Anthropic says 80% of engineers now run self-improving agent loops

Redo · x · 2026-07-27

Anthropic’s Head of AI Lab says 80% of the company’s engineers are running self-improving agentic loops, and the team has started building Graphs for multi-agent orchestration over the past three months.

The post frames this as a practical shift away from linear prompting toward graph-based agent workflows, and points to a 1-hour podcast plus an article on “Graph Engineering with Claude” for more detail.

Related event: Anthropic Engineers Detail Self-Improving Agent Loops(5 posts)→

Original post →

More from coding & agent

coding & agent channel →