In brief: Users with domain expertise achieve comparable success rates with Claude Code as software engineers, regardless of their programming experience.
Anthropic analyzed approximately 400,000 Claude Code sessions between October 2025 and April 2026 to examine how agentic coding works in practice. Finding: success depends on domain expertise, not programming experience.
Anthropic conducted an analysis of approximately 400,000 privacy-preserving Claude Code sessions to evaluate the composition of tasks, human-AI collaboration, and success rates. The study reveals a clear division-of-labor pattern: humans predominantly make planning decisions (What needs to be done), while Claude predominantly makes execution decisions (How to do it).
The more domain expertise a person brings to a session, the more work Claude performs per instruction. On coding tasks, representatives from all major professional groups achieve similar success rates as software engineers: they accomplish what they set out to do, demonstrably confirmed through passing tests or committed work. The difference between intermediate and expert users is modest. Over the seven-month observation period, the proportion of time spent debugging fell by nearly half. Usage shifted toward end-to-end agentic deployment, code execution, data analysis, and non-code documentation authoring.
Simultaneously, the estimated value of typical tasks—determined by comparison with freelance job postings—rose over seven months across nearly all task categories by an average of approximately 25 percent. This indicates that Claude Code is increasingly handling more complex and valuable tasks. The share of GitHub projects with coding-agent activity has more than doubled since late 2025; Claude Code users spend an average of 20 hours per week with the tool.
For CTOs, this finding is relevant: agentic coding does not replace domain expertise; rather, it amplifies it. Teams with deep problem understanding leverage such agents more effectively. This means that focus on competency development in specific domains can be more valuable long-term than programming training alone. Technological implementation becomes increasingly delegable—strategic and domain-specific knowledge becomes scarcer.
Source: www.anthropic.com · Published June 17, 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrase and classification via Lumi News Pipeline v1.7.1.