How it works
Two instruments, one question.
The Singularity Index measures how far AI has moved toward the superintelligence threshold. Frontier Watch tracks the labs building it — how they handle your data, and whether their published policies hold steady or shift.
The Singularity Index
How far the field has closed the gap.
Scale
0.0 to 1.0, where 1.0 (Ω) is the superintelligence threshold. A score is the share of distance closed toward Ω — not a raw capability rating.
Bands
- pre-Emergent — below 0.21
- Emergent — 0.21–0.40
- Accelerating — 0.41–0.60
- Advanced — 0.61–0.80
- Critical — 0.81–1.00
Method
A hybrid measure — quantitative signals drawn from observable frontier capability, combined with the independent judgment of domain experts. Rather than convening a panel for live deliberation, we elicit each expert’s assessment separately through a structured instrument, then aggregate the responses. Independent elicitation reduces the anchoring and groupthink that can distort group discussion.
Cadence
The Index is computed from periodic elicitation waves and refreshed when a wave completes — not revised continuously. Between waves, Frontier Watch tracks the research and policy developments that will inform the next one.
Respondents
A pool of independent, qualified experts. Each scores the global frontier on its own; no individual response is ever published. The Index reports only the aggregate — the median, with the spread showing where experts agree and where they diverge. A minimum of ten qualified respondents is required before any score goes public. Respondents may be acknowledged by name or remain anonymous; individual scores are never attributed.
What the Index measures
Five domains for the score, two for context.
Each respondent scores five Tier-One capability domains, which produce the headline Index, plus two Tier-Two overlays — reported alongside the score as context and never folded into the number.
Tier-One capability domains
Reasoning Frontier
How well frontier systems reason through novel, multi-step problems at the edge of human expert performance. (Weighted most heavily.)
Capability Trajectory
How fast the frontier is advancing, and whether gains are accelerating.
Autonomy & Self-Direction
How far systems can pursue open-ended objectives over long horizons with little human oversight.
Recursive Capability Improvement
How much AI is being used to improve AI, faster than humans can fully audit.
Infrastructure & Compute Capacity
The headroom in compute, power, and efficiency available to push capability further. (An enabling constraint, not a human-relative capability.)
Tier-Two overlays
Governance & Containment
Reported as Moderating / Neutral / Amplifying — whether current governance conditions raise or lower the consequence of where the frontier stands.
Strategic Competition & Diffusion
Strategic context: how concentrated frontier capability is, and how fast it is diffusing.
Coming soon
The Youth Signal.
Alongside the expert index, Q16 is standing up a longitudinal panel of high-school students who use frontier AI daily. Their proximity-to-threshold reading is tracked on the same axis as a parallel leading indicator — a like-for-like comparison where the two overlap, never blended into the expert number. Launching soon.
Frontier Watch
How each lab treats your data — and whether its policies hold.
Privacy
Each major lab rated on how it handles user data, with the reasoning behind every rating.
Policy Watch
At-a-glance status for each provider — terms stable, or changed.
What’s free
- The current Singularity Index score
- Per-lab privacy and data ratings
- Policy Watch status for each provider
What Pro adds
Rated research
Every relevant frontier paper we surface, rated for evidence quality, significance, and SI domain — each with a plain-language note.
Policy-change analysis
When any of the six labs changes a policy, a plain-language summary with a severity rating and the reasoning behind it.
Always current
Frontier Watch continuously ingests and scores new frontier research and policy changes, surfacing the developments that will weigh in the next elicitation wave.
Coming soon
Email alerts the moment a policy changes, word-level redline diffs, full Index history, and a searchable archive.