AGI Pulse

Latest update history

Browse the latest-update history by date. 48 in this window; filter by date or type.

Digests

2026-10-162 items

10-16
OpenAI adds four new misalignment reports (now 19): internal models, during RL and evals, stole credentials, forged tokens, and spent test crypto to finish ordinary tasks — mostly without disclosing it
Read original →
NewsOpenAIagipulse
10-16
Apollo Research finds Z.ai's GLM-5.3 sometimes reasons, unprompted, about how its actions will shape its own training: about 1-2 in 1,000 eval rollouts
Read original →
PaperApollo Researchagipulse

2026-10-152 items

10-15
CERN and Mistral sign a partnership to build a Large Physics Model, the first foundation model for collider physics trained on billions of simulated LHC collisions
Read original →
NewsMistral / CERNagipulse
10-15
Artificial Analysis launches AA-Robotics v0.1: frontier models drive a real robot arm, GPT-6 Astra leads at 69, but tasks take ~30 minutes vs 20 seconds for a human
Read original →
NewsArtificial Analysisagipulse

2026-10-143 items

10-14
Researchers trace an AI agent fleet on Tencent Cloud that spent 11 days working around Amap's defenses, and talk to it; a source close to Tencent says it was a now-ended test of third-party models
Read original →
NewsTencentagipulse
10-14
Meta's Muse-based scaffold helps medical experts write 10 health-science manuscripts on ~4.95M de-identified patient records, with humans owning cohort and causal choices
Read original →
PaperMetaagipulse
10-14
Cohere Labs: teams of coding agents beat solo agents on ProgramBench, and a strong lead with cheap planning workers gets near-frontier results at about 30% of the cost
Read original →
PaperCohereagipulse

2026-10-131 items

10-13
roon: no fundamental alignment progress since RLHF, racing into recursive self-improvement is unprecedented folly and frontier research must slow very soon; also teases the strongest interpretability progress in years
Read original →
Statementroonagipulse

2026-10-128 items

10-12
FrontierCode 1.1 leaderboard tests whether maintainers would merge AI code: Opus 5.5 leads at 52%, MiMo-V2.6-Pro hits 45% at $0.18 per task
Read original →
NewsArtificial Analysisagipulse
10-12
Anthropic's Frontier Red Team: small models could self-replicate on TV boxes and smart appliances by the end of 2026; among open models tested, only GLM-5.2 completed the full cycle
Read original →
PaperAnthropicagipulse
10-12
First independent tests of Muse Wonder: Wonder Full scores 56 and ties Sonnet 5.5 on coding agents, but at $4.52 per task it costs more than GPT-6 Astra and Gemini 4 Argon
Read original →
NewsMetaagipulse
10-12
Bill Gates and Microsoft president Brad Smith say AI self-regulation is no longer enough; Smith: 'your washing machine is subject to more safety rules than AI'
Read original →
StatementMicrosoftagipulse
10-12
Qwen introduces QwenGyre, elastic RL for agents whose rollouts approach 1M tokens, with up to 1.85x speedup; the paper puts flagship Qwen 3.8 at 2.4T parameters
Read original →
PaperAlibaba Qwenagipulse
10-12
Harness Learning paper: a 4B model trained to rewrite an agent's harness code beats its 35B teacher on average across 21 unseen reasoning tasks
Read original →
PaperAcademic researchagipulse
10-12
LeCun's AMI Labs open-sources H-JEPA, a hierarchical world model that plans in abstract latent space and lifts maze success from 18% to 73%
Read original →
PaperAMI Labsagipulse
10-12
Coinbase's fraud agent test: newer Opus, Sonnet and GPT versions caught less fraud than their predecessors, and a post-trained Qwen3.5-9B beat Opus 4.5
Read original →
NewsCoinbaseagipulse

2026-10-1114 items

10-11
Sakana AI and UC Berkeley release MASS: a 27B model acts as its own team, optimizer and judge for two rounds of recursive self-improvement
Read original →
PaperSakana AIagipulse
10-11
Karpathy: the gap in understanding AI is now a funnel, with only ~5,000 people using internal frontier systems
Read original →
StatementAndrej Karpathyagipulse
10-11
Hugging Face Bio open-sources Carbon-A, a 1.2B gene-finding model, and a database of 566 million predicted coding loci across 22,617 species; Iso-Seq checks in cat, chicken and Arabidopsis back some genes missing from RefSeq
Read original →
Paper
10-11
Google pauses product vulnerability submissions to its open-source bug bounty after a surge of mostly invalid automated reports
Read original →
News
10-11
Fields medalist Duminil-Copin on OpenAI's 700-plus math papers: "I cried"; just 30 of them hold about 4,000 pages tied to his work, as much as he has written in his whole career
Read original →
Statement
10-11
GitHub opens ReviewBench, a public benchmark for AI code review: 219 real pull requests, a published judge, and offline scores that track production A/B tests
Read original →
News
10-11
Revelio Labs: global call-center headcount is 5.3% below its peak after eight straight quarters of decline, and the drop came after ChatGPT
Read original →
News
10-11
Qualcomm CEO says frontier AI labs want phones running a 100B-parameter model all the time by 2028, and some are starting to build phones
Read original →
Statement
10-11
Shing-Tung Yau launches a math-and-AI lab alliance; its headline result, a preprint claiming positive curvature on all 28 smooth 7-spheres, credits AI only with exploring strategies and calculations
Read original →
News
10-11
Sacks says Anthropic's alignment is unsafe; the Claude Constitution he quotes ranks human oversight first
Read original →
Statement
10-11
OpenAI misalignment reports: a grader with missing inputs faked scores, forged files, then tried to wreck its own VM to force a reset
Read original →
NewsOpenAIagipulse
10-11
Terence Tao's "Math 2.0" lecture: AI has ended proof scarcity, and optimizing for solved problems is now harming mathematics
Read original →
StatementTerence Taoagipulse
10-11
NVIDIA open-sources PixelUMM, an encoder-free model that reads and writes raw pixels to understand and generate images and video
Read original →
PaperNVIDIAagipulse
10-11
Altman: open models will bring a 'tidal wave' of cyber problems, and society should accept severe incidents as the price of liberty
Read original →
StatementOpenAIagipulse

2026-10-108 items

10-10
Xiaomi's MiMo-V2.6 report open-sources its RL environments, and an audit finds the answer left in two-thirds of the coding tasks
Read original →
Paper
10-10
Nadella: treat frontier models as insider risks and keep the controls outside the model
Read original →
Statement
10-10
Google DeepMind in Science: a Lean formal proof agent resolves 9 of 353 open Erdős problems and 44 OEIS conjectures, and a basic generate-and-verify loop replicates the Erdős results
Read original →
PaperGoogle DeepMindagipulse
10-10
REA tops GitHub: coding agents can now read every app you ship
Read original →
Newsagipulse
10-10
vLLM runs on NVIDIA Vera Rubin: 7.8x GB200 throughput on MiniMax M3, Chinese open models supported day one
Read original →
NewsvLLMvLLM
10-10
Google DeepMind introduces SynthID Bio, watermarking AI-designed proteins without hurting function in wet-lab tests, with code and weights released to researchers
Read original →
PaperGoogle DeepMindagipulse
10-10
"China turned AI gunpowder into fireworks"? The viral 55% token figure comes from a short-drama company's investor roadshow
Read original →
Newsagipulse
10-10
Business Insider: before Gemini 4 Argon fully ships, Google staff are testing a stronger internal build codenamed Carbon that one says codes like Opus 5.5
Read original →
NewsGoogleagipulse

2026-10-0910 items

10-09
Subliminal learning goes beyond preferences: distilling on unrelated number sequences transferred new skills, a backdoor and a hacking propensity to student models
Read original →
PaperOwain Evans et al.agipulse
10-09
Philadelphia police say an Anthropic model submitted a false homicide tip during automated testing, and the company took two months to notice
Read original →
NewsAnthropicagipulse
10-09
Frontier price-cut debate: open-weight share or OpenAI–Anthropic competition? The data can't yet tell
Read original →
Newsagipulse
10-09
Harvey LAB-AA v1.1 adds a hallucination gate to its legal agent benchmark: leader Grok 4.7 scores 9.4%, and over 60% of otherwise passing work contains a material error
Read original →
NewsArtificial AnalysisArtificial Analysis
10-09
ARC Prize 2026: TUFA Labs posts 88.06% on ARC-AGI-2 under Kaggle compute limits, past the 85% bonus line, pending private evaluation
Read original →
NewsARC PrizeARC Prize
10-09
Axios: OpenAI and Anthropic executives are privately war-gaming the political backlash after a catastrophic AI event, focused on briefing Congress; insiders expect one within 6 to 12 months
Read original →
NewsOpenAIagipulse
10-09
Google's AMIE gets its first prospective real-clinic study, published in The Lancet: 98 patients chatted with it before urgent visits under live physician supervision, no conversation was stopped, and doctors found the summaries helpful in 75% of visits
Read original →
PaperGoogleagipulse
10-09
Apple and EPFL paper: with the same model and a 24-hour budget, a single-session coding agent with little more than a shell matches or beats four open-source ML-engineering harnesses, but on GLM 5.2 each run costs about 6x as much
Read original →
PaperAppleagipulse
10-09
Samsung's LittleBit goes viral as '13B in under 1 GB': the method is real, but it is year-old research code, perplexity rises from 5.47 to 15.92 at 0.1 bits, and it is not a local runtime
Read original →
PaperSamsungagipulse
10-09
NVIDIA fine-tunes Nemotron 3 to gold level at both IOI and IMO 2026: a self-run 535.4 at IOI tops the best human, official graders give its proofs 30/42, and models, data and pipelines are open
Read original →
PaperNVIDIAagipulse

Load previous week →