A Feed for Tiny Tools: Following, Updating and Trusting AI-Built Web Apps

The challenge: tiny tools have no RSS feed to track changes and updates. AI coding assistants have made it easy to build small, single-purpose web tools quickly, but no way to keep track of them. Simon Willison published dozens of tools on a single page, but how to follow them and know when they improve or become malicious was unclear.

Matt Webb posed the question in his April 2026 post Syndicating vibes. The solution is a JSON Feed of tools, each with an Install button. Tool_feed.py is a Python script that turns any tools index page into a subscribable, change-tracked feed. It extracts the list of tools, identifies each tool by its canonical URL, detects changes since the last run, checks for risky new behavior, and publishes the feed with a changelog per tool.

The script parses the page into a tree, groups linked children by signature, and scores the parent by its biggest group. It skips navigation menus and focuses on the real list of tools. Each tool page is fetched, fingerprinted, and compared to the previous version. If the hash changes, the tool gets a new version number and a changelog entry.

External hosts touched by the tool are collected and compared to the previous version. If new domains are found, the update is flagged and listed. The feed format is a plain JSON Feed 1.1, making it easy for any feed reader to subscribe. The script uses enums, frozen dataclasses, and immutable values to keep the code clean and efficient.