AI coding agentsAI Coding Agents Lift Pull Requests 35.7%, Harvard Study Finds
A Harvard study of 720,542 engineers at 655 firms finds AI coding agents raised pull requests 35.7% while resolved issues rose a borderline 13.7% and headcount showed no detectable change. Code review is the bottleneck.
November 7, 2026 · 10 min readspan of controlMeta Tried 50 Employees per Manager. Now It Wants Managers Back.
Meta built its Applied AI division at up to 50 employees per manager. Six months later it wants managers back. Size 2027 spans by the manager work AI did not absorb.
September 13, 2026 · 11 min readAI coding agentsClaude Matched the Patch. Qwen Overshot. Score the Scope.
A study of 14,922 agent trajectories across five coding benchmarks finds Claude succeeds by matching the human patch's file scope while Qwen succeeds by exceeding it at every scale. Equal pass rates buy unequal diffs, and your ticket template is the control.
September 3, 2026 · 14 min readAI coding agentsAgent Teams Hit 65 PRs a Week. Nobody Got Time Back.
Linear's first cross-company dataset tracked 6,887 teams for two years: those with a coding agent went from 21 pull requests a week to 65, those without went 8 to 10. Engineering time on triage, comments and review went up, not down.
August 19, 2026 · 12 min read