"AI Tooling access"

[tomstdenis]

Quick msg. One of the devs asked to connect an "AI-Native" tool to the repo which I rejected. I get that I'm not active here anymore but given how/where LibTom projects are used I strongly discourage agentic unsupervised coding efforts on projects like this. For instance, LibTomMath is still actively used in the EDA scene where people care very deeply about the quality of the code that goes into their tools.

Also hooking up random third party CI/CD tools with write access is just asking for supply side attacks to be a thing.

It's one thing to use an LLM to help review or reference things and quite the other to let an unguided missile through your bread and butter.