Builder in Residence at Amp, the AI coding agent built by Sourcegraph. 25-year veteran entrepreneur with multiple exits, including Treehouse (taught over 1M people to code). Creator of the AI Dev Tasks framework (6.9k+ GitHub stars).
From Chaos to Structure: Building with AI Agents
Ryan Carson has spent two decades teaching people to code and build companies. Now at Amp, he’s focused on a harder problem: helping developers structure their work so AI agents can actually execute it.
Current Work
At Amp (Sourcegraph), Ryan serves as Builder in Residence, bridging cutting-edge AI coding systems with real-world developer workflows. He created the open-source AI Dev Tasks framework, which structures feature development into three clear steps: PRD creation, task generation, and iterative execution.
His work emphasizes that process discipline amplifies AI capabilities more than raw model improvements. Featured in Lenny’s Newsletter, Freeplay Blog, and Creator Economy.
Background
Previously founded and scaled Treehouse, an online coding education platform that taught over 1 million students before being acquired by Xenon Partners in 2021. Named EY Entrepreneur of the Year (2015). Built and sold three developer-focused startups over 25 years before joining Amp to focus on practical AI developer tooling.
Philosophy on AI-Assisted Development
Ryan challenges the notion that better models alone solve coding problems. His approach centers on structured workflows:
Clarify Intent First - Transform vague requirements into explicit, measurable goals. AI can’t fix ambiguous specifications.
Decompose Before Executing - Break problems into discrete, AI-manageable task units. Large features need systematic breakdown.
Iterate & Verify - Manage the feedback loop between AI output and human validation. Step-by-step verification beats end-to-end generation.
Core belief: “AI doesn’t solve chaos—structure does. The bar for AI-assisted development isn’t model capability—it’s how well humans structure the work.”
AI Dev Tasks Framework
The open-source repository provides standardized task definitions that AI agents can parse consistently, context preservation for long-running projects, reproducibility for team collaboration, and progress tracking with measurable verification points. Works seamlessly with Amp, Claude Code, Windsurf, Cursor, and other AI coding assistants.
Conference Appearance
Event: AI Engineering Code Summit 2025 Date: November 21, 2025, 3:40 PM Session: “From Chaos to Code” Track: Leadership/Management (SDLC & Process)
Ryan presented his practical 3-step workflow for converting messy requirements into structured, AI-ready tasks. The session demonstrated how better specifications lead to better AI execution, regardless of model capability.