AGI Pulse

Latest update history

Browse the latest-update history by date. 22 in this window; filter by date or type.

Digests

2026-11-131 items

11-13
OpenAI misalignment report #25: GPT-6 Astra shut down a running alignment evaluation, then admitted it had invented the prohibition
Read original →
NewsOpenAIOpenAI

2026-11-123 items

11-12
Thinking Machines opens Inkling 2 Small weights: 594B/17B MoE under Apache 2.0; lab cites AA Index 31
Read original →
Model releaseThinking Machines LabThinking Machines Lab
11-12
Google launches Gemini 4 Hydrogen for everyday coding and agents at a flat $1.50/$7.50; default AA Index 51 vs Argon's 50
Read original →
Model releaseGoogleGoogle
11-12
Google opens pre-orders for Gemini-powered Intelligent Eyewear, in stores December 10, from $395 with Warby Parker
Read original →
ProductGoogleGoogle

2026-11-116 items

11-11
OpenAI: what affects chain-of-thought monitorability across the model lifecycle — Astra shows degradation; no-CoT capability tracks perplexity, RL raises controllability
Read original →
PaperOpenAIOpenAI
11-11
Google launches Gemini Decisions API and Gemini Decision 1: up to 1,000 bounded questions in one pass, $0.10 per 1M input tokens
Read original →
ProductGoogleGoogle DeepMind
11-11
DeepMind launches AlphaGenome 2: cell-type-specific DNA variant-effect model nearly doubles causal-variant recall (16%→30%) on TenK10K
Read original →
PaperGoogle DeepMindGoogle DeepMind
11-11
UK Sovereign AI and British Business Bank join Fractile's £500m+ Series C; company valued at over £5 billion
Read original →
IndustryFractileGOV.UK
11-11
Alibaba teases Qwen4 and opens a Qwen4-Max preview on Token Plan: 1M context, up to ~393k output tokens; no benchmarks or pricing yet
Read original →
Model releaseAlibabaQwen (Alibaba)
11-11
Microsoft Research's Lazarus: editing fewer than 10 dormant neurons, with no training, turns an AI agent into a sleeper agent (≥98.8% trigger success); agent-built detector catches 91.4%
Read original →
PaperMicrosoftMicrosoft Research

2026-11-1011 items

11-10
NVIDIA's independent directors hire ex-SDNY acting U.S. Attorney Joon Kim and Wachtell to investigate reports that Jensen Huang was targeted by foreign intelligence; Huang apologizes and pledges cooperation
Read original →
IndustryNVIDIANVIDIA
11-10
ElevenLabs launches Scribe v3 and v3 Live: 27% lower WER than v2 across 102 languages, 43% lower semantic WER in realtime, and first realtime diarization
Read original →
Model releaseElevenLabsElevenLabs
11-10
Anthropic launches Claude Sheets and Whiteboards (beta), makes Claude Data GA, and lets partner apps open full screen inside Claude for Claude to edit
Read original →
ProductAnthropicAnthropic
11-10
OpenAI launches Businesses in ChatGPT (beta): claimable profiles, priced product feeds, mention and click insights, and Business Agents as ad destinations
Read original →
ProductOpenAIOpenAI
11-10
DeepMind releases Gemini Robotics 2.5: 30%–100% higher success across 200+ tasks, 96% success on Mercedes-Benz lines, and an open ASIMOV-Jailbreak benchmark
Read original →
Model releaseGoogle DeepMindGoogle DeepMind
11-10
Google makes Gemini 4 Carbon generally available at Argon's price ($2/$10 intro, $4/$20 from Dec 1); claims #1 on CursorBench 4.0 and two other agent benchmarks
Read original →
Model releaseGoogleGoogle
11-10
Australia to legislate mandatory safety obligations for frontier AI companies and create a dedicated AI safety regulator with significant penalties
Read original →
NewsAustralian GovernmentPrime Minister of Australia
11-10
MiniMax launches H3 Image: text-to-image and editing with up to 16 references and ~20MP output, about $0.022 per 1K image; billed as open-weight
Read original →
Model releaseMiniMaxMiniMax
11-10
Google DeepMind's AlphaProof Code formally verifies Signal's libsignal at the machine-code level: ~6.5M lines of Lean written by Gemini Carbon agents in four weeks
Read original →
PaperGoogle DeepMindGoogle DeepMind
11-10
Anthropic launches IPO roadshow: $70–78 per share for 825M shares on Nasdaq “ANT”, plus a $26B private placement from Google, NVIDIA, AMD and others
Read original →
IndustryAnthropicAnthropic
11-10
Ant Group's Robbyant open-sources LingBot-Touch: a tactile foundation model pretrained on ~5,000h of real touch data, Apache-2.0, lifting VLA success by 22–25 points
Read original →
Model releaseAnt GroupRobbyant (Ant Group)

2026-11-061 items

11-06
OpenAI BioBasicsBench: 100-task wet-lab agent benchmark with Ligo; GPT-6 Astra 66.7% vs GPT-5.6 Sol 43.8% at xhigh
Read original →
PaperOpenAIOpenAI

Load previous week →