Loading
Papers and notes on things I've been digging into.
Can LLMs actually patch security vulnerabilities automatically? I looked into it — here's what works, what doesn't, and what scared me.
A messy but effective way to combine Google Search with LLMs to reduce hallucinations. Tested it, measured it, and it actually worked.
A design proposal for secretless application access: a Git-native layer that lets workloads use credentials without permanently owning them.
Working on a few things. Will post here when they're done.