Anthropic AI Updates: June 17, 2026

1. Anthropic Research Finds Domain Expertise, Not Job Title, Drives Agentic Coding Success

Anthropic. Anthropic published economic research analyzing roughly 400,000 Claude Code sessions from October 2025 to April 2026, examining how people and agents split technical work. The study found a consistent division of labor in which users make about 70 percent of planning decisions on what to build while Claude handles roughly 80 percent of execution decisions on how to build it. Expertise mattered more than occupational background: every major occupation succeeded at nearly the same rate as software engineers, and verified success climbed from 15 percent for novice-rated sessions to 28 to 33 percent at intermediate and expert levels, with expert users triggering an average of 12 Claude actions per prompt versus 5 for novices. The estimated economic value of a typical session rose about 25 percent over the seven-month window, led by building, operating, and fixing tasks. Source