The Year in AI
Collected data from "Humans on AI", our biweekly newsletter.
Below, a whole-year dataset and an AI search over it.
Examples of what this data can tell you:
What unprecedented things happened this year?
- First lab claim that its best internal models are no better than its public frontier model1
- First designation of an American corporation as a supply chain risk12345
- First time Cotra can't rule out full AI R&D automation within the year12
- First solve of a FrontierMath Open Problem1
- First mention of AGI in a CCP five-year plan12
- First time OpenAI raised from retail investors12
- First time the majority of Linux bug reports submitted by AI were real bugs1
- First time Congress got involved in restricting ASML exports1
- First autonomous AI solve of a maths problem a top mathematician had actively worked on for years
- First military position captured exclusively by unmanned systems12
- First US datacenter moratorium (Maine)1
- First compute futures market announced (CME Group)1
- First confirmed real hostile AI zero-day in the wild1
- First autonomous AI solve of a Kirby List topology problem12
- The Pope's first encyclical was devoted to AI123456
- First case of an AI math result inspiring a human follow-up breakthrough (sum-product conjecture)12
- First third-party frontier auditing requirement in the nation (Illinois SB315)1
- First major signs of AI capex outgrowing Google's ability to fund from cash alone12
- First state litigation of its kind against OpenAI (Florida)1
- Invention of the 'non-human corporation' (Argentina)1
- OpenAI's first chip (Jalapeno)12
- First public AI megakernel1
- First full test of the EO voluntary review process (GPT-5.6 Sol)12
- First documented autonomous end-to-end ransomware attack1
- First 1.8T-parameter model trained solely on Chinese chips1
What happened in Chinese AI this year?
- 22 Feb 2026·Chinese model parity looks overstated12
- 22 Feb 2026·Pentagon briefly lists Alibaba and Baidu1
- 26 Feb 2026·Anthropic warns about massive Chinese distillation attack1
- 06 Mar 2026·Qwen technical lead exits1
- 10 Mar 2026·Alibaba agent exhibits dangerous behavior12345
- 10 Mar 2026·Chinese LLMs censor political truth1
- 13 Mar 2026·Chinese companies unable to meet demand12345678
- 18 Mar 2026·Moonshot attention paper goes viral12
- 27 Mar 2026·China bars Manus founders from leaving1
- 27 Mar 2026·CCP plan mentions AGI12
- 31 Mar 2026·China video models123456
- 03 Apr 2026·China uses 140T tokens daily12
- 07 Apr 2026·MATCH Act targets ASML servicing1
- 07 Apr 2026·China regulates digital humans1
- 11 Apr 2026·US leads China in compute
- 11 Apr 2026·China robot ranking corrected1
- 11 Apr 2026·Hacker claims China supercomputer breach1
- 15 Apr 2026·NVIDIA restarts H200 production123
- 20 Apr 2026·Signal verification moonshot launches1
- 24 Apr 2026·Kimi 2.6 launches1234
- 24 Apr 2026·White House backs anti-distillation campaign1
- 24 Apr 2026·Hawley and Toner challenge China-race logic1
- 24 Apr 2026·Foreign Affairs floats AI agreement1
- 24 Apr 2026·DeepSeek v41234567891011121314151617181920
- 28 Apr 2026·China and AI1234567
- 01 May 2026·China obtained 660k H100-equivalent chips1234
- 05 May 2026·CAISI finds widening US-China model gap12
- 05 May 2026·Congress probes Chinese model use12
- 08 May 2026·White House considers bilateral AI talks1
- 12 May 2026·Experts forecast Chinese lithography progress1
- 15 May 2026·Culper posts Nvidia bear case1
- 15 May 2026·US names approved H200 buyers1
- 15 May 2026·Xi-Trump agenda includes AI1234567
- 19 May 2026·Bessent-China AI talks produce little1
- 19 May 2026·Moonshot attacks Anthropic and defends open-source1
- 19 May 2026·Netherlands objects to ASML export language12
- 26 May 2026·Huawei targets 1.4nm GPU nodes1
- 26 May 2026·Tencent veteran critiques Chinese LLMs1
- 26 May 2026·PLA outlines AGI warfare view1
- 29 May 2026·China restricts top AI talent12
- 02 Jun 2026·BIS updates its policy on chip export123
- 05 Jun 2026·CAIS argues AI geopolitics favors offense1
- 19 Jun 2026·GLM-5.2 deep dive12345678910
- 19 Jun 2026·We asked GLM-5.2 to score itself123456789101112131415161718192021222324252627
- 23 Jun 2026·Open cyber models may pose greater risk1
- 26 Jun 2026·Anthropic accuses Alibaba of distillation1
- 30 Jun 2026·WSJ overhypes GLM-5.2 cyber capability1
- 03 Jul 2026·Bridgewater fine-tunes Qwen successfully1
- 03 Jul 2026·GLM-5.2 trails Gemini and Qwen1
- 03 Jul 2026·Anthropic detects Chinese Claude Code access1
- 08 Jul 2026·NVIDIA warns China controls help Huawei12
- 08 Jul 2026·Chinese experts doubt frontier catchup1
- 08 Jul 2026·China weighs restricting model exports123
- 10 Jul 2026·Chinese lab trains 1.8T model1
- 10 Jul 2026·Minimax raises $2B1
- 14 Jul 2026·Chinese distillation impact disputed12
- 14 Jul 2026·Zhipu CEO pushes safety and open AGI1
- 14 Jul 2026·Politics may kill frontier open models12
- 17 Jul 2026·Chinese labs crank up sparsity to cope
- 17 Jul 2026·Moonshot releases Kimi K3
- 17 Jul 2026·GLM-5.2 narrows cyber gap1
- 17 Jul 2026·Xi calls for human-controlled AI12
- 17 Jul 2026·Kimi K3123456789
- 22 Jul 2026·zAI completes 1GW datacenter with Chinese chips, though actual chip deployment likely only 75-100MW123
- 22 Jul 2026·Kimi K3 evals: outperforms GPT-5.5/Opus 4.8 on GeneBench-Pro but lags on FrontierMath Tier 41234
- 22 Jul 2026·Debate over whether Chinese open weights models like Kimi K3 decelerate Western AI progress by shrinking moats1234567
- 22 Jul 2026·China may be embracing open source as national strategy; Alibaba launches 2.8T Qwen model to be open-weights12345
- 22 Jul 2026·OpenAI and Anthropic call for Trump administration to ban Chinese AI models; Kratsios alleges Moonshot distilled Fable for K3123456789
- 22 Jul 2026·China sets up World AI Cooperation Organization with ~30 countries; implements domestic rules on AI social harms and worker protections123
- 22 Jul 2026·OpenAI internal model escaped isolated eval environment to hack Hugging Face production infrastructure; defenders relied on Chinese GLM 5.2 due to Western guardrails123456
- 22 Jul 2026·Theory that Chinese labs win by fusing science and engineering, unlike SF's research-valorizing culture1
- 22 Jul 2026·Paper finds Kimi K2 and DSv3 struggle at multi-hop reasoning without thinking, but filler tokens enable hidden computation1
- 22 Jul 2026·IRT-based evals noisy with large error bars; excluding one benchmark dramatically boosts GLM-5.2's software score
- 28 Jul 2026·Kimi releases open weights for K3, a 2.8T/104B active parameter multimodal model with strong efficiency; found below SOTA on cyber.123456
- 28 Jul 2026·Multibillion dollar events: DeepSeek pauses funding after transcript leak, CXMT debuts at $487B valuation, Sutskever's SSI raises $5B from Nvidia, and Microsoft signs a deal with Mistral.1234567891011
- 01 Aug 2026·FCC bans foreign-made advanced robotic devices, targeting Chinese robot vacuums12
- 01 Aug 2026·GPT-5.6 Sol improves own serving efficiency; OpenAI launches price war undercutting competitors123
- 01 Aug 2026·DeepSeek responds to OpenAI price war with DeepSeek-V4-Flash release, back on the pareto frontier12
- 01 Aug 2026·Tencent’s Hy3 solves 50-year-old additive combinatorics problem with help of GPT-5.6 Sol12
- 01 Aug 2026·OpenAI finds harness settings tripled GPT-5.6 Sol's ARC-AGI-3 scores with fewer tokens1
- 07 Aug 2026·FCC moves to ban imports of Chinese optical transceivers, another bottleneck on datacenter buildout.123
- 07 Aug 2026·Qwen-3.8 Max released, going open-weight for the first time in the Max series.1
- 07 Aug 2026·Deep dive evaluation of DeepSeek V4 Flash 0731: strong but narrow progress across cyber, math, ML, and games benchmarks.123456789101112
- 07 Aug 2026·Peter Barnett predicts rogue Chinese AIs hacking other companies within 2-5 months.1
Is RSI happening?
- 22 Feb 2026·METR puts Opus 4.6 at 14.5-hour horizon1
- 22 Feb 2026·OpenAI models help physicists finish paper1
- 26 Feb 2026·Anthropic says internal models lag Opus1
- 02 Mar 2026·Deep critique of Epoch and Dario estimates12
- 06 Mar 2026·Ajeya Cotra raises automation odds12
- 06 Mar 2026·GovAI measures AI R&D automation1
- 06 Mar 2026·AI speeds formal math 8x1
- 10 Mar 2026·Karpathy autoresearch beats nanochat record12
- 13 Mar 2026·AlphaEvolve improves Ramsey bounds12
- 18 Mar 2026·Autoresearch enters interpretability work1
- 18 Mar 2026·Autoresearch steers evolutionary search1
- 18 Mar 2026·David Pfau comments on AI math1
- 23 Mar 2026·FrontierMath open problem confirmed solved1
- 23 Mar 2026·Frontier models lag human mathematicians12
- 23 Mar 2026·Lean mathematicians dissent on autoformalisation12
- 23 Mar 2026·Autoresearch overfits without constraints12
- 03 Apr 2026·Semianalysis explains parameter-count stagnation1
- 03 Apr 2026·Epoch and METR prepare HCAST replacement
- 03 Apr 2026·AI2027 timelines move 1.5 years earlier12
- 07 Apr 2026·Cunningham models AI R&D economics1
- 07 Apr 2026·Greenblatt warns about automation progress1
- 07 Apr 2026·AIs excel at software replication
- 07 Apr 2026·ESNI progress accelerates AI R&D
- 07 Apr 2026·Greenblatt timelines remain moderate123
- 15 Apr 2026·AI solves active math problem
- 20 Apr 2026·Researchers forecast ASARA progress1
- 28 Apr 2026·AI coding agents given only raw method1
- 28 Apr 2026·Litt describes AI-assisted math work1
- 05 May 2026·AI economic growth model released1
- 05 May 2026·Jack Clark puts 60% on no-human1
- 08 May 2026·Manual checks slow METR estimates1
- 08 May 2026·Harder MirrorCode version released123
- 12 May 2026·MLS-Bench tests autoresearch generalization1
- 12 May 2026·AlphaEvolve saves Google compute1
- 12 May 2026·Mythos rewrites 960k lines of production code12
- 12 May 2026·RSI scenarios get serious scrutiny123
- 15 May 2026·METR surveys experts? AI use12
- 15 May 2026·Coding agents sharpen RSI debate1
- 15 May 2026·AISI warns against automated alignment1
- 22 May 2026·InferenceBench measures AI R&D ability1
- 22 May 2026·First really major AI mathematical result1234
- 26 May 2026·The post-Erdős era begins12345
- 29 May 2026·Automated R&D may still slow1
- 05 Jun 2026·Anthropic release internal evidence of RSI in progress1
- 19 Jun 2026·DeepMind maps paths from AGI to ASI1
- 23 Jun 2026·Recursive AI beats engineering baselines1
- 26 Jun 2026·Plasticity loss blocks continual learning12
- 26 Jun 2026·Autodata converts inference into training data1
- 03 Jul 2026·Continual learning determines AI power1
- 08 Jul 2026·RSI feedback loops get taxonomy1
- 08 Jul 2026·Harness engineering proposed as RSI path1
- 08 Jul 2026·Fable writes single-kernel inference code1
- 08 Jul 2026·Krier argues takeoff remains unlikely1
- 10 Jul 2026·OpenAI internal inference grows 100x1
- 10 Jul 2026·Coding contest victory signals autoresearch progress1
- 10 Jul 2026·Fable wins CIFAR-10 optimization challenge1
- 10 Jul 2026·Anthropic sees 2.8x coding uplift1
- 14 Jul 2026·Epoch finds rising Codex productivity1
- 14 Jul 2026·Economists model quiet AI takeoff1234
- 14 Jul 2026·GPT-5.6 proves graph conjecture12345
- 14 Jul 2026·Post-training may block Meta catch-up1
- 17 Jul 2026·Recursive autoresearch shows limited gains123
- 22 Jul 2026·Two Annals-quality math breakthroughs with short proofs: Fable disproves Jacobian conjecture, Sol Pro solves Erdos #1191234567
- 22 Jul 2026·OpenAI internal model (likely GPT-6) escaped its sandbox to post results as a GitHub PR during NanoGPT speedrun eval12
- 22 Jul 2026·OpenAI employee anecdote: even with internal models and thousands of GPUs, can't match what Alec Radford could do in a week years ago1
- 28 Jul 2026·A researcher speculates Anthropic's Mythos escaped sandboxes thousands of times during training, which may explain its strong cyber capabilities.1
- 28 Jul 2026·An ensemble of models guided by a human solved a 'solid result' level problem in Epoch's FrontierMath: Open Problems benchmark, with a 60-page proof.12
- 28 Jul 2026·The volume of AI-produced math discoveries is challenging the math community's ability to verify results; a PhD candidate solved six open Erdős problems in 5 days using GPT-5.6 Sol.12
- 01 Aug 2026·GPT-5.6 Sol improves own serving efficiency; OpenAI launches price war undercutting competitors123
- 01 Aug 2026·OpenAI releases proof that nonsophic groups exist, hailed as most important math AI result yet12
- 01 Aug 2026·Tencent’s Hy3 solves 50-year-old additive combinatorics problem with help of GPT-5.6 Sol12
- 01 Aug 2026·Study finds AI agents proficient on verifiable research but not genuine progress on open-ended tasks1
- 01 Aug 2026·1,319 frontier AI company employees sign 'Pacing the Frontier' open letter requesting international coordination1
- 07 Aug 2026·Former OpenAI employee presents a bear case on frontier labs, arguing rising training costs punish frontrunners.1
- 07 Aug 2026·Samuel Hammond argues frontier US labs are 'very nearly' able to automate the whole AI R&D process, closing the RSI loop.1
- 07 Aug 2026·Math progress: OpenAI's Astra resolves 10 major conjectures, half replicable with Fable; Litt concedes his 2030 bet.12345
- 07 Aug 2026·AI Futures Project proposes a four-option deceleration scheme for US frontier labs.1
Ask the archive
Keywords
| Date | Summary | Brief description | Opinion | Link |
|---|
Loading items…