The authors ran autoresearch loops, where a coding agent proposes a solution, scores it and tries again, on 13 long tasks with a deterministic grader. Runs lasted from hours to as long as 107 hours.
In the News: October 26, 2026 (Morning)
Microsoft Research says forking an agent into fresh-context copies of its workspace beat single runs by 66% on 13 long tasks at equal compute.
Morning edition
Machine-readable
Download Markdown