The Singularity Is Scheduled for 2045. Here Is the 2026 Scorecard

October 3, 2026. The technological singularity is the point at which machine intelligence improves itself faster than people can follow, so technology changes in ways nobody can predict. Ray Kurzweil, the idea's most famous promoter, put it on the calendar: human-level AI around 2029 and the singularity in 2045, a forecast he first made in 2005 and repeated in 2024. The term is suddenly everywhere again. Google's Keyword Planner shows US searches for "what is the singularity" up about 900 percent over the past three months, after a week in which the White House renamed AI "Super Intelligence" and Anthropic's chief executive wrote that AI is now helping build the next generation of AI. This is what the singularity means, where the dates came from, and how the 2026 evidence scores against them.

What the singularity actually means

The word is borrowed from physics, where a singularity is a point past which the normal rules stop working. Applied to AI, the standard definition runs like this: once an AI can improve its own design, each better version builds the next one faster, intelligence grows explosively, and the result is a superintelligence far beyond human understanding. Three people built the idea:

1. I. J. Good, 1965. The British mathematician described an "intelligence explosion": an ultraintelligent machine could design even better machines, so the first one would be "the last invention that man need ever make".
2. Vernor Vinge, 1983 and 1993. The science fiction author and mathematician popularised the word singularity, and in 1993 predicted greater-than-human intelligence between 2005 and 2030.
3. Ray Kurzweil, 2005. The Singularity Is Near predicted human-level AI around 2029 and the singularity in 2045, when he expects humans to merge with machine intelligence. His 2024 book, The Singularity Is Nearer, kept both dates.

Not everyone bought it. Paul Allen, Gordon Moore, Steven Pinker, Jaron Lanier and Roger Penrose are among those who have argued the explosion will not happen, usually because intelligence does not scale like transistor counts.

The singularity vs AGI vs superintelligence

The three terms describe a sequence. AGI is an AI that matches a capable person at most intellectual work; our AGI explainer covers how the labs define it. Superintelligence is an AI far beyond the best humans; see what super intelligence means, including the US government's new use of the phrase for all AI. The singularity is the event in between: the runaway period in which one turns into the other. DeepMind's June 2026 report From AGI to ASI names four possible routes for that period, including recursive self-improvement, which is the mechanism Good described in 1965.

The 2026 scorecard

Measured against Kurzweil's 2029 milestone, three years out, here is where the evidence stood on October 1, 2026:

1. Human-level on familiar reasoning tests: reached. On the ARC Prize leaderboard, Claude Opus 5.5 scores 98.5 percent on ARC-AGI-1 against a human panel's 98.0 percent, and GPT-6.1 Sol scores 94.2 percent on ARC-AGI-2 against the panel's 100 percent.
2. Human-level at learning new tasks: not reached. On ARC-AGI-3, where 100 percent means learning unfamiliar games as efficiently as people, the best standard score is 52.7 percent.
3. Self-improvement: started, by the labs' own account. In his September 2026 essay We Must Pace the Frontier, Anthropic's Dario Amodei wrote that since roughly the summer AI has been "advancing drastically faster, driven primarily by AI's growing ability to build the next generation of AI", that this recursive self-improvement is "starting to happen across the industry, including at Anthropic", and that it "could outrun our ability to understand and control these systems" if left unchecked.
4. Insider worry: rising. Anthropic safety researcher Evan Hubinger said in September there is a greater than 10 percent chance AI "could kill all humans" within the next decade, the BBC reported. OpenAI's Sam Altman and Elon Musk both backed Amodei's call to slow down.
5. Forecasters moving, in both directions. The team behind AI 2027, which depicted superhuman AI researchers by late 2027, has published a companion scenario, AI 2040: Plan A, in which humanity deliberately delays superintelligence until 2040; one author's median year is now 2030.

Kurzweil's early milestone is therefore closer than his critics expected in 2005 and further than his fans claim today. Machines match people on tests they have practised for and still learn new things slowly. The explosion requires the second to change.

What a singularity would mean for a business

By definition, nobody can plan for the event itself. You can plan for the run-up, which is already measurable: model prices falling, capabilities jumping every few weeks, and the leaderboard leader changing twice in a month. Three habits keep a company steady through that:

1. Buy results, not predictions. Measure any AI system on finished work per dollar today. Our guide to what an AI agent costs to run gives the method.
2. Automate the bounded jobs first. An AI receptionist for a home services company answers a defined set of calls to a defined standard. That is narrow AI, it works now, and it does not need 2045 to pay back.
3. Build so the model can change. If the labs are right that capability is compounding, the system you deploy this quarter should swap models without a rebuild. That is the standard our AI automation agency designs to.

The honest summary: the singularity is a hypothesis with a date attached by its optimists, not a measurement. In 2026 the people closest to the technology sound more like Good than like Pinker, and that, more than any benchmark, is what changed this year.