Two meanings, one word
The timeline above deliberately mixes two uses of the word that are easy to conflate, and labels each one.
Research meaning. Intelligence that substantially exceeds human cognitive capability, often written ASI. When Safe Superintelligence Inc. names itself after it, or Meta says its vision is “personal superintelligence,” this is the sense they mean. It is a stated goal. No current system has achieved it, and nothing on this page suggests otherwise.
Federal label. Since Executive Order 14434 of September 29, 2026, U.S. executive agencies use “Super Intelligence” and “SI” where they previously wrote “Artificial Intelligence” and “AI.” The order defines SI by reference to the existing statutory definition of AI, so for now the label describes the same technology the law already defined. Companies that renamed products or divisions to match (SpaceXSI, SIForce) are following the label, not announcing a new goal. For every agency that has applied the label, including pages and titles below the bar for an event, see Which Federal Agencies Have Switched From AI to SI?
Every event on this page says which meaning applies. For the full distinction, read What Is Super Intelligence (SI)? and Super Intelligence vs AI.
Foundations. The idea is older than any organization that adopted it. Three works that defined it, I. J. Good’s 1965 paper, Vernor Vinge’s 1993 essay and Nick Bostrom’s 2014 book, open the timeline as foundations. They are shown for context and are not counted in the figures or the chart, which cover organizational adoption from 2023 on.
Noted but not logged
OpenAI. OpenAI has not renamed a division or product. Its three events are the 2023 “Governance of superintelligence” post on its own site, the formation of the Superalignment team that July, and the team’s dissolution in May 2024. Sam Altman has also used “superintelligence” in the research sense in personal essays since January 2025, and said in October 2026 that OpenAI has “no plans” to rebrand but “we have been talking about super intelligence a lot, and we will continue to talk about it.” Personal essays do not meet the bar for an event; a formal mission or division change would.
Anthropic, Thinking Machines Lab. Neither has adopted the term in a name, mission or product as of the last update. Anthropic’s stated target is “powerful AI.”
Methodology
What counts. An event requires an official act by the organization itself: incorporating under the name, naming a division or team, publishing a mission statement, renaming a product, or issuing a policy instrument. For government bodies, an executive order, an agency page change or an official website qualifies. A preprint or peer-reviewed publication from an organization’s own researchers is logged, and labeled as a publication rather than a renaming.
What does not count. Interviews, podcast remarks, personal essays and social-media commentary by executives do not create an event on their own, with one exception: a chief executive announcing an organizational or product rename on an official channel (as Musk and Benioff did on X) is logged, because the announcement is itself the act. Marketing copy that uses the word without a name, mission or product change is excluded. Third-party descriptions (“a superintelligence startup”) are never sufficient.
Every event carries a quote. Each entry shows the organization’s own words, verbatim, from the source. If we cannot find the sentence in which the organization used the term, the relevance is inferred rather than stated, and the event does not qualify. Quotes keep their original spelling and are never translated in place; translated pages show a translation beneath the original.
Reversals are events. A team dissolved, a goal walked back or a term dropped is logged with the same bar as an adoption, marked “Team dissolved” or “Reversal” and linked to the event it reverses. A record of adoptions alone would read as cheerleading.
Dates. The date of an event is the earliest verifiable public use we could find, not necessarily the first use that occurred. Dates are revised backward when earlier evidence appears. Where an organization’s current materials no longer use the term (Reflection AI), the event stays, because the record is of adoption events, not current branding.
Sources. Every event links to a primary source where one exists (the organization’s own page, the executive order, the Federal Register), and otherwise to a named news organization reporting the primary statement verbatim. Sources are listed with each event and, combined, in the references panel.
What the chart shows. One dot per event, placed on the date it happened, in the lane for the meaning it uses. Events that fall close together stack upward so none is hidden. The chart does not plot a cumulative total or a trend line: a running count of curated events would rise by construction and would add a vocabulary change to a company mission as if they were the same thing. The honest picture is the dots themselves.
No scoring. Events are ordered by date only. Position implies nothing about an organization’s capabilities, seriousness or standing.
Change log
- 2026-10-25: Added the formation of the White House’s Super Intelligence Force (October 4). See What Is the Super Intelligence Force?
- 2026-10-24: Added five federal-label events: the Justice Department memo (October 6), the State Department cable (October 19), the American SI Exports Program rename (October 20), the Navy’s NAVADMIN message (October 22) and the Marine Corps message (October 23).
- 2026-10-20: Every event now carries a verbatim quote. Added three foundations (Good 1965, Vinge 1993, Bostrom 2014), OpenAI’s 2023 “Governance of superintelligence” post, the Superalignment team’s formation (2023) and dissolution (2024), and Microsoft’s March 2026 statement. Reversals are now logged. The data is also available as CSV, under CC BY 4.0.
- 2026-10-20: Page launched, replacing the Terminology Tracker. Same fifteen events, now with a chart, per-event sources, and a JSON copy of the data.
- 2026-10-19: Terminology Tracker first published.
Corrections and additions, with a source, to corrections@sibuilders.io.