Solo Methodology

AI-driven solo development — how one person ships products using LLM agents, harness engineering, and automation.

Pages

• harness-engineering-summary — Harness engineering: agent mistake → fix harness, 3 components, 6 adoption steps

• agent-mistake-fix-harness — Core loop: CLAUDE.md → linters → structural tests. Ratchet effect

• context-engineering — Context as code: CLAUDE.md as TOC, progressive disclosure, dynamic context

• stream-six-layers — STREAM: principles as dependency graph, not flat list

• antifragile-life-design — Barbell for solo: 90% consulting + 10% products

• one-pain-one-feature-launch — Ship one thing in days, not platforms in months

• context-graphs-summary — Context graphs: agent decision trajectories as compounding asset

• decision-traces-compound — Each agent action improves future ones via precedent retrieval

• dev-principles-summary — 993-line reference: SOLID, TDD, DDD, Clean Arch, i18n, infra, code quality

• cli-first-testing — Every project gets CLI mirror, make integration mandatory

• schema-guided-reasoning — SGR: schemas → logic → UI. Agent reads schemas first

• agent-self-discipline — Drift detector, complexity thresholds, evolution = commit

Pending Ingest

• Schema-guided reasoning deep dive (from 1-methodology/sgr/sgr-complete.md)

Open Questions

• Optimal agent team size for solo dev?
• When to use subagents vs sequential work?
• How to prevent context window waste?

Key Sources

• 0-principles/manifest.md — Core values
• 1-methodology/harness-engineering.md — AI agent development
• 1-methodology/dev-principles.md — TDD, DDD, SOLID, Clean Arch
• 1-methodology/context-graphs.md — Agent trajectories