This category contains 6 tracked predictions. Each page includes the original claim, current evidence, counterevidence, and what would change our assessment.

Agents struggle with long-horizon tasks Confirmed
2025 and early 2026 Updated 2026-11-09 85% confidence in assessment

Agent-1 is bad at even simple long-horizon tasks (page 6, Early 2026 section). Also: agents in Mid 2025 are 'impressive in theory but in practice unreliable.'

AI model capable of autonomous self-replication Emerging
January 2027 Updated 2026-11-07 55% confidence in assessment

The safety team finds that if Agent-2 somehow escaped from the company and wanted to “survive” and “replicate” autonomously, it might be able to do so.

Best AI agents cost hundreds of dollars per month Confirmed
Mid 2025 Updated 2026-11-07 95% confidence in assessment

The better agents are also expensive; you get what you pay for, and the best performance costs hundreds of dollars a month.

Computer-using agents marketed as 'personal assistants' Confirmed
Mid 2025 Updated 2026-11-07 85% confidence in assessment

Advertisements for computer-using agents emphasize the term “personal assistant”: you can prompt them with tasks like “order me a burrito on DoorDash” or “open my budget spreadsheet and sum this month’s expenses.”

METR time horizon doubles every 4 months On Track
~4 month doubling from 2024+ Updated 2026-11-07 85% confidence in assessment

METR time horizons doubled every 7 months from 2019-2024 and every 4 months from 2024-onward (Appendix G, page 50). The acceleration from 7-month to 4-month doubling is a key claim.

Unreliable but useful AI agents emerge Confirmed
Mid 2025 Updated 2026-11-07 95% confidence in assessment

AI agents become increasingly useful for real tasks but remain unreliable on complex, multi-step workflows.