Agentic coding and persistent returns to expertise

2026

Open publication workspace · Sign in to read the full PDF.

AI-generated summary

This report analyzes agentic coding sessions to understand how AI coding assistants are changing work, highlighting the persistent value of domain expertise.

* Analyzes ~400,000 Claude Code sessions to evaluate task composition, human-AI collaboration, and success rates.
* Finds a clear division of labor where users plan and AI executes, with domain expertise amplifying effectiveness.
* Observes a shift towards more end-to-end agentic use and an increase in task value over seven months.

Tags: AI, coding, expertise, agentic, labor market

Check the original publication for accuracy and context.