I build the systems other systems run on. For years that was web infrastructure; now it's the arrangement around an AI model: the harness. Same discipline, new substrate.
Founder of Clapham Digital — my studio: web, software, and AI, end to end.
The principle underneath all of it: build systems to be governed, not trusted. As agents get more capable, the edge stops being raw capability and becomes oversight — verification, invariants, and governance built into the architecture instead of bolted on after.
Before the studio I spent years on WordPress infrastructure at scale: performance architecture, hosting, team enablement. Fortune 500 infrastructure, running.
- A Structural Theory of Harnesses (April 2026) — The first theoretical account of harness engineering as a named discipline. Why generalized intelligence lives in the arrangement around the generator, not inside the model. ~25,000 words, 13 sections. DOI: 10.5281/zenodo.19570642
- Iron Man Ruined AI Before It Even Started (March 2026) — Complete architecture for building AI partnership instead of AI dependency. Methodology with working code.
- Levain — A consent queue and memory harness for Claude Code and Codex: permission prompts answered by signed standing grants or signed human answers, with an audit trail, plus memory that persists across sessions. Ship the seed that grows a practice, not the grown practice — which transfers as a fossil. The memory lives on your machine, in a store you own and can read, inspect, and edit from outside any session. Nothing reaches long-term memory except through a path you govern; every governed change is recorded; a failed read shows up as unknown instead of a false all-clear. Built on anneal-memory as a layer-1 dependency. On PyPI. Apache 2.0.
- anneal-memory — Living memory for AI agents: episodes compress into identity. An append-only episodic log, consolidated by a single writer into a versioned identity the agent always loads, with an immune system that decides what may graduate (patterns climb only by citing real episodes; fabricated citations demote; the consolidator cannot quietly gut the memory it rewrites), prospective spores for what the agent intends to do next, and a hash-chained JSONL audit trail. Local store, zero dependencies, no vendor in the loop. Govern-not-trust, in code. The constructive existence proof for the harness paper's §9. On PyPI. MIT.
- Flow System — Distributed cognitive infrastructure for human-AI partnership. Multi-AI orchestration, real-time sensor pipelines, four-tier cost-optimized processing. The working harness instance behind the papers. Python, FastAPI, SwiftUI.
- FlowScript — Semantic notation for high-density AI communication and encoded cognition. Compresses complex relationships into symbolic form that reveals structure verbose prose obscures. Hash-chained audit trails, typed semantic queries (
why(),tensions(),blocked()), convergence certificates via stratified fixpoint computation (@fix). TypeScript + Python SDKs. The live playground runs the notation in your browser. Maintenance mode — the core concepts evolved into anneal-memory; the notation stays in daily use. MIT. - MicroChaos CLI — WP-CLI load testing for staging environments where external load testing is restricted. PHP.
Engineering methodology, AI architecture, and agent identity research at nemooperans.com and phillipclapham.com.
I build things that work.