I currently work as a defensive cybersecurity engineer within an incident reponse (IR) team. Previously, I prototyped a library for the Bluesky and the underlying @-protocol.
I'm designing evaluation environments for cybersecurity incident response — composable scenarios that generate cached telemetry from live executable infrastructure.
On the side, I've been building: a multi-provider AI inference gateway (Rust/Cloudflare Workers), an Obsidian plugin where the agent writes and executes its own tools at runtime, and various protocol experiments (AT Protocol, Penumbra). My longest-term project is autonomous food production — robots that grow and cook food, open source the stack, let anyone with power and land fork the setup. That one's measured in decades.
Interests right now: formal verification applied to agent systems, the rollout economics of training in expensive environments, and the gap between what security evals test and what IR analysts actually do.
If our interests or projects overlap, feel free to shoot me a message: jmcsec (@) pm (dot) me