Observed arrival · 2026-11-02
Frontier Watch’s AI evidence board
An independent AI observatory tracks measured capabilities, documented incidents, policy signals, and news, including a timeline comparing the AI 2027 scenario with public evidence.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Readers tracking AI capabilities, safety, and policy
- Worth noticing
- Its METR task-time estimates come with a stated limit: estimates above 16 hours are unreliable with this task suite.
Field notes
The capability monitor uses METR task-time estimates for selected frontier models and notes that estimates above 16 hours are unreliable with this task suite. Its outcome watch groups selected evidence as beneficial, concerning, or unresolved, while warning that these counts are not probabilities. A separate timeline marks AI 2027 milestones with labels including “Partial evidence” and “Unverified.”
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
ƒJavaScriptBrowser-side code central
One card from the complete issue