I validated hundreds of coding patterns with AI agents and kept watching them fail on documentation in the same ways. They'd hit long or auth-gated content and hallucinate an answer. They'd pull outdated API references because the docs site didn't signal what was current. They'd miss entire sections because the page structure made them invisible to automated retrieval.

These weren't model failures. They were infrastructure failures. And nobody was systematically studying them or proposing solutions.

So I started building the research and standards infrastructure to fix it.

What I've built so far:

• Agent-Friendly Documentation Spec — a 21-check specification defining what makes documentation accessible to coding agents, built from empirical observation across 578 coding pattern validations
• afdocs — an open-source CLI tool that audits any documentation site against the spec. Also pairs with a companion library to use in test suites for ongoing testing
• Agent Skill Report — qualitative analysis of 673 public Agent Skills, with findings on spec compliance across the ecosystem
• AE Shift — practitioner-focused commentary on what's actually changing in the agent ecosystem
• An automated research pipeline that tracks 20+ themes daily across RSS feeds, arXiv, and GitHub, producing synthesized analysis of emerging patterns

I also have two other ecosystem-wide analysis reports in progress, and additional agent-related reports and tools in development.

Why this matters now:

The agent ecosystem is forming right now. The standards, best practices, and quality signals that get established in the next year or two will shape how agents and documentation interact for a long time. This research helps that foundation get built on evidence rather than guesswork.

Why sponsorship matters:

All of this work is published openly and the tools are free. Sponsorship is what makes it sustainable. Your support helps me keep doing independent research, maintaining the spec and tooling, and publishing findings that benefit the whole ecosystem.