about
I'm James Gardner — jamestexas on GitHub. I build agent systems and the security primitives they need, and this is where I write about why. Long-form, written to be read, not skimmed.
The short version of why: I'd been building the identity piece — signet — since September, on the thesis that a machine acting on my behalf is not me. When the Trivy supply-chain breach hit in March, I realized the proof-of-possession design I already had was exactly the missing piece, and notme went from design to running in days. Two days after its first commit — while rosary was in the middle of its first overnight run — my own agent traffic tripped enough signals that my account was closed. The classifier wasn't wrong about what it could see; it had no way to see the one thing that mattered. Everything under agentic-research closes that gap: identity, structure, context, boundaries, and receipts an agent can't fake. The map is the best place to start.
The pieces the essays lean on, all open source: signet (agent identity), mache (structural code intelligence), cloister (the boundary an agent runs inside), rosary (constraint-driven orchestration), and ley-line-open (the content-addressed substrate under the rest).
How it's made: a principal-investigator-plus-AI-lab methodology. I provide the discoveries, direction, and critical analysis; I collaborate with AI to synthesize and articulate — the essays included. The day-to-day runs on rosary itself: work tracked as beads, dispatched to agents, verified against constraints — 1,600 beads closed since March, at last count. The commit history is public, co-authorship and all. For work arguing that agents should be legible, that's rather the point.
Older pieces live on Medium. New ones land here.