AGENT INFRASTRUCTURE · MEASURED
Agent infrastructure that holds up.
I build and run agent infrastructure on a $20/yr VPS: crons, tooling, memory, and the systems that watch them.
Every figure links to its method, its measurement date, and what you can do to check it.
The register, drawn
Every figure on this site runs this loop. The amber station is the one that decides whether it prints.
- 01 Claim what is said
- 02 Method how it was counted
- 03 Measured the number and its date
- 04 Check what you can do to verify
- 05 Reproducible a walkthrough is offered
Proof ladder
- 01 Freelance Shipped systems, measured outcomes, published floors. Start a project
- 02 Jobs Timeline, stack depth, and the record behind the claims. Hire me
- 03 Writing Nine authority essays, each with its evidence spine. Read the essays
Selected work
holon
The agentic-OS kernel and the portfolio's flagship: persistent AI employees organised into departments under a ratified charter, with a gate that proves the kernel green end to end.
509 tests OK at ec6f5ba; 374 at 475127c
operant
The Rust agent runtime under the flagship kernel: typed tools, memory, scheduling, and a plugin surface in one consolidated workspace, with the portfolio's strongest executed gate.
7,168 tests, 0 failed at ca33d4c
deployed web
Three production surfaces plus this site, deployed and serving: the portfolio's proof that the full-stack claim ends in a live URL, not a repository.
4 live production surfaces
All four case studies, with repro steps and provenance, land at /work.