GitHub
The GitHub repo
goes live July 20.
I am preparing the public Security in Public release. It will include the Grounded Confidence code, a simple Email Investigation AI Agent, and the Golden Eval Set.
Notify me by email when the GitHub repo is live.
This uses Substack to subscribe you to Security in Public. You may be asked to confirm by email.
01 Grounded Confidence code
A naive implementation of confidence grounded in expert-labeled evals.
02 Simple Email Investigation AI Agent
A small agent for testing investigation behavior on phishing and BEC scenarios.
03 Golden Eval Set
The expert-labeled cases used to evaluate where the agent performs well and where it should abstain.