Skip to content
TheBestSuperintelligence

Methodology

How we weigh evidence

These standards apply to every guide on this site. They were previously repeated at the bottom of each article and now live here once.

A rigorous comparison begins by asking what would count as evidence. For an empirical claim, look for a clear task definition, a relevant baseline, disclosed evaluation conditions, and results that another party could reproduce. For a market or organizational claim, use filings, official documentation, and dated disclosures. For a forecast, identify the event being forecast, the time horizon, the probability, and the assumptions that could make the estimate wrong.

A common error is to combine metrics that answer different questions. Peak hardware throughput is not the same as sustained model-training performance; a benchmark score is not the same as reliability in production; a company's stated safety policy is not the same as independent evidence of safe behavior. These distinctions are especially important in a field where systems, products, and corporate strategies change quickly.

Key terms

  • AGI: Artificial general intelligence, a term used for systems with broad capability across many cognitive tasks; definitions vary.
  • ASI: Artificial superintelligence, a hypothetical level of capability beyond the best human performance across a broad range of domains.
  • Evaluation: A defined procedure for measuring a system against specified tasks and criteria.
  • Operational reliability: The degree to which a system performs its intended task correctly under real conditions, including edge cases and failures.
  • Scalable oversight: Methods intended to help humans evaluate or supervise systems when direct evaluation becomes difficult.

How we verify

Checked on 2026-10-20. A claim that something does not exist is only as strong as the search behind it, so each one below says what was searched and where the search stopped.

No generally accepted test for artificial superintelligence

Used in the models, companies, and timelines guides.

  • What we found: proposals, not a standard. "Levels of AGI" (Morris et al., Google DeepMind; arXiv:2311.02462, first posted 4 November 2023, latest version 24 September 2025) proposes levels of AGI based on performance and generality. Two preprints propose tests aimed at superintelligence itself: SuperARC (arXiv:2503.16743) and The Einstein Test (arXiv:2501.06948). Each is a research group's proposal, not an adopted standard.
  • What we did not find: a definition of, or test for, superintelligence adopted by a standards body or used across labs, in the sources we could read.
  • Where the search stopped: web search plus the three papers above. The ISO/IEC 22989:2022 terminology standard is paywalled and we did not read it, so we cannot say whether it defines the term. This was not a systematic literature review.

No ETF dedicated to verified superintelligence exposure

Used in the ETF guide.

  • Method: a full-text search of SEC EDGAR for "superintelligence" in fund filings (forms 497K, 497, 485BPOS, and N-1A) filed from 1 January 2025 to 20 October 2026.
  • Result: none found in the sources searched as of 2026-10-20. The search returned two filings, both by Themes ETF Trust (a 497K and a 485BPOS dated 26 June 2026). In both, the word appears in a description of Meta Platforms ("to advance its vision to deliver personal superintelligence for everyone"), not in a fund's name or objective.
  • Where the search stopped: a fund that describes itself with other terms would not appear in this search. We did not search fund-screener sites (ETF.com, ETFdb, Morningstar) or funds listed outside the U.S. "Verified superintelligence exposure" has no agreed definition, so the statement is that we found no fund claiming it.

No established lag between AGI and ASI

Used in the timelines guide.

  • What we found: the question is treated as open. A Metaculus essay by Marius Hobbhahn (2 October 2022) defines takeoff speed as "the length of the period between reaching roughly human-level AI and reaching superintelligence," arbitrarily calls "less than two years between AGI and superintelligence" a fast takeoff, and concludes that "the community is still very uncertain about takeoff questions."
  • Where the search stopped: one essay and our own search. We did not review forecasting databases or expert surveys for a published AGI-to-ASI interval.

Primary-source checks

Every figure in the comparison tables on the stocks, ETF, companies, models, and timelines pages comes from a filing, fund document, or lab publication fetched on 2026-10-20, and the table cell gives its as-of date. Some sites (OpenAI's pages and Metaculus) block automated requests; where we could only read an Internet Archive copy of a page, the sources table says so. Each source is named by filing and section in the page's sources table, and each quotation was checked against the text of the document. Where a source did not support a claim, we reworded or removed the claim. Fund holdings and weights change daily; treat them as dated snapshots.