---
title: "Notes on agent discovery"
date: 2026-07-10
description: "What I've learned watching crawlers hit sites — and what actually matters for being discoverable."
---

Placeholder post. Replace this with a real one when you have something to say.

Some threads I want to pull on:

1. **`llms.txt` is the new sitemap.** Not because anyone standardized it, but because it works. Agents look for it.
2. **Markdown mirrors matter.** Serving `/writing/[slug]/raw` next to `/writing/[slug]` costs almost nothing and gives agents a clean substrate.
3. **The `.well-known/ai-catalog.json` file is where identity lives.** Think of it as your machine-readable business card.

The site you're on is running all three. Check `/agents` to see which crawlers noticed.