FEATURED ESSAY · 12 MIN READ
Autonomy Is Not the Absence of Process
Why reliable AI agents need bounded freedom, external verification, and systems that preserve intent beyond a single model session.
Research & Systems Architecture
Field work on governed agent systems, retrieval integrity, and agent-readable web infrastructure.
Axe 01 · Governed Agent Systems
How to grant agent autonomy without losing control
Read the Essay
CONCEPT PAPER · VERSION 1.1
Shokunin: A Governed Operating Method for AI Work
The complete methodological foundation: bounded authority, durable state, lean context, controlled process, independent challenge, and proof over promises.
REFERENCE IMPLEMENTATION · V1
Shokunin Harness Architecture
The multi-agent delivery harness implementing role separation, script-captured evidence, and Swarm HOLD pattern in production TypeScript.
Axe 02 · Agent-Readable Web
Making web information accessible, structured, and verifiable
EMPIRICAL BENCHMARK · 50,665 OBS
AUTOPSY V4 Benchmark
Empirical study of 2,000 domains and 50,665 public llms.txt links. Uncovering the discovery vs delivery gap in AI web retrieval.
OPEN PROPOSAL · RFC DRAFT
index-ai Open Specification
A verifiable open standard for agent-readable web content, introducing structured Agent Views, token economics, and publisher policies.
RETRIEVAL EVIDENCE SYSTEM
Agent View Platform
The developer tool and evidence platform for making AI retrieval observable, verifiable, and free of JavaScript rendering gaps.