AI R&D Capabilities Rising Fast, METR Update Finds
METR’s Q4 2026 update says public evidence can’t rule out drastic AI R&D acceleration, with task time horizons doubling roughly every 3.6 months.
METR’s Q4 2026 update says public evidence can’t rule out drastic AI R&D acceleration, with task time horizons doubling roughly every 3.6 months.
METR added harder, open-ended tasks to its time horizon plot. Frontier models still improve fast on verifiable work but fall short on judgment-heavy tasks.
One engineer lays out how to run swarms of coding agents: ambitious goals, self-hosted infrastructure, blunt prompts and a codebase stripped of friction.