Ryan Carson is a software founder whose work spans developer education, conferences and AI-assisted development. He founded Treehouse and Carsonified and currently builds Untangle for family-law firms. His open-source projects turn specifications, task lists and product reports into workflows for coding agents.[3][4][5]
Start with AI Dev Tasks →
X Career
Carson says he earned a computer-science degree in Colorado before moving to the UK to work as a web developer. He later built Carsonified’s technology conferences and Treehouse’s online computer-science school. His current career history records Treehouse leadership from 2010 to 2021 and identifies Untangle, founded in 2024, as his current legal-technology venture. The same history lists a 2025–2026 builder-in-residence role at Amp and ongoing AI developer-community work with Intel.[2][3]
Contributions
From a feature request to reviewable work
AI Dev Tasks organizes development into a product requirements document, a detailed task list and incremental implementation. The point is to make the requested behavior explicit before asking an assistant to change code, then give the developer smaller changes to review. Compound Product extends that loop into ongoing product work. It reads performance reports, selects a priority, plans the change and produces a pull request for review. Its README credits earlier compound-engineering and agent-loop ideas on which the implementation builds.[4][5]
A small agent implementation to study
Code Editing Agent is a TypeScript implementation of an Amp tutorial by Thorsten Ball, known as mrnugget. It exposes a short set of file operations to a language model: list files, read a file and edit text. This makes the tool-use loop inspectable without a large application around it. Carson’s repository is an implementation and teaching example of that tutorial, useful for seeing how a model request becomes a concrete file operation.[6]
Tools, writing & talks
Tools
Start here
AI Dev Tasks
github.comA specification, task-list and implementation workflow for developing features with an AI assistant.
Start here to practice writing a clear feature brief and reviewing one task at a time. The Markdown workflow is readable without an account; trying it requires a coding assistant and a project you can edit.
Repository owner[4]Original source- Amp Autofix Examplegithub.com
Reference setup showing Amp AI detecting and fixing production errors automatically through GitHub Actions.
Catalog workOriginal source - Code Editing Agentgithub.com
A compact TypeScript implementation of an agent with file-listing, reading and editing tools.
Implementation of Thorsten Ball’s Amp tutorial[6]Original source - AI PR Reviewgithub.com
GitHub Actions workflow that runs AI code reviews on pull requests using Amp or Claude Code.
Catalog workOriginal source - Compound Productgithub.com
A workflow that reads product reports, selects a priority and implements a planned change as a reviewable pull request.
Repository owner; implementation built on credited prior methods[5]Original source - ClawChiefryancarson.com
Opinionated skills that turn an OpenClaw agent into a chief of staff for your day, projects, and follow-ups.
Catalog workOriginal source