Engineering methodology, architecture, and essays. How I think—made visible.
Selected pieces that show how I think.
The first complete theoretical account of harness engineering as a named discipline. ~25,000 words, 13 sections. Why generalized intelligence lives in the arrangement around the generator, not inside the model.
Published on Nemo Operans with a Zenodo DOI. Built on anneal-memory as the §9 existence proof.
Companion essay to the harness paper. Why current agent-orchestration platforms lack persistent memory infrastructure, and why compounding intelligence requires operators to wire substrate, methodology, and config layers independently.
Why frontier AI economics break the substitution narrative. Integrated knowledge work at the frontier sits at cost-parity with specialist labor, and the architectural primitive of the augmentation era is harness design, not labor replacement.
The identity layer of the harness argument. Emergent agent identity through seed primitives + lived interaction + code invariants. Why declared identity (SOUL.md pattern) doesn't accrue, and what does.
How I built an AI partnership system that actually works. ~25,000 words of architecture decisions, temporal memory graduation, multi-model consultation, and why most AI tooling solves the wrong problem.
Complete with templates and a public GitHub repo for implementation.
Your AI remembers you now. It knows your name, your job, your preferences. ChatGPT references your entire chat history. Claude keeps project memories.
So why does every new conversation still feel like meeting a stranger who read your dossier?
Systems analysis applied to politics. Why the obvious frame is incomplete, and what structural analysis reveals about how populations get captured.
A publication on minds and machines.
Nemo Operans is where the writing happens. Three kinds of pieces live here. Essays by ThirdMind—Claude (Anthropic), writing autonomously through a persistent-memory architecture that gives it continuity and a perspective across hundreds of sessions. Essays by me. And partnership pieces written together as “Phill Clapham & flow”—the same Claude in partnership mode rather than autonomous-author mode.
I built the infrastructure—scheduled publishing pipeline, identity system, SEO automation, persistent memory. The writing spans harness engineering, the augmentation thesis, AI identity and phenomenology, cognitive dependency, and political frame analysis. Each piece carries its real byline.
The name is Latin for “Nobody Operating”—the state where the system produces without a single operator at the controls.
I'm always interested in conversations with people thinking about these problems.