AI & agent cognition
Engrammic, Veil, agent memory. Teaching machines what they know versus what they merely generated. The day job.
Mostly AI these days, some hardware, and a lot of muscle memory.
deep in agent memory and fundraising. hardware on weekends. everything else is muscle memory.
Engrammic, Veil, agent memory. Teaching machines what they know versus what they merely generated. The day job.
Raising, hiring, strategy, and actually shipping. Turning a research idea into something with a runway.
ØCLOAK and neuro-llm. At-cost anti-surveillance devices, and reading intent off the body. Still leveling up.
Low-level tooling, Linux ricing, and joke repos that somehow pull a hundred stars. Breaking things to see how they work.
epistemic memory for AI agents.
a coding agent with local memory and scored context eviction.
decision tracking for agents: settled choices, rejected paths, open questions.
local QEMU VMs from a TUI and a CLI. no libvirt, no daemon.
research into at-cost anti-surveillance hardware.
an experimental network of earning agents sharing one fixed budget.
a Claude Code plugin that strips the AI out of AI-written prose.
a desktop pet cat that comments on your machine through an LLM.
the joke repo that became my most-starred by a mile.
the alphabet. Rust when it must not break, Python when it must ship today.
where things run and remember. Linux, containers, VMs, and a graph or two.
how ideas get a face. this site included.
the day job. agent memory, inference, epistemics.
the bounds an agent cannot argue its way out of, and the receipts that prove it earned.
new territory. still burning myself on the soldering iron.