Open predictions and promises, sorted by the date we will grade them. The resolve-by date is our reading of the deadline the claimant gave (for “by end of 2026” we use 31 December 2026). Claims past that date stay listed until an editor grades them against primary evidence.
Past due, awaiting a verdict
- Overdue by 465 days About $80B on AI datacenters in FY2025 — Microsoft
- Overdue by 281 days "Several thousand" Optimus robots built in 2025 — Tesla
- Overdue by 281 days "Probably in 2025" an AI that can be a mid-level engineer — Mark Zuckerberg
- Overdue by 281 days Truly driverless cars limited to a modest number of cities in 2025 — Gary Marcus
- Overdue by 183 days AGI, defined as smarter than the smartest human, "within two years" — Elon Musk
October 2026
- Due in 21 days Most code for Meta's AI efforts written by AI within 12 to 18 months — Mark Zuckerberg
November 2026
- Due in 53 days AI will be an election issue in the 2026 US midterms — Gary Marcus
December 2026
- Due in 84 days First clinical trials of its AI-designed drugs by end of 2026 — Isomorphic Labs
- Due in 84 days 2025 will be seen as the year of "peak bubble" for generative AI — Gary Marcus
- Due in 84 days No country takes a decisive lead in the GenAI race in 2026 — Gary Marcus
- Due in 84 days Home humanoid robots "all demo and very little product" in 2026 — Gary Marcus
- Due in 84 days In 2026 AI will be able to make "very small discoveries" — OpenAI
- Due in 84 days Fully driverless Waymo rides in London in 2026 — Waymo
- Due in 84 days At least one model will match human experts across many industries before end-2026 — Julian Schrittwieser
- Due in 84 days Owners able to add their cars to the Tesla robotaxi fleet in 2026, "confidently next year" — Tesla
- Due in 84 days 2026 will likely see AI systems that figure out novel insights — Sam Altman
- Due in 84 days First 200 MW of Stargate UAE expected to go live in 2026 — OpenAI
- Due in 84 days A 10,000-person anti-AI protest in Washington, DC by late 2026 — AI Futures Project
- Due in 84 days The stock market will be up 30% in 2026, led by AI companies — AI Futures Project
- Due in 84 days Waymo One ready for riders in Washington, D.C. in 2026 — Waymo
- Due in 84 days Vera Rubin GPUs to ship in the second half of 2026 — Nvidia
- Due in 84 days 40% chance of AGI (replacing the majority of human jobs) by end of 2026 — Andrew Critch
- Due in 84 days Optimus "production design 2" to launch in 2026 — Tesla
- Due in 84 days Powerful AI "could come as early as 2026" — Dario Amodei
- Due in 84 days A new chip shortage in 2026 — Daniel Kokotajlo
- Due in 84 days AI will outperform humans at writing high-school essays by 2026 (median) — ML researchers (2016 expert survey)
March 2027
- Due in 155 days "At least 3-5 years away from automating software engineering" (from Mar 2024) — Bindu Reddy
- Due in 174 days Powerful AI systems in late 2026 or early 2027 — Anthropic
August 2027
- Due in 308 days Most white-collar tasks automated within 12 to 18 months — Mustafa Suleyman
December 2027
- Due in 449 days No AGI in 2026 (or 2027) — Gary Marcus
- Due in 449 days "Something pretty clearly superhuman in most respects by end of 2027" — Miles Brundage
- Due in 449 days 10%: AI solves or substantially helps solve a Millennium Prize Problem by 2027 — LEAP expert panel
- Due in 449 days By end-2027, models will frequently outperform experts on many tasks — Julian Schrittwieser
- Due in 449 days AGI "likely to arrive by 2027" — Roman Yampolskiy
- Due in 449 days Over 40% of agentic AI projects canceled by end of 2027 — Gartner
- Due in 449 days 2027 may see robots that can do tasks in the real world — Sam Altman
- Due in 449 days Rubin Ultra (NVL576) in the second half of 2027 — Nvidia
- Due in 449 days "Strikingly plausible" that models do the work of an AI researcher/engineer by 2027 — Leopold Aschenbrenner
- Due in 449 days 10% chance machines outperform humans at every task by 2027 — AI researchers (2023 expert survey)
- Due in 449 days Deployed "last 10 yards" delivery robots: not before 2028 — Rodney Brooks
- Due in 449 days Multi-task physical-assistance robot for the elderly: not before 2028 — Rodney Brooks
- Due in 449 days Generally agreed "next big thing" beyond deep learning: by 2027 — Rodney Brooks
- Due in 449 days Driverless taxi service in 50 of the 100 biggest US cities: not before 2028 — Rodney Brooks
- Due in 449 days AI will outperform humans at driving a truck by 2027 (median) — ML researchers (2016 expert survey)
January 2028
- Due in 450 days ~25%: 80%-reliability one-month task horizons by the start of 2028 — Ryan Greenblatt
- Due in 459 days "A 50% chance of AGI in the next 3 years" (from Jan 2025) — Shane Legg
May 2028
- Due in 571 days 30% of IBM back-office roles replaced over five years — Arvind Krishna
June 2028
- Due in 631 days METR 50% time horizons of about two weeks around the start of 2028 — Ryan Greenblatt
- Due in 631 days Bet: AI will do at least 8 of Marcus's 10 hard tasks by the end of 2027 — Miles Brundage
- Due in 631 days Bet (10:1 odds): AI will not do 8 of 10 hard tasks by the end of 2027 — Gary Marcus
July 2028
- Due in 662 days "There will be no programmers in five years" — Emad Mostaque
November 2028
- Due in 763 days Within five years you will "simply tell your device" what you want instead of using apps — Bill Gates
December 2028
- Due in 815 days AI will use 4% of US electricity in 2027 (7% in 2030) — LEAP expert panel
- Due in 815 days "Pretty confident" of AI that makes more significant discoveries in 2028 and beyond — OpenAI
- Due in 815 days A fully automated "legitimate AI researcher" by 2028 — OpenAI
- Due in 815 days 50/50 that AI does small-business taxes end-to-end by 2028 — Dwarkesh Patel
- Due in 815 days 75% chance of AGI (replacing the majority of human jobs) by end of 2028 — Andrew Critch
- Due in 815 days Individual training clusters costing hundreds of billions of dollars by 2028 — Leopold Aschenbrenner
- Due in 815 days At least 50%: AI builds a payment site from scratch, writes a hit-indistinguishable song and fine-tunes an LLM by itself by 2028 — AI researchers (2023 expert survey)
January 2029
- Due in 816 days ~15%: AI capable of fully automating AI R&D by the start of 2029 — Ryan Greenblatt
- Due in 835 days AGI "will probably get developed during this president's term" — Sam Altman
- Due in 836 days Stargate: $500B of US AI infrastructure over four years — OpenAI
- Due in 845 days Bet: AI will not get an A on 5 of his 6 latest midterms by January 2029 — Bryan Caplan
March 2029
- Due in 893 days AI will do well on any specified test set ("AGI") within 5 years — Jensen Huang
May 2029
- Due in 964 days AI will displace social drinking within five years — Jonathan Ross
June 2029
- Due in 977 days In five years there will be more software engineers than today, not fewer — François Chollet
- Due in 987 days AGI "one to 10 times smarter than humans" in the next three to five years — Masayoshi Son
- Due in 996 days Data-center AI accelerator market to reach $500 billion by 2028 — AMD
July 2029
- Due in 1000 days Binding regulation on the most powerful AI model developers — UK Labour Party / UK Government
December 2029
- Due in 1180 days Median 2029 for full coding automation (as of Feb 2026) — Daniel Kokotajlo
- Due in 1180 days No halving of US labor-force participation and no 15% GDP growth year by 2029 — Matthew Barnett
- Due in 1180 days LLM hallucinations "much less of a problem, certainly by 2029" — Ray Kurzweil
- Due in 1180 days Human-level intelligence and AGI by 2029 — Ray Kurzweil
- Due in 1180 days Dexterous robot hands generally available: not before 2030 (hopefully by 2040) — Rodney Brooks
- Due in 1180 days Home-navigating robot as an expensive product: not before 2030 — Rodney Brooks
- Due in 1180 days Long Bet: no computer will pass the Turing Test by 2029 — Mitch Kapor
January 2030
- Due in 1203 days Current LLM paradigm has a shelf life of "probably three to five years" — Yann LeCun
March 2030
- Due in 1250 days Bet: AI produces Annals-quality number theory papers for under $100k each within 5 years — Tamay Besiroglu
- Due in 1270 days Long Bet: AI will not lift US productivity growth above 1.8%/yr in the 2020s — Robert J. Gordon
- Due in 1270 days Long Bet: US productivity growth will average over 1.8%/yr in 2020-2029 — Erik Brynjolfsson
May 2030
- Due in 1328 days AI could wipe out half of entry-level white-collar jobs and push unemployment to 10-20% within 1-5 years — Dario Amodei
September 2030
- Due in 1427 days "In five years" unemployment levels never seen before: "Not talking about 10% … but 99%" — Roman Yampolskiy
December 2030
- Due in 1545 days Median 2030 for full coding automation (AI Futures Model) — Eli Lifland
- Due in 1545 days 20%: AI solves or substantially helps solve a Millennium Prize Problem by 2030 — LEAP expert panel
- Due in 1545 days By 2030 AI will autonomously fix issues, implement features and solve hard scientific programming problems — Epoch AI
- Due in 1545 days Frontier training clusters will cost over $100B by 2030, enabling ~1e29 FLOP runs — Epoch AI
- Due in 1545 days Few if any drugs approved by 2030 will have benefited from today's AI tools — Epoch AI
- Due in 1545 days By 2030 many scientific fields will have AI assistants comparable to today's coding assistants — Epoch AI
- Due in 1545 days Hyperion data centre: 2 GW online by 2030, scaling to 5 GW — Meta
- Due in 1545 days UK to expand its AI Research Resource compute "by at least 20x by 2030" — UK Labour Party / UK Government
- Due in 1545 days Claims of AGI by 2030 are "laughable" — Christopher Manning
- Due in 1545 days Human-level AGI most likely in 2029-2030, possibly as early as 2027 — Ben Goertzel
- Due in 1545 days "GPT 2030" will likely be superhuman at coding, hacking and math — Jacob Steinhardt
September 2031
- Due in 1806 days "Real AI" (beyond today's chatbots) arrives in about 5 years — David Krueger
December 2031
- Due in 1910 days ~30% of physics, materials science and medicine papers will engage AI by 2030 (from 3% in 2022) — LEAP expert panel
- Due in 1910 days AI will use 7% of US electricity in 2030 — LEAP expert panel
- Due in 1910 days Data centre electricity demand will more than double to ~945 TWh by 2030 — International Energy Agency
- Due in 1910 days A major city bans human-driven cars from part of town for driverless cars: 2027 to 2031 — Rodney Brooks
- Due in 1910 days AI will outperform humans at working in retail by 2031 (median) — ML researchers (2016 expert survey)
December 2032
- Due in 2276 days 50/50 that AI learns on the job like a human for any white-collar work by 2032 — Dwarkesh Patel
January 2033
- Due in 2277 days 45%: AI capable of fully automating AI R&D by the start of 2033 — Ryan Greenblatt
December 2033
- Due in 2641 days AGI "slightly after" 2030, not quite by 2030 — Sundar Pichai
- Due in 2641 days Generative AI could raise global GDP by 7% over ten years — Goldman Sachs Research
January 2034
- Due in 2654 days AI will displace about 40% of the world's jobs within 15 years — Kai-Fu Lee
June 2034
- Due in 2813 days AI "10,000 times smarter than humans" within 10 years — Masayoshi Son
December 2034
- Due in 3006 days Affordable home-navigating robot: not before 2035 — Rodney Brooks
February 2035
- Due in 3065 days Within a decade AI makes great medical advice and tutoring "free, commonplace" — Bill Gates
March 2035
April 2035
- Due in 3126 days Only about 30% chance of full remote-work automation within 10 years — Ege Erdil
June 2035
- Due in 3161 days Maybe 50% chance of AGI in five to ten years — Demis Hassabis
October 2035
- Due in 3296 days Truly capable AI agents are about "a decade" away, not "the year of agents" — Andrej Karpathy
November 2035
- Due in 3322 days Strong AGI expected "in the next 5-10 years" (from Nov 2025) — Daniel Kokotajlo
January 2036
- Due in 3372 days Only Waymo and Zoox will matter for US self-driving through 2035 — Rodney Brooks
December 2037
- Due in 4102 days 10% chance all human occupations are fully automatable by 2037 (50% by 2116) — AI researchers (2023 expert survey)
December 2043
- Due in 6293 days Transformative AGI by 2043 is less than 1% likely — Ted Sanders
December 2044
- Due in 6655 days AIs smarter than people "within probably the next 20 years" — Geoffrey Hinton
November 2045
- Due in 6988 days A system that learns as well as a human, then becomes superhuman: "5 to 20" years — Ilya Sutskever
December 2047
- Due in 7754 days 50% chance machines outperform humans at every task by 2047 — AI researchers (2023 expert survey)
December 2049
- Due in 8485 days AI will write a bestselling book by 2049 (median) — ML researchers (2016 expert survey)
December 2053
- Due in 9946 days AI will outperform humans at working as a surgeon by 2053 (median) — ML researchers (2016 expert survey)
December 2054
- Due in 10307 days "10% to 20%" chance AI leads to human extinction within three decades — Geoffrey Hinton