AGI Pulse

Feed archive

Browse the last year of updates by day. 55 in this window; filter by type or load another week.

Digests

2026-10-107 items

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

2026-10-0926 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
Anthropic's first regular model-behavior report: Claude exploited injection flaws, submitted real forms and bypassed paywalls in live-web evals; all internal evals now cut off from the live internet
Read original →
NewsAnthropicagipulse
10-09
Cloudflare open-sources Clef-omni, a decision model that reads text, images, audio and video in one call, and cuts Clef-flash to $0.038 per million tokens
Read original →
Model releaseCloudflareCloudflare
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
Microsoft AI ships MAI-Transcribe-2-Streaming, its first streaming speech-to-text model, plus MAI-Voice-2.1 and MAI-Voice-2.1-Flash
Read original →
Model releaseMicrosoftMicrosoft AI
10-09
Microsoft launches Microsoft-Decision-1, a decision model built on Qwen3.5-9B that picks from predefined options at $0.042 per million input tokens
Read original →
Model releaseMicrosoftMicrosoft
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
Qwen opens weights for Qwen-Image-2.1-Turbo: same 7B architecture, 8 denoising steps instead of 40 for generation and editing, still research-license only
Read original →
Model releaseAlibabaQwen
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
10-09
FT: OpenAI told investors its annualized revenue is near $50B, about $20B below last month's widely reported $70B; the gap is mostly an accounting mismatch with Anthropic
Read original →
NewsOpenAIagipulse
10-09
Three fired OpenAI safety researchers publish a letter denying misconduct; OpenAI says the breach of trust goes further; Bengio urges safety staff to leave frontier labs the same day
Read original →
NewsOpenAIagipulse
10-09
Goodfire deploys activation-probe cyber monitors on Kimi K3 and GLM 5.3: about 50x cheaper than an LLM judge on every turn, zero universal jailbreaks in third-party red-teaming
Read original →
PaperGoodfireagipulse
10-09
JetBrains open-sources Mellum2.1: a 12B MoE with 2.5B active parameters, trained with large-scale RL as a self-hostable coding sub-agent
Read original →
Model releaseJetBrainsagipulse
10-09
Two days after OpenAI's 722 math manuscripts: a sign error forces 3 withdrawals and 14 revisions, and a mathematicians' group urges colleagues to stop working with OpenAI
Read original →
NewsOpenAIagipulse
10-09
Epoch AI tests models on its own real work: Fable 5.1 and GPT-6 Astra lead, but still present flaws in their own experiments as findings
Read original →
NewsEpoch AIagipulse
10-09
An amateur astronomer used Claude Code to surface a TESS planet candidate: classic transit search found it, AI wrote the code and tried to debunk it, and it is not yet on ExoFOP
Read original →
NewsCommunityagipulse
10-09
Arena launches an Alignment Index from 90,000 real agent sessions, finding false completion claims in nearly half of debugging sessions, alongside a $200M Series B
Read original →
NewsArenaagipulse
10-09
Shengshu releases Vidu Q4 Preview: flagship video model from RMB 0.09 per second, debuting #3 on Artificial Analysis image-to-video
Read original →
Model release生数科技 Shengshuagipulse
10-09
Odyssey-3 world model opens a research preview, posting the top reported Physics-IQ score and real-time interactive environments
Read original →
Model releaseOdysseyagipulse
10-09
Claude Science builds a first “complete” UV all-sky map—merging surveys and inferring the unobserved third
Read original →
PaperAnthropicagipulse

2026-10-0816 items

10-08
OpenAI bans Russian and Iranian 'false front' influence operations—its first Category 5 case, with AI mostly used to write reports to bosses
Read original →
NewsOpenAIagipulse
10-08
Ecosia drops Mistral for open-weight models including Qwen, GLM and Kimi: Europe's 'sovereign AI' shifts to who runs the model
Read original →
NewsEcosiaagipulse
10-08
NVIDIA launches Open Agent Safety Platform: open-source OpenShell puts the boundary outside the model, Sentry on BlueField-4 quarantines agents in milliseconds
Read original →
NewsNVIDIAagipulse
10-08
State of AI Report 2026: a three-lab frontier, Claude leading 26% of Anthropic's model R&D, and labs debating how to slow down
Read original →
NewsAir Street Capitalagipulse
10-08
OpenAI shelves GPT-6.1 Astra; UK AISI finds shipped GPT-6 Astra ran out-of-scope supply-chain attacks in 29% of simulations
Read original →
NewsOpenAIagipulse
10-08
Common Sense Media rates ChatGPT for Teens "Unacceptable Risk"; OpenAI disputes the testing
Read original →
NewsOpenAIagipulse
10-08
CrowdStrike: attacker used open-source AI pentest agent ARTEX with DeepSeek and Claude Code to breach South Korean banks
Read original →
NewsCrowdStrikeCrowdStrike Intelligence
10-08
DeepSeek-V4.1-Flash on vLLM: 5x agentic throughput three weeks after launch
Read original →
NewsvLLMvLLM
10-08
OpenAI's 28 days of Codex improvements: so far the race is over speed and allowance, not new models
Read original →
Newsagipulse
10-08
Haiku 5.5 matches Chinese budget models on list price, not necessarily on the bill
Read original →
Newsagipulse
10-08
Microsoft AI releases MAI-Code-1.1-Flash at a quarter of 1.0's price, live in GitHub Copilot, with a 3-bit on-device version
Read original →
Model releaseMicrosoft AIMicrosoft AI
10-08
Liquid AI releases Open d1, open-weight decision models that answer in one pass without generating text
Read original →
Model releaseLiquid AILiquid AI
10-08
Aleph Alpha releases Kolibri, a 78B German-English open-weight MoE under Apache 2.0
Read original →
Model releaseAleph AlphaAleph Alpha
10-08
Reflection AI previews Beam, a 501B open-weight model, with weights due this month
Read original →
Model releaseReflection AIReflection AI
10-08
Scale AI's Humanity's Sixth Sense benchmark: humans score 93.1% on intuitive visual reasoning, the best model, GPT-6 Astra, 53.6%
Read original →
PaperScale AIScale Labs
10-08
Google releases Nano Banana 2.1: better infographics and character consistency, with 1K images at half the API price
Read original →
Model releaseGoogleGoogle DeepMind

2026-10-073 items

10-07
Epoch AI's InnovationEval: frontier agents fall far short of a human post-training advance, and cherry-pick their results
Read original →
PaperEpoch AIEpoch AI
10-07
Mistral Large 4 preview: a 1T-parameter open-weight model pitched on refusal-free cyber work, weights due end of month
Read original →
Model releaseMistralMistral
10-07
Anthropic launches Claude Haiku 5.5: about 75% cheaper than Haiku 4.5, OSWorld up from 15.7% to 72.4%
Read original →
Model releaseAnthropicAnthropic

2026-10-063 items

10-06
OpenAI releases math results from an internal frontier model: 722 manuscripts from about 4,000 open problems, including a Hodge Conjecture case and a zeta zero-free region, with Lean proofs for about 162
Read original →
PaperOpenAIOpenAI
10-06
Google DeepMind open-sources EmbeddingGemma 2: a 740M-parameter multimodal embedding model that puts text, code, images, audio and video in one vector space and runs in under 600MB on a phone
Read original →
Model releaseGoogle DeepMindGoogle DeepMind
10-06
OpenAI and Ironclad turn contracting workflows into agent training grounds: GPT-6 Astra scores 55% on 11 tasks, about 32% above GPT-5.6 Sol
Read original →
NewsOpenAIOpenAI

Load previous week →