A first-person field report on why output-only auditing misses the attacks that matter, and the trajectory-level checks I added to my agent stack to catch them.
Your Agent Returned the Right Answer. That's Why It Was the Worst Possible Outcome.
Related
Build and Deploy Java AI Agents with Google ADK
Set up Java and Google ADK on Debian 13, run an agent locally, test it with JUnit, and deploy it to Cloud Run.
The five primitives I run a whole company on
I run a consumer product company by myself. Physical inventory, two storefronts, multiple...
I built a local LLM that runs entirely in your browser. No install, no GPU, no server
A few months ago I got obsessed with a question: can you run a real LLM entirely inside a browser...