SCALING LONG-HORIZON CODING AGENTS

Summary from the TLDR AI newsletter

A framework from Meta for test-time scaling in coding agents summarized past rollouts into structured representations, enabling better selection and reuse to improve benchmark performance.