the commit count is mostly automated or the users are mostly one-time hobbyists.