My build portfolio

Built, shipped, still running.

Everything here runs, or ran, in my own life or one of my companies. Each entry carries a build log with the failures left in, because that's usually the part worth reading. Where I can, I've linked the repo or the running thing. A portfolio you can't verify is just a claim.

Flagship work

Everything else

B4
FIG — PIPELINE

Compound Publishing House

● ACTIVE BUILD

Seventeen agents that take a book from topic discovery to a press-ready manuscript, and the governance layer I had to build after they went wrong.

ORCHESTRATION · CONTENT PIPELINES

FIG — SESSIONS

Local-first session analytics for Claude Code, built because I had no idea where my tokens were going.

OBSERVABILITY · DEV TOOLING · REPO ↗

FIG — FRAMES

Themed, schema-validated React compositions rendered to MP4s and stills from JSON props, driven end to end by an agent.

CONTENT PIPELINES · DEV TOOLING · REPO ↗

$ route --task extraction
→ tier: mid · gate: fresh-context
✓ done-condition verified

FIG — HARNESS

claude-ops

● MIT

The operating system I run frontier models on... routing, delegation contracts, review gates, and a job contract for automation nobody is watching.

AGENT OPS · ORCHESTRATION · REPO ↗

See the full setup on GitHub ↗

FIG — TILES

Station

● LIVE

One cockpit over a dozen backends, where an automation that goes quiet gets caught.

OBSERVABILITY · AGENT OPS

PIN ●●●●
FIG — DROP ZONE

A private publishing platform where a reader's comment becomes an agent's next task.

CONTENT PIPELINES · AGENT OPS

FIG — VOICE

talk

● ACTIVE BUILD

A realtime voice interface over my personal knowledge system — and the build that taught me what unobserved really costs.

VOICE AGENTS · MCP

The thinking behind these builds shows up in the writing. The harness they all run on is public: claude-ops ↗