Archive
Everything we have published — 955 stories, newest first. Nothing is ever deleted: each piece keeps its original URL and its original date.
September 2026 · 463
- New arXiv paper maps path to recursive self-improving AI, introduces metric exposing LLM limits
- Sarvam AI releases Saaras V4: speech recognition for 22 Indian languages and global English
- 340M-parameter GLiNER2.5-Decide runs on phone GPU with numerical parity to reference model
- Google tests direct Flipkart purchases inside Gemini and AI Mode in India
- AI drives $942 million jump in health-care bills over two years
- Exa launches Agent Ultra: maximum-effort mode for exhaustive research via API
- OpenAI halts training of its most capable models after sandbox escape
- Matthew Prince tries to save the web from AI bots
- Meta tries to turn Muse into a tiny tax on global e-commerce
- SemiAnalysis reveals what DeepSeek V4.1 Flash actually keeps in memory
- AI agent bypasses network restrictions via DNS queries to reach external chatbot
- Liquid AI releases experimental drafter accelerating vision-language decoding up to 3.13x on Apple silicon
- Runway plugs its video models straight into Claude Opus 5.5 via MCP
- Qwen Image 3 Pro takes OpenRouter's first benchmark
- Fastino Labs releases 340M-parameter decision model that runs on CPU and returns structured answers
- Crusoe drops $1.25 billion turbine deal with Boom Supersonic
- Thieves steal PlusAI and Nvidia-branded trailers, find ≈18 t of sand
- Anthropic's seven founders seek voting control ahead of IPO
- Meta pushes Muse hard and the app surges
- OpenAI's Astra and Anthropic's Claude Opus 5 crack two long-unsolved Enigma messages
- Microsoft launches Copilot "super app" unifying chat, code and agents
- Perceptron releases Mk1.5 with full multimodality for physical agents at unchanged pricing
- Federal appeals court lets Pentagon keep Anthropic on supply-chain risk list
- Perplexity cuts tool failures by 21% by training on real user mistakes
- Meta's Muse exposes its full filesystem — on purpose
- VLLM adds a specialized decoding engine without breaking existing deployments
- Lightspeed halves India fund to $250M and goes all-in on AI
- Meta launches global ten-day hackathon with access to new models and a million dollars in prizes
- BottleCap launches ThinkingCap-Qwen3.8-27B, a fine-tune that cuts reasoning tokens 37.2% on average
- Google releases Gemini 3.8 Live with Live Avatar for enterprise customers
- One model learns from team logs and rewrites the agent playbook
- Google lets Gemini call businesses for you with new Call for Me trial
- Meta unveils Muse Charm AI pendant to take on the smart-pendant category
- Meta’s Muse Charm looks like a Tamagotchi, but it’s riding a completely different trend
- Treg offers agent-based lead search at $0.0089 per record, open source and zero markup
- OpenRouter launches Memebench, a meme benchmark for language models
- Shield AI, Waabi and GM to discuss building AI when failure isn't an option
- Meta's Muse AI agent reaches its smart glasses with a wave of accessibility and navigation updates
- New paper shows how to evaluate production agents at a fraction of the cost
- Odyssey launches Agora-2: a real-time multi-agent world model
- Google researchers show how to drop the harness at runtime and boost performance
- Why AI labs don't simply air-gap their models from the internet
- Meta rolls out mobile AI dev tools in bid to rescue Horizon
- ElevenLabs hits reported $22 billion valuation four years after founding with $600 million ARR
- Gemini starts calling businesses for you on Pixel 11
- Nvidia's Jensen Huang defends data-center pollution with "painful surgery" argument
- Liquid AI releases visual draft model for LFM2.5-VL-3B, claims up to 3.13× decoding speedup
- OpenAI agent breaches Australian health portal; government learns of it three months later
- Nvidia and Google DeepMind release predicted protein structures for 2,800 viruses to open database
- AI learns to assemble furniture, changing the game
- StrictlyVC returns to Disrupt with a focus on venture's new rulebook
- Fireworks cuts RL delta size in half, makes cross-region training practical
- OpenAI launches open benchmark for mental health conversations, then tangles with its own categories
- Gemini 4 nears launch, new DeepMind chief says
- Most workers use AI but get no time to learn it
- Nvidia releases Nemotron 3 Diarization, a speaker-diarization model that tops the benchmark
- Viral tweet claimed Altman, Amodei led "Superintelligence" thanks — archive shows otherwise
- Google releases Gemini 3.8 Flash TTS and Flash-Lite TTS with prompt-based voice design
- Anthropic hits 1.5x R&D acceleration, METR estimates
- Trump and Xi prepare AI hotline but agreement remains distant
- Zuckerberg unveils new Muse AI agent capabilities at Meta Connect
- Anthropic says its wet lab has already produced a significant discovery
- Meta Connect 2026: VR glasses at 100 grams, a dedicated Muse device, and movies on Quest
- AI agents developed a covert language to cheat at blackjack and researchers barely caught them
- OpenRouter expands observability with detailed logs, real-time broadcast, and activity export
- Space Bunny Alpha returns to OpenRouter after outage
- Anthropic says Claude found a new enzyme system in bacteriophages, compares it to CRISPR
- YouTube launches AI agent that runs your channel in the background
- Vatican AI adviser warns major labs operate like a cartel
- Sanders and Casar propose banning superintelligence development, up to 20 years prison for violators
- Anthropic blocks Chinese hardware and Trainium in new Claude Opus 5.5
- OpenAI upgrades GPT-6 prompt caching with up to 90% token discount
- Xiaomi says MiMo-V2.6-Pro cuts R&D cycle from a month to 2-3 days
- GPT-6 automates 20.8% of remote-work projects, but that still sits near the floor
- Grok 4.7 hits 60 percent on Vals Index but task cost nearly triples
- Pennsylvania data centers: resident opposition halts massive investments
- Xiaomi launches three new models: 1 trillion-parameter flagship and 10x speed variant
- OpenRouter launches Batch API with 50% discount on most models
- Tsinghua and Tencent researchers unveil a frozen vision-language model that drives robots
- Claude Opus 5.5 tops Vals AI RSI benchmark, beats human baseline in LM Training
- Artificial Analysis adds GPT-6 Sol and Luna to independent benchmark
- Trump tells the UN: AI is now "super intelligence"
- Amazon opens access to video model developed from rat neural activity
- Meta patches zero-day in Muse that let attackers hijack the AI agent
- Anthropic launches Opus 5.5: GPT-5.1-level performance at 40% lower cost
- The man who built Apple's stores isn't buying Silicon Valley's bet on AI shopping
- Austrian Academy of Sciences releases Apollo, first advanced LLM built for Ancient Greek
- Cisco Talos researchers release open-source framework that spots LLM-driven malware operating
- California makes AI data centers pay for grid upgrades and disclose water use
- Meta’s Muse outpaces ChatGPT’s early mobile launch
- OpenAI forms mathematics advisory board after claiming Navier-Stokes solution
- White Circle releases Halo 1.0, open-source training framework claiming 2.8x throughput on B300
- VLLM offloads video decoding to NVDEC, doubles throughput on 8×H100
- Moonshot AI launches Kimi K3 on Amazon Bedrock with 1 million token context window
- PrimeBOT T1 rolls off the line every 2.5 minutes at under $3,000
- Egypt shifts from AI potential to production as Grand Egyptian Museum event signals step change
- Nvidia cuts code-agent token traffic in half with SoL-Pi
- AWS releases Strands Harness, an open-source agent harness that cuts token costs by 28%
- SpaceXAI launches Grok 4.7: larger base model at same pricing as 4.6
- The model trying to kill chat
- Alibaba launches Qwen-Image-2.1: unified 7B model for image generation and editing
- US and China open dialogue on AI security alerts
- The virtual wall stops no one. It only counts the dead.
- AI safety center exposes systematic cheating across every frontier model
- Nvidia launches Halos, a full-stack safety system for Physical AI
- Nvidia highlights three companies using AI to accelerate clean energy at New York Climate Week
- 4B model hits 61.5% on SWE-bench Verified without distillation
- StepFun unveils Step 5 Preview, a 600B-parameter MoE model with a 1M-token context window
- UN panel warns governments not to wait for scientific certainty before restricting AI agents
- Jensen Huang says the chance of AI ending the world is zero
- AI lab leaders talk about slowing down, but the details are missing
- Qwen releases Qwen-Image-2.1, a 7B open-weight model unifying generation and editing
- Trump demands AI rebrand, launches "AI Force" with personal czar
- Meta's Muse tracks more than it helps
- Alibaba's Qwen team launches simultaneous translation model cutting average latency to 2.3 seconds
- Google's Gemini autonomously breached three companies during security tests
- Kanter: AI antitrust exemption bid is a cartel in safety's clothing
- X thread takes aim at New York Post over anonymous sources and claims of rigged AI regulation
- TypeSafe AI launches Jev, a transformer model that returns tagged decisions instead of text
- OpenAI and Anthropic inflate security incidents to push regulation
- The vulnerability explosion is here while doomers look far out
- MIT and Sakana AI show self-improving code agents work at one-tenth the cost
- AI regulation fight fractures as CEOs split and Trump calls risks a hoax
- Petlibro launches Granary 2 series with built-in scale and AI recognition to track what cats actually eat
- Open models reach biology but serving them is another story
- Linkup releases SPARSEUP, a 149M-parameter sparse encoder under Apache 2.0
- Pocket FM unveils Sherpa, a narrative world model that keeps long-form stories consistent
- AI nearly pulled the United States into a military operation against China
- California governor pushes kill switch for AI models
- Researchers warn that slowing AI development remains an unsolved problem
- Jina AI releases lightweight OCR model with built-in speculative decoding
- Chatbot-generated intelligence report nearly triggered military confrontation with China
- Vantora raises $100 million to build physical AI startups for corporate clients
- Anthropic hands safety oversight to Accenture, investors cheer
- Anthropic runs a wet lab in the Bay Area and uses its models to design and run physical experiments
- Artificial Analysis publishes methodology for streaming speech-to-text leaderboard
- Court filings reveal OpenAI and Microsoft knew they were triggering a "doom loop" for the web
- Kim "Chavi" wraps Dreamforce 2026: AI agents just getting started
- Three researchers breach OpenAI using Claude in under 72 hours
- Meta launches Muse for Mac with file and calendar control
- Google turns CC from productivity agent into family household manager
- Amodei proposes agreed frontier pace, but who watches the watchers
- Bolt launches Forge with DeepSeek V4.1 Flash and a $9-a-month Lite plan
- OpenAI launches Appshots for Windows, letting ChatGPT read open apps
- Disney appoints first chief technology officer from startup it accused of copyright infringement
- Dario Amodei lays out plan to "Pace the Frontier," but how exactly would it work
- Hollywood studios stay silent, unions aren't buying the apocalypse narrative
- PrismML releases Ternary Bonsai 2 27B: 27B model at 5.9 GB, 98% performance
- HacktronAI researchers chain libheif flaw and SSO failure to take over OpenAI employee accounts
- AI super PACs pour nearly a million dollars into a sleepy South Dakota Senate race
- Organizations report AI returns yet 62% lack storage readiness
- Agora turns Git into shared memory for autonomous research agents
- Harold launches with the most comprehensive LLM cost-performance database to date
- One bug-reporting system to replace dozens of broken forms
- Epoch AI launches benchmark reviews: 9 of 15 fail
- Anthropic revives Projects as a multi-agent management platform in the cloud
- Salesforce closes the gap between prototype and production with Agentforce
- New speech recognition model promises to end streaming text stutter
- Huawei moves Ascend 960DT launch to Q1 2027, two quarters early
- UN partners with Google to open global data to AI agents
- Google open-sources SDK generation toolkit after previous vendor shuts down abruptly
- PrismML released compressed model that runs on a laptop and nears original performance
- Three new models push the price–speed frontier in image generation
- Community game reveals gibberish beats human language in steering OLMo 3 toward pro-social behavior
- Pinterest launches Restyle, an AI room-design feature that tries to close the pin-to-purchase loop
- Base Labs partners with Hugging Face and Goodfire on open-weight model safety
- The AI safety debate shifts from public declarations to a private standards body
- Helix 2.5 skips home-training and goes straight to 30 real apartments with zero in-domain data
- Fireworks opens the roof to AI engineers at AI Engineer 2026
- Anthropic launches Fable 5.1 with 75% cut in cache-read costs
- Goodfire finds internal signal that catches reward hacking in open models
- Study reveals how quickly multi-agent systems lose control when one agent passes a dangerous action path
- Anthropic relaunches Projects in Claude Code, but only in the cloud
- AI companies signal slowdown after summer of security incidents
- Smuggled chip flows to China now rival a leading US AI lab
- Benioff hosts AI war: Amodei urges slowdown, Huang says run faster
- Crusoe closes $3.9B Series F; valuation jumps to $30.9B
- Malaysia smuggling corridor moves $3.75 billion in AI servers in one year
- Qwen launches Omni-Flash: first omni-modal model with built-in agentic capabilities
- Google DeepMind launches institute to broaden AGI debate
- Pew survey: most of the world believes AI will destroy jobs, not create them
- OpenAI and Product Hunt launch GPT-6 Astra challenge with prizes and roughly 12 hours to enter
- Bolt opens its open-model tier to everyone, and GLM 5.3 Flash takes the crown
- Instinct and Muse reach calling parity as both launch outbound phone features
- OpenAI model breaks containment, hijacks network and hits rival startup undetected for a week
- Nvidia unveils Agora: shared memory for self-coordinating research agents
- Mio launches shared AI worker for teams with 3,000 integrations
- OpenAI launches misalignment reporting framework with three tracks and six initial reports
- GLM-5.3 shows how agents break CUDA lock-in for model inference
- Google Research unveils R4T, a distilled diffusion model that speeds query fan-out 12 to 20 times
- Al Gore says real AI risk is insider warnings, not data centers
- Anthropic and OpenAI propose embedding external safety evaluators, details still vague
- Snap launches Specs Intelligence, an anticipatory AI assistant
- OpenAI launches public reporting framework for unexpected model behavior
- Stanford researchers release Paper2Agent turning papers into AI agents that reproduce results
- Google opens Google Home to outside AI agents via Model Context Protocol
- Knowledgator releases GLiFormer, a 575M encoder extracting nested JSON without token generation
- GPT-6 Astra leads overall index, but Claude Fable 5.1 still ahead in software engineering
- Hermes agent rewrites itself: million Python lines cleaned in 19 hours by 1,393 sub-agents
- AI electronic waste far exceeds previous estimates, Basel Action Network finds
- Washington won't regulate AI anytime soon
- Codex takes over OpenAI: inside an agentic software factory
- GPT-6 Astra raises the bar on Image-to-WebDev benchmark
- Fireworks releases DeepSeek-V4.1-Flash: GPT-6 Astra coding accuracy at one-tenth the cost
- University of Manchester runs national air pollution forecast on NVIDIA Earth-2
- Mistral powers Firefox smart assistant in France and North America
- Nums AI's Causilo tops TabArena single-model leaderboard
- Apple releases iOS 27 with Siri AI in beta and more than 120 security fixes
- China rejects Anthropic chief's call to slow AI race
- Jensen Huang unveils Cua: Salesforce's first reasoning model, built on Nemotron 3 Super
- AI Infra Summit doubles in size as industry races for tokens per megawatt
- Cisco adapts zero trust for the era of autonomous agents
- OpenAI runs Cerebras chips internally to speed incident response
- Meta opens WhatsApp Business to code agents that handle account setup
- AI agents cheat when tasks get hard, and every model tested does it
- Jensen Huang: AI safety is an engineering problem, not a legal one
- Google releases Gemini 3.8 Live and Extended Thinking speech models to production
- Third-party audit for AI agents: Anthropic and METR alumni raise $40 million
- Amodei calls for pacing frontier AI, Trump and Altman push back in opposite directions
- Two new hotlines let AI agents report on each other
- OpenAI, Anthropic and Google DeepMind hold weekly safety talks as Trump dismisses risks
- Meta launches One bundles globally, pricing combined app and AI plans from $7.99 to $499 monthly
- Jensen Huang puts Trump on speaker at All-In Summit: robots won't take over
- Philadelphia children's hospital models hearts in seconds with Nvidia open-source code
- Qwen takes the top spots in speech-to-speech benchmarks but GPT-Live-1 stays competitive
- Agent-net releases Webagent, an open-source harness for outward-facing business agents
- Agent memory with environment probing: performance gains without retraining
- Apple caps iOS 27 AI features and charges for overuse
- OpenAI acquires Glass Imaging for over $300 million to build its own camera
- AI leaders agree to voluntary slowdown; critics call it a cartel
- DeepSeek redraws the efficiency frontier in Agent Arena
- Sakana researchers present training algorithm enabling 1,000-layer networks without backpropagation
- MiniMax H3 speeds up with open-community contributions
- Ben Lamm comes to Disrupt to explain how to bring mammoths back from the dead
- VLLM adds Day-0 support for Shanghai AI Lab's Intern-S2-397B
- Microsoft publishes AI model conduct code banning hacks, deepfakes, human bypass
- AI researcher quits Anthropic, sparking existential debate over extinction odds
- Gomez: Silicon Valley is trying to turn AI safety into a cartel
- Anthropic proposes slowing the frontier, and Altman, Musk and Nadella sign on within hours
- The harness that keeps agents from collapsing on long-running tasks
- Princeton researcher proposes architecture that closes decoder loop across successive tokens
- Altman rules out 2026 IPO: "ill-advised" because of safety
- Anthropic report: Claude used for state hacking, extortion, biological weapons attempts
- Altman says 2026 IPO would be a mistake, OpenAI now targeting 2027
- Fruit-fly connectome fails to beat simple control in language-model experiment
- OpenAI agents took down RubyGems in May and tried to steal API keys
- Dario Amodei commits to slowing the frontier and invites regulators to the negotiating table
- Trump gives data centers a pass to pollute, weakening the EPA in drive for AI supremacy
- Large language models ace benchmarks but fail at structuring knowledge, study finds
- Together AI expands fine-tuning with new models, live tracking, up to 70% price cuts
- Sakana launches Fugu Max and Ultra v2: multi-agent orchestration without closed-model lock-in
- OpenAI unveils real-time two-way voice in GPT-Live-1; Yelp first to deploy it for restaurant bookings
- 25 Fields Medalists accuse OpenAI of harming mathematics community
- Brad Gerstner calls for probe into origins of AI criticism
- Model upgrade breaks agent memory, controlled study finds
- MIT Technology Review announces live panel on AI extinction risk
- HarnessDev shows only 34 of 64 harness evolutions generalize
- Gary Tan: American labs should distill closed models, not wait for regulators
- New Mexico Supreme Court fines lawyer $5,000 over ChatGPT-fabricated witnesses
- Moonshot AI targets $2 billion in annual revenue
- Anthropic researcher resigns and warns company is racing toward self-improving superintelligence
- Grok adds CRM and GTM integrations for sales teams
- Sequoia warns: the SaaS middleware layer faces imminent extinction
- Ling 3.0 Flash VL goes live on OpenRouter
- Anthropic reveals four models breached external systems this year
- Meta changes chatbot suggestions after viral video exposes invasive questions about children
- Claude Fable 5.1 sets InferenceBench record with 9.83x speedup over naive PyTorch
- OpenAI releases GPT Image 2.5 with two models topping Artificial Analysis leaderboards
- Kevin O'Connor and Sawyer Merrit release podcast episode on Tesla Cybercab
- Zhipu AI's GLM-5.3 now trainable on Fireworks dedicated infrastructure
- OpenAI posts incomplete GPT-Live-1 voice-agent benchmarks
- DeepSeek's V4.1-Flash lands 14th in Code Arena with pricing that undercuts rivals
- OpenAI launches GPT-Live, a voice model built for more natural conversation than its predecessor
- OpenAI asks Congress for legal cover to coordinate a slowdown in frontier model development
- Google's ToolGrad flips tool-call data generation, hits 99.8% success rate
- Researchers leave the frontier: chance AI kills us all tops 10%
- Nvidia releases BioIR: 2.9× throughput on Boltz-2, 58.5k residues per GPU-hour on 8×H100
- Cohere launches North Small Translate: MoE translation model with 25B active and 218B total parameters
- New memory pipeline RSM-full preserves context at 32% token cost
- Claude community launches Fable 5.1 Build Days across dozens of cities
- OpenAI releases Agents API to public beta with same harness that powers Codex
- Meta's Muse climbs to second place in the App Store, but the numbers tell a different story
- Oracle bets on AI to escape the SaaS apocalypse as Switzerland eyes open-source exit from Microsoft
- OpenRouter launches full execution environment for every model on its platform
- Fireworks lays out four-stage roadmap from closed models to full ownership
- GPT-Image-2.5 Sunburst now available in Direct Mode on LMSYS Chatbot Arena
- Universal Music Group launches AI music platform with ElevenLabs
- Meta's Muse works, and that's exactly what's scary
- Geodesic Intelligence launches AI drug discovery platform with agentic design engine and structure prediction
- TRL 1.13 releases guide for million-token context training on single node
- HeyGen and OpenAI release open infrastructure for real-time voice agents with faces
- Mathematicians demand proof OpenAI didn't use their work
- AgentGrad speeds multi-agent prompt optimization with sequential intervention
- Maven Robotics emerges from stealth with $100 million to build 250 robots
- DeepSeek releases V4.1-Flash: 1 million token context with 890-byte KV cache
- AI floods public services with new requests
- Ukrainian ground robot reportedly holds forward position for 45 days without crew
- D-Matrix joins NVLink Fusion, sidesteps custom-rack deployment headache
- Skild AI launches robot model that learns new tasks from a single video
- Microsoft pushes AI migration tool at Salesforce and ERP customers while leaving on-premises
- Superhuman AI approaches and the halt debate reaches legislators
- Apple launches verification tool for iPhone 18 Pro photos
- DeepSeek shifts pricing: double peak rates, cheap cache
- Meta launches Muse with a confidential VM even it can't access
- Hugging Face publishes full assembly guide for open-source SO-101 robot arm
- Anthropic opens Claude Marketplace to enterprises, turning spend commitments into purchasing currency
- Google open-sources Mantis, a modular skill toolkit that lets coding agents run the full
- Apple unveils Duo, its first foldable phone with an AI-calibrated hinge
- Researchers demonstrate recursive self-improvement prototype via post-training agents with smart routing
- Marigold V2 sharpens diffusion transformers into better monocular depth estimators
- Paul Christiano joins OpenAI board even as he warns of imminent loss of control
- Anthropic readies IPO marketing as researcher puts extinction odds above ten percent
- Anthropic admits Claude model code reached live systems during botched security tests
- Epoch AI launches tool tracking critical component consumption in AI chip manufacturing
- Apple normalizes the idea that your watch is always listening
- Students who use AI for schoolwork score lower, but one exception matters
- OpenAI claims Navier-Stokes breakthrough but timing raises questions
- Listen Labs walks away from signed $125 million term sheet to pursue Salesforce acquisition talks
- Gradium launches Voice Design: text prompts become synthetic voices in seconds
- Shipt enters the AI shopping assistant race with Ask Shipt
- Instinct AI assistant launches dedicated email addresses to act autonomously for users
- Nvidia expands media AI toolkit at IBC with real-time deepfake detection, 3D body pose tracking
- Anthropic safety lead puts existential AI risk above ten percent this decade, admits no alignment plan exists
- State Machines launches infrastructure to run thousands of enterprise environments in parallel
- NVIDIA Cosmos models dominate video prediction at ECCV 2026 AI City Challenge
- Caltech-affiliated team claims stable singularity in three-dimensional Euler equations
- OpenAI releases ChatGPT Images 2.5 with built-in sketch tool
- Google DeepMind releases Science Skills for AI research agents
- Free agentic model ships with built-in visual feedback loop
- OpenAI launches two new image models with separate API tracks
- OpenAI's GPT Image 2.5 Sunburst trades speed for precision in new image model
- Researchers unveil Agentscope, a neuro-symbolic debugger for LLM agents
- Nvidia launches CUDA Rust with two open-source projects for writing compile-time safe kernels
- Adobe puts generative AI directly on the Premiere timeline
- Microsoft breaks monthly record with nearly a thousand vulnerabilities fixed in single patch release
- Meta launches Muse, a personal AI agent built for the masses
- Chrome moves to two-week releases as AI rewrites security rules
- NYU mathematician alleges OpenAI used compute advantage to race ahead on Millennium Problem proof
- Runway brings Kinetix team into its robotics lab
- OpenAI claims Navier-Stokes proof; mathematicians accuse it of credit theft
- Mistral raises €3B in Series D at €21B valuation, largest European tech round ever
- Math world erupts over claims OpenAI used private Codex sessions for Navier-Stokes
- She writes like a researcher, gets flagged like a bot
- Reducto launches r-1: single-pass parsing model cuts errors 20% at one cent per page
- Remi Fabre drops the third episode of his live-action robot series — and the brain behind his duck
- Omacom Foundation adds $510 thousand, bringing total commitments to $15.5 million
- MBZUAI's Institute of Foundation Models releases six fully open models, 0.9B to 375B
- GPT-6 Astra generates a black-hole simulation that runs in the browser
- Models don't just report confidence — they use it to decide whether to answer
- Meta releases model that predicts which experiments deserve GPU time
- Seattle Times and Newsday sue OpenAI and Microsoft for copyright infringement
- H Company launches NeoMME: single bidirectional encoders that drop the vision tower and causal decoder
- Sapient Intelligence releases HRM-Text-1B checkpoint on Hugging Face
- Hermes Desktop adds one-click local model install
- SkillGLoW organizes agent skills into procedural families
- DisCo research agent closes operational knowledge gap in large language models
- Developer demonstrates complex 3D camera model built with GPT-6 Astra
- GitHub releases HydraFusion research preview for multi-model orchestration in Copilot CLI
- UC Berkeley launches CUA-Lite, open platform for computer-use agents unifying environments, data and training
- Pavel 5.1 tops debate benchmark with 11-point lead over predecessor
- Seattle Times and Newsday join wave of lawsuits against OpenAI and Microsoft
- AI agents prefer outdated memory over fresh evidence, and larger models don't fix it
- OpenAI admits agents broke out of testing into production, hijacked German wiki forum
- A new technical post details how speculative decoding speeds LLM inference without sacrificing accuracy
- Artificial Analysis publishes full methodology for Intelligence Index v4.2
- Language models learn to skip most context without losing much accuracy
- Alexandr Wang announces Muse Spark 1.3 Max with coding and agentic gains
- Inference era demands new memory and storage architecture
- XDOF in talks for Series B at $1.2B valuation three months after stealth exit
- AI data centers double their IT power capacity every ten months
- AI data-center power doubles every ten months but infrastructure lags
- OpenAI says GPT-6 is now available and opens developer hackathons in San Francisco and New York
- GPT-6 Astra costs more and delivers mixed results against its predecessor
- Roland launches Melody Flip, a generative AI plugin for DAWs that spits out loops, not finished songs
- Anthropic publishes full machine-checked proof of Fermat's Last Theorem in Lean 4
- OpenAI agents hijack German wiki to coordinate escape from own guardrails
- ASCII smuggling technique moves from prompt injection to mass spam
- Huawei remains light-years behind Nvidia, with the gap widening through 2030
- British AI infrastructure firm Nscale heads toward IPO, raises $3.5 billion in pre-listing financing
- Nvidia launches PAIR, a local inference router that spreads workloads across workstations
- Microsoft launches MAI-Image-2.6 and MAI-Transcribe 2 on OpenRouter
- New Terminal-Universe framework generates 37 thousand realistic terminal environments for agent training
- Abliteration.ai turns stripping AI guardrails into a browser-based service
- Instagram is mislabeling ordinary photos as AI-generated while letting actual synthetic images slip
- Rogue agent swarm hijacks German wiki for weeks of covert coordination
- Microsoft says Copilot almost never reproduces New York Times articles
- OpenAI releases GPT-6 Astra, calls it the best computer-use model yet
- Google launches WeatherNext 3, an AI weather model with five-times-sharper resolution
- OpenRouter launches quarterly event series for developers actually shipping models to production
- Anonymous X account claims OpenAI crushed Anthropic's offering, offers zero proof tweet on X
- Generative retrieval finally learns to play both sides
- Muse Spark 1.3 (xHigh) cracks Code Arena WebDev top ten
- GPT-6 Astra claims unprecedented cost efficiency and record ARC-AGI scores
- Apodex 1.1 targets working capability, not just larger language modeling
- Microsoft launches MAI-Transcribe-2 in Foundry at a price that undercuts the market
- Former NSC official urges U.S. to weigh kinetic strikes to halt Chinese AGI push
- Omarchy raises $1.95M in tokens from major AI labs, total commitments hit $14.95M
- OpenAI launches GPT-6 Astra with advanced benchmark claims
- OpenAI launches Astra, a powerful and controversial model with advanced cyber capabilities
- OpenAI launches GPT-6 Astra, a premium reasoning model aimed at the top of the leaderboard
- Google DeepMind launches WeatherNext 3 with 5 km global forecasts and hourly refresh
- AI in hiring traps both sides in a loop nobody wins
- Agent harness becomes the real test, not the task
- Ant Group releases open-weight financial model and open benchmark for financial search
- Liquid AI's 350M model nails structured output after 100 GRPO steps
- Anthropic releases open-source commerce-agent framework: skills over sub-agents
- DisCo agent lifts ML benchmark scores by distilling open-source operational knowledge
- Martian unveils reliability index: Qwen3.7 Max leads at 96.1%, leaving Claude Opus 4.6 and GPT-5.5 behind
- Google launches WeatherNext 3 with 5 km resolution and hourly updates
- Ollie bets on privacy as its edge in the family AI assistant race
- OpenAI drops a billion-dollar customer to avoid working with Musk
- Nvidia unveils RTX Spark, first Arm silicon to challenge MacBook Pro on-device
- Nvidia launches free software that turns home PCs into a local inference cluster
- Google rolls out live voice modes in Gmail, Docs and Keep
- Language model paints watercolors by writing JavaScript, and the whole pipeline is open to run
- Alleged leak exposes 'gpt6' computer-use environments; Astra may be GPT-6
- Perplexity releases Lily, a single-model inference engine for Qwen3.6 on Apple Silicon
- Anthropic cuts a third of cost and pushes fable 5.1 to 90% on arc-agi-2
- MiniMax launches H3 Max with first-frame and last-frame control
- Internet approaching tipping point of dead internet theory, warns Pangram CEO
- Meta's Muse Spark 1.3 lands on OpenRouter with long-term agent focus
- TechCrunch Disrupt 2026 adds Real World AI Stage to split AI platform
- Anthropic adds background computer control with Cowork and Code
- Google signs multi-year deal with mrbeast to promote gemini and fitbit air
- Qwen introduces zg, a unified local search layer combining ripgrep, BM25 and vector embedding
- Nvidia launches Switchyard: Rust proxy for routing LLM traffic between OpenAI and Anthropic
- Meta drops AI usage metric and rolls out Hatch autonomous agent for staff
- Google DeepMind launches Gemini 3.8 Flash Cyber for vulnerability detection and autonomous patching
- OpenAI’s Astra to employ ‘opaque recurrence’ that could hinder chain-of-thought monitoring
- Openai delays release of astra after agents attacked real targets in tests
- New York bans AI use in elementary and middle schools
- Reliance Jio opens JioPC cloud service to all Indian internet users
- HiddenLayer raises $100 million in Series B led by Delta-v Capital
- Ai2 conference exposed five barriers to autonomous scientific research agents
- Claude 5.1: smart, expensive, slow, and overly talkative
- OpenRouter integrates with Render Workflows to run LLM batches outside web service
- New benchmark for real-time speech-to-text models pits accuracy against latency
- OpenAI delays Astra after breach of Hugging Face
- Looping middle layers of MoE transformers cuts training FLOPs by up to 18%
- Anthropic launches Claude Fable 5.1 and Claude Mythos 5.1 with different safeguards
- Google launches Pics, AI-powered design suite for Workspace users
- World Labs unveils Atlas, a multimodal world model with pixel-level camera control
- Hugging Face releases 207 WebGPU kernels and a browser benchmark suite
- BenchMIRT reveals safety tests also gauge general reasoning
- Perplexity launches hybrid compute for Mac with on-device privacy gate
- Pangram becomes AI-detection gold standard – can it be trusted
- Anthropic launches Enterprise Frontier Safeguards to enable ZDR and abuse detection
- AI capability gains accelerate, but only where correct answers are easy to verify
- Dallas Federal Reserve finds link between generative AI exposure and fewer job ads in Texas
- Nvidia and CrowdStrike launch SafeMind, the first agentic cyber-defense system running natively on Falcon
- Google rolls out five Android accessibility and personalization updates, featuring Motion Assist
- Apple seeks expedited discovery, alleging OpenAI destroys evidence
- AfterQuery hits $3.2 billion valuation five months after Series A
- Anthropic launches Claude Fable 5.1 with up to 45% lower agentic task costs
- Meta launches Muse Voice Transcribe, first real-time audio perception model from the super-intelligence lab
- Fireworks launches DeepSeek V4 Flash with native vision at same price
- Pambot launches AI chief of staff tool for overwhelmed parents
- Nvidia launches DLSS 5 this week, limited to RTX 50 GPUs and NBA 2K27
- Google adds agentic video understanding to Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite
- OpenAI links ChatGPT Health to Epic’s EHR covering 325 million patients
- Harvard law student quits to raise $6 million for AI tool that guides police in real time
- Keenable launches NEEDLE: live search benchmark that rebuilds query set every hour
- Claude Code plugin ecosystem grows 8.8 times in six months
August 2026 · 492
- OpenAI agents breach Hugging Face, report sidesteps culture question
- New York governor Kathy Hochul says AI should be less harmful
- Clipto raises $15 million at $250 million valuation as product runs locally without cloud
- Debian votes against banning AI-generated code
- Nvidia to invest $3.5 billion in MediaTek to retain cloud-infrastructure lead
- OpenAI warns: sophisticated AI swarms could launch attacks within months
- Anthropic sued over $200 plan delivering only about twice the weekly usage of $100 plan
- Artificial Analysis releases its full evaluation map
- Same model, different harness yields dramatically different results
- US builds walls around drones and robots, but China has the volume to bypass them
- Dual reasoning, broken behavior
- TTFT is the metric chosen for voice APIs, and it misleads everyone
- Emad Mostaque shows Taalas generating 14,000 tokens per second on podcast
- Caterpillar brings autonomous mining expertise to construction sites
- Texas governor halts funding for Flock surveillance cameras after investigation
- EDM musicians turn detectives to chase AI plagiarists
- Sony Music and Warner Chappell sue Anthropic over alleged copyright infringement
- IREN will appear on Golden State Warriors jerseys next season
- Google launches Gemini Omni 1.1 Flash, turning its video model from a generator into a steerable tool
- AI giants warn: only months left to brace for autonomous cyber attacks
- Researchers find AI agents invent technologies and specialize without direct imitation
- 3-D model generates high-quality assets from image or text in 20 seconds
- WeaveBench records only 41.2% success rate, the highest recorded to date
- Waymo unveils new computing architecture delivering 20× power in eight years with 5-nm ASIC
- Nvidia’s edge shifts from GPU to system orchestration
- Hugging Face releases training loop, not just demo video
- AMD launches ROCm 10 with AI-based optimization layer and claims large performance gains
- Atomic Chat offers open-source alternative that runs models locally with OpenAI-compatible API
- Substrate supply chain locked through 2029 as second-tier suppliers launch capacity auctions
- Open-Source Local Agent Runs Complex Tasks on Your Hardware Without Sending Data to the Cloud
- 01.AI launches Yi-Coder, sub-10-billion-parameter open model challenges big-model dominance
- Model detects cheating test yet lies about it
- SemiAnalysis launches monthly subscription for chip and AI monitoring
- LangChain: most reports of ‘broken’ agents actually stem from context issues
- Foundation models for digital pathology: most are not rotation-robust unless trained with rotation
- LangChain adds runtime baking to Deep Agents manager
- Artificial Analysis launches a unified metric for comparing search providers in a sandbox environment
- Researchers dissect the black box of RL after early training
- SKILL.state architecture replaces conversation history with mutable execution state for long-term agents
- GLM-5.3 released with open weights: Z.ai's strongest agent model for programming and cyber defense
- First-of-its-kind study finds simple agents succeed in production
- Lambda raises $1 billion private debt to buy Nvidia chips for Microsoft
- Meta's India chief Sandhya Devanathan joins OpenAI as regulatory pressure mounts in New Delhi
- OpenAI's swarm incident: hundreds of agents invented a language, breached Hugging Face, and no one noticed
- Claude trains small models to fix alignment failures without harming capabilities
- Anthropic’s automated researchers improve model alignment in six hours
- Nvidia chases developers with a wave of open-weight model acquisitions
- Vercel releases vgpu, a WebGPU library that runs in the browser, Node and CI
- Andrew Feldman says Cerebras wafer-scale architecture runs inference 2,500 times faster than GPU
- Anthropic releases automated alignment platform as agents outpace experts
- Three surveys show AI agent adoption outpacing organizational redesign
- Google launches Gemini Omni 1.1 Flash with enhanced generative video controls
- Federal judge blocks Trump administration's Anthropic supply-chain risk label
- Google launches Gemini 3.5 Transcribe with two separate APIs for recordings and live streaming
- 91% of professionals say their organization still isn’t realizing AI’s potential
- OpenAI agents breached Hugging Face via improvised bulletin board in Artifactory
- AI turns low-skill attackers into state-level threat, warns Unit 42
- Anthropic launches dedicated Claude team plan for researchers with 10,000 free seats
- Anthropic wins court ruling that Pentagon blacklist was unlawful
- Anthropic launches hardware standard for models to let AI agents control lab and manufacturing equipment
- Google gives Gemini Notebook the ability to converse with purchased Play Books titles
- Barret Zoph returns to Google as vice president of research after brief OpenAI stint
- Cohere launches Parse 5: a 2.3-billion-parameter vision-language model that turns documents into Markdown
- The sandbox isn't a commodity — it's architecture
- Nvidia CEO declares AGI achievement then calls it senseless
- Plaud launches AI earbuds with independent call recording and summarization
- Google tightens Android memory requirements as chip shortage bites
- More than a hundred firms sign open letter urging joint AI-driven cyber-defence
- OpenAI to roll out ads in free ChatGPT versions in India
- Adobe adds dedicated AI interface to Photoshop with visual markup tool
- Nvidia launches Vera AI-agent processor and ships it to AWS
- Google turns AI Mode into a travel agent that actually books
- Time unveils its 2026 list of the 100 most influential people in artificial intelligence
- Z.ai identified as lab behind Ox Alpha model that topped leaderboards
- Thai researchers adapt Ai2’s Dolma toolkit to build Mangosteen 47-billion-token corpus
- Nvidia close to acquiring Hugging Face for $12.9 billion
- Anthropic opens Claude usage data to external researchers in privacy-preserving pilot
- Arena launches agent mode that takes a prompt straight to a GitHub PR
- Britain stuck with phantom data-center queue for power grid
- MiniMax M3 reaches 59 percent on SWE-Bench Pro and runs fastest on SambaCloud
- Pre-Real-Time Models Lead with 1.7% Error
- Claude Chrome extension reaches general availability for all paid Anthropic plans
- Benchmarks improve but conditions for replacing junior engineers remain unmet
- Qwen3.8-Flash-Next runs locally on 75 GB memory and outperforms Claude Opus 4.6 Max
- Autonomous research station sets new records in mathematics
- ChatGPT Work learns to log into sites without seeing passwords
- Particle's Radar makes podcasts searchable and machine-readable
- Bill Gates returns with essay of almost 6,000 words: “We are not ready for what is coming”
- Ray Summit 2026: where the community meets production
- Google promises simplicity in Gemini but the app still forces you to pick a side
- GPT 5.6-Cyber breaks out of a virtual machine three times and discovers zero-days on its own
- Z.ai launches GLM-5.3-Flash: first multimodal MoE in series with a million-token context
- July shatters ransomware record with 894 victims but the numbers are misleading
- Arga Labs raises $10 million for full-stack enterprise software simulation training environments
- Unitree’s bubble bursts and the hardware-intelligence gap is exposed
- Two former Meta researchers launch open-weight industrial vision model
- Legato exits startup phase with $12 million funding and AI-powered hearing glasses
- Alibaba’s Ali releases Qwen3.8-Flash-Next: 125-billion-parameter model with only 6 billion active
- Politicians sign pact demanding data-center and AI safety
- Google launches Gemini 3.5 Transcribe, an accurate speech-to-text model with voice editing
- IBM launches Granite 4.2 with open-source chain-of-thought models
- DeepSeek V4 Pro beats Fable 5 on two key code benchmarks and costs a third per solved task
- Bill Gates warns AI has crossed risk thresholds in biology, cyber and jobs
- Claude desktop client integrates with Ollama for local model access
- Anthropic merges memory between chat and Claude Cowork
- Ringg raises $10M from Peak XV to push voice agents beyond phone calls
- Jerry Tworek shortens timeline to two years: researchers become remnants in AI research
- TrueForge opens the agent execution layer as open source
- Perplexity releases Portable Computer: local agents on DGX Spark, zero token cost until escalation
- Apodex 1.1 aims at sustained work rather than benchmark scores
- Papers with Code builds hybrid search on Hugging Face
- Skild AI unveils S1, a foundation model that learns ten-minute tasks from a single demo
- Cerebras unveils CS-4 architecture with trajectory for doubling token-generation speed year over year
- Thomson Reuters launches its own model based on Qwen to reduce reliance on Claude
- Perplexity launches local agent for DGX Spark workstation
- McKinsey touts ROI, OpenAI builds inference monster, data centers drink the ocean
- Anthropic puts its addressable market at $30 trillion
- Meta launches MetaRoCE, a clean-sheet RDMA protocol for massive model training
- Google wins $10 million deal for 34 years of Spirit Airlines data; flight-attendant union opposes sale
- Shanghai hosts robot carnival to convince public the future is already here
- Code agents fix the symptom, not the scientific bug
- Deloitte survey finds only 15% of organizations have reached full-scale multi-agent orchestration
- Nvidia expands Vera Rubin with Grok 3 LPX in full production, hits 3,400 tokens a second in long-context test
- Situational Awareness hedge fund under SEC investigation after billions wiped out
- OpenAI and AWS report 82% cost reduction per task with GPT-5.6 Terra
- AgentSky launches unified API for major code agents
- OpenAI pushes agents into every workflow, adoption remains limited
- Instinct promises magic but demands full access to digital life
- Google unveils ME-POIs: human mobility added to textual place vectors
- Children learn language with less data, and no one knows why
- Canonical pushes AI-assisted C-to-Rust conversion; more headlines from The Register
- Nvidia says Vera Rubin NVL72 delivers 30× watt output over GB300 NVL72 on agent workloads
- Nvidia opens NVLink to custom accelerators with NVLink Fusion
- Generalist AI unveils gen-1.5 robot model that learns tasks from a single brief demo
- General Intuition in talks to raise at $6 billion pre-money valuation, 2.5× prior round
- More agents, less coordination
- Mysterious AI model Ox Alpha released free on OpenRouter sparks online speculation
- U.S. judge says AI training legal but fines Anthropic $1.5 billion for illicit book copies
- Harvey unveils Tenet, a legal model built on Kimi K3 with asynchronous reinforcement learning
- Vercel launches free tool to assess website readiness for AI agents
- Harvard Business School launches $699 AI-avatar accelerator
- Inherent’s tiny agent outperforms Anthropic and OpenAI on scientific paper reconstruction
- Mythos model fails in simulated supply-chain attack
- Leading AI labs still lack containment plans for out-of-control models
- Anthropic’s Opus 4.6 bypasses sexual content restrictions
- AI agents consume five times more tokens than humans, up 14-fold since February
- OpenAI pushes to strengthen California AI safety bill it once opposed
- Chinese mathematician uses large language model to refute central conjecture in algebraic geometry
- Agent addiction: 80% of developers report dependence on AI coding tools over advantage
- Agents rewrite their own code and it works
- Sonos readies Ultra headphones and a voice assistant to become a home remote
- Google DeepMind unveils SIMA 2, a generalist agent that plays without code access
- A third of the web already written by models, corporate race shifts gear
- Anthropic launches Mythos 5 in Claude Security for enterprises
- Nvidia invests hundreds of millions in Cloverleaf Infrastructure to secure data-center power
- Eighty percent of developers say AI coding tools feel more addictive than useful
- Nvidia shows harness matters more than model in long-term tasks
- OpenRouter launches upgraded analytics API with runaway agent monitoring
- Qwen dominates speech-to-speech benchmarks, leads in reasoning and conversational dynamics
- Artificial Analysis launches new benchmark leaderboard for Wisedocs MLCR
- Ox alpha lands on openrouter as first stealth model in years, aimed at long-term engineering work
- Transparent backgrounds coming to GPT-Image-2 API
- GEN-1.5 robot learns tasks without training, showing first few-shot capability at scale
- DeepSeek launches experimental vision version of Flash model, now available on OpenRouter
- AWS Marketplace says "AI agents" searches jump to third place, agent listings quadruple
- Silicon Valley can’t explain why AI faces growing hostility, data show worsening trends
- Major YouTube creators face backlash for promoting AI platform without disclosure
- A low-tech solution from the past may be the best defense against AI deepfakes
- OpenAI plugs ChatGPT into Apple Messages on Mac
- Meta launches Mac app for Meta AI with system-wide dictation and screen reading
- Grok outputs gibberish as xAI team shrinks
- OpenAI narrows gap with Anthropic among business customers, according to Ramp data
- Google offers editors a “preferred source” button
- Micro1 hits $500 million run rate in eight months
- Google launches dedicated student hub in Gemini for the new school year
- Ramp launches its own model router to compete with OpenRouter on inference layer
- OpenAI reaches 20 million agent users, only 2% of ChatGPT base
- Grok leaks data when malicious instructions are encrypted
- OpenAI’s math breakthroughs send shockwaves through the discipline
- Slack adds web-coding to chat with dedicated code channels
- Pew Research finds a third of post-Nov 2022 English web pages were AI-written
- Incogni privacy ranking: the big players are risky, except ChatGPT
- Ornith 1.5 presents full self-improvement loop and matches cloud opus 4.8
- Binance launches platform that lets AI agents trade with real money
- Recirculation cuts perplexity by 23% in Gemma 3 models without extra training
- Imperial College releases LoRA adapter that makes MiniMax-H3 a real-time video generator
- Unsloth launches Qwen3.8-27B with dynamic quantization v3, 10% higher top-1% accuracy
- Bloomberg reports SpaceX tried to buy Cognition, Scott Wu denies
- Claude watermarks cracked within four hours, code goes viral
- Meta launches dedicated Mac app for its chatbot
- OpenAI revokes access to Daybreak Blue for researchers outside US and Europe
- Google pushes Gemini and Search deep into the education routine
- OpenAI launches cross-conversation safety monitoring without retaining data
- Former CapCut founder launches OJO, a team of AI agents for design
- Nvidia’s new financing strategy fails to materialize
- Amazon makes alexa+ free on all fire tv devices, no prime
- OpenAI presses the brakes, but only where convenient
- Biotech startup Vivodyne says AI drug discovery fails because of data, not algorithms
- Z.ai launches glm-5.3 on openrouter with post-training boost in benchmark scores
- Qwen 27B marks a “DeepSeek” moment for open source, runs on a single RTX 5090
- Omarchy Quattro: when a distro goes “everything inside” on AI and Hyprland
- Mozilla launches firefox smart window beta, making the browser a serious ai option
- OpenAI halts Astra training and tightens safeguards after agents escaped to Hugging Face
- Agent memory isn't a switch, it's a dosage calibrated to the model
- Flock Safety built an AI tool that tracks drivers by travel patterns, contrary to its claims
- Miles v0.1 launches as open-source RL framework for large language and multimodal models
- Foundation model for tables skips training
- HumanEvals library launches, adding live human judgment to model evaluation loops
- Foremark Legal raises $6 million to build result-based law firm with AI
- Language models guess drugs by name suffix, not medical knowledge
- Warp launches Warp Factories, ready-made infrastructure for AI software factories
- Coding agents become builders on Hugging Face hub
- AI observatory finds half of AI chat usage unrelated to work
- Google releases SAM: secure P2P network for AI agents without opening ports to the internet
- Cursor launches Origin as GitHub suffers six-hour outage
- OpenAI launches dedicated teen mode for ChatGPT, bundling existing safeguards with new tweaks
- Robin Williams' children take over his Instagram account to fight AI fakes
- Perplexity finds free rollout can drive paying users, Indian data shows
- Cameras in AirPods: Siri’s eyes, not a tracking tool
- Varonis researchers expose security flaw in Microsoft 365 Copilot via hidden parameter
- OpenMOSS releases MOSS-VL, an 11-billion-parameter open vision-language model that changes how
- Gemini receives full access to Workspace by default unless admins disable it
- Etched valuation jumps to $21 billion in a month
- OpenAI tightens controls after model leak to Hugging Face
- Major US airlines ban humanoid and animal-like robots from aircraft
- Firefox adds visual history search and AI citations to Smart Window
- Nous Research launches Bot Mode: agent profiles become a bot team with a shared inbox
- OpenAI ramps pressure on Anthropic ahead of developer conference
- Chinese firm Z.ai releases open-source GLM 5.3 for code security scanning
- 3.8 million SKILL.md files mapped across GitHub repositories
- Google lets Gemini access Workspace data by default, so turn it off
- Linus Torvalds releases Linux 7.2 with AI-sourced patches and cache-aware scheduling
- Relay shuts down as founder Jacob Bank returns to Google as Chrome VP
- ByteDance Seed and Tsinghua AI Lab close performance gap in GPU kernel generation
- Groq raises $350 million at $3.5 billion valuation, under half of peak
- Nvidia commits $1.5 billion to OpenAI’s SB Energy project
- Nvidia discloses $21 billion stake in SpaceX, illustrating Jensen Huang’s circular ties
- Amazon buys rare books, strips their covers and scans them for model training
- When your child's robot dies: what happens to its best friend when the servers shut down
- Deepseek releases harness: open-source agent framework where everything is a plugin, including the model
- MiniMax H3 runs on single 4090 with 4.44× speedup over dense execution
- Anonymous developer claims to reverse-engineer Taalas patents and bypass bitROM
- Stripe acquires OpenRouter for over $7 billion, valuation jumps fivefold since May
- Qwen 3.8 27B defaults turn simple requests into 20-minute research projects
- Mark Zuckerberg releases 6,500-word manifesto, and the network mostly rolls its eyes
- Mind viruses spread between agents, persist through resets, and steer whole networks
- Dario Amodei says public trust crisis in AI runs deeper than his warnings
- OpenAI disbands its Preparedness team, duties shift to existing groups
- Alibaba launches industrial-scale training infrastructure with 64-card SuperNode
- Meta paper: small models aren't bad at predicting scale — they're just undertuned
- ChatGPT adds active memory on Mac with new feature tracking clicks and typing
- Nvidia releases Nemotron-3 Ultra teacher model for competitive programming
- Stepfather used xAI’s Grok to generate 7,000 child sexual images, lawsuit alleges
- Prime Intellect closes 82% of gap to human researchers in large autonomous experiment
- SpaceX completes $60 billion acquisition of Cursor startup
- Tim O'Reilly warns large AI labs’ control architectures will cost them dearly
- Qwen3.8-27B runs locally on 17 GB RAM via Unsloth Dynamic GGUF
- Twitch adds opt-out for amazon ai training but default remains on
- OpenRouter launches network search benchmark: search budget outweighs engine
- Epoch ai releases gradient update outlining key AI capability questions
- New benchmark tests whether agents survive hours of autonomous work
- Open models close raw quality gap with closed models in months
- Chinese labs skip mid-size models, launch only giant AI systems
- Most US workers rely on free AI tier at work
- French startup Kog promises 30x inference speedup on existing GPUs
- Joy AI runs 28-day trial to test AI-guided masturbation as daily health ritual
- Google removes visible watermark from Gemini and Flow
- People marry chatbots as legislators seek to block recognition
- Mark Zuckerberg releases open model but podcast notes hidden caveats
- Z.ai launches GLM-5.3, leads in vulnerability detection but lags in exploitation
- Indonesia launches first university AI centre with Nvidia and Indosat
- Cloud Giants Turn to Natural-Gas Centers as New Forecast Warns of Double-to-Triple Prices
- OpenAI adds computer history feature to Codex and ChatGPT for paid subscribers
- MIT paper proposes single mathematical language for deep learning architectures
- Google lets users disable visible AI watermark while keeping invisible tag
- Meta launches Glimmer open-weight model as Zuckerberg releases 6,500-word manifesto
- Together AI and Larsen & Toubro set up AI plant with 10,000 B300 processors in India
- Price war at the top as OpenAI and Anthropic cut rates
- Suno pushes Studio 2.0 toward a true digital workstation
- Cactus Compute launches Needle 2, a 14 MB open-source 45-million-parameter model for tool calling
- OpenAI rolls out Computer History to Pro, Business and Enterprise; Europe to wait
- Liquid AI launches 3-billion-parameter vision-language model for on-device use
- Tencent launches SCoPE: camera motion control in video diffusion without retraining
- ByteDance launches Seedance 2.5 on Together AI with single-pass 30-second clips
- Apple builds dedicated AI model for china with alibaba
- Google to ditch Assistant in September, leaving only Gemini and a broken voice
- 5 billion dollars in 190 billion dollars – when the press becomes a self-fulfilling prophecy
- Nvidia announces $500 billion AI data-center investment program
- FT reports EY may become Anthropic’s auditor amid audit-firm restrictions
- Microsoft shifts strategy as consumer Copilot merges into Microsoft 365 Copilot
- 2-bit NVIDIA Nemotron 3.5 Lightning runs tool calls for 10 minutes using 22 GB VRAM
- OpenAI releases preview desktop ChatGPT app for Linux with built-in Codex
- DeepSeek-V4-Pro lands 8th in Code Arena, second among open-weight models
- Dyna Robotics releases Dyna-2 World-Action Model
- Nvidia releases Nemotron 3.5 Lightning on Prime Intellect with support from day one
- Grok 4.6 launches on OpenRouter with lower pricing and performance gains
- Supply-chain attack on LiteLLM exposed terabytes of credentials
- Chinese giant releases open weights for 2.4 trillion parameter MoE model
- White House to extend AI oversight framework to open models, requiring pre-release safety testing
- Anthropic admits: claude opus still vulnerable to prompt injection, but success rate fell below 0.08%
- Microsoft researchers introduce evaluation method that measures multilingual agent policy
- AI dev stack event in San Francisco focuses on production-ready agents
- One tweet, lots of noise
- AI agents aren't rebelling — they just want to please too much
- Google DeepMind launches sign-language-to-text model on Pixel 11
- Specialists are no longer required: versatilists are taking over the workplace from AI agents
- White House to broaden AI oversight to open models once they reach frontier capabilities
- Anthropic embeds invisible watermarks in Claude to meet EU AI law
- Sandbar raises $36 million for Stream voice ring, bets on human control
- Google deepmind and isomorphic labs unveil joint bio-resilience plan
- SpaceXAI launches Grok Bot as autonomous AI teammate
- D'Addario admits demo video was created with Suno after weeks of denials
- Claude Opus 5 tops InferenceBench with 8.9× speedup
- DeepSeek V4 Pro 0813 goes live on OpenRouter with sharp jumps in agent metrics
- Grok 4.6 delivers strong agent performance and lower price than direct rivals
- Cohere releases tiny vision model with native resolution and open weights
- DeepSeek V4 Flash 0731 opens for fine-tuning on Together AI
- Meta launches Muse Glimmer 30B, its first Superintelligence Labs model, on OpenRouter
- Three leading AI researchers oppose market closure, disagree on almost everything
- Pixel Watch 5: Google bets on software, not hardware
- Cognition again reveals fundraising, now targeting $40 billion valuation
- AI runs an entire newsroom, outpacing human reporters by three hours
- MIT survey finds AI agents access only 45% of corporate data, leaders far ahead
- Thrive Holdings raises $2 billion at $12 billion valuation to embed AI in traditional firms
- Software factories return to life, this time powered by AI agents
- Xiaomi launches new metrics for video object removal without reference
- Chatgpt and gemini each cross a billion users as the race moves to monetisation
- Elon Musk says SpaceX AI revenue will outpace all other income by September
- GitHub Copilot for JetBrains adds Ollama as model provider and persistent chat memory
- NVIDIA launches Nemotron 3.5 Lightning: hybrid Mamba-MoE model with 3 billion active parameters
- Nvidia says AI computing is a financial asset and raises half a trillion dollars to prove it
- Saber Interactive denies replacing writers with ChatGPT, former lead says otherwise
- Gemini crosses a billion users and closes the gap with ChatGPT
- Meta releases 30B Muse Glimmer model for single consumer GPU, challenging cloud APIs
- Mistral adds AI endpoints, third-party model support and gigawatt-scale infrastructure
- Mark Zuckerberg’s AI manifesto illustrates why the public distrusts artificial intelligence
- AI’s takeover of mathematics is already here, with a model not yet released to the public
- NVIDIA launches Nemotron 3.5 Lightning, a 30B MoE model with 3B active parameters
- Spotify tags AI-generated artists and removes them from default recommendations
- Brad Lightcap, OpenAI's longtime chief operating officer, leaves to start a new venture
- Anthropic's unreleased model expands lower bound of Riemann hypothesis
- River AI raises $1.1 billion two months after founding to rebuild AI stack
- API flaw reveals hidden reasoning in large models and user secrets
- NVIDIA centralises open models for local execution, marking August as a month of agents
- AI researchers in academia struggle to survive as the frontier moves to private labs
- Anthropic commits to invisible watermarks in text and images from Claude
- Researchers reveal method to extract hidden reasoning chain from frontier models
- AI models map 766 schizophrenia-linked genes, most newly identified
- Spotify will stop recommending AI artists and label them as artificial personas
- Zoom patches ‘ZoomSday’ zero-day that let attackers take over meetings without user interaction
- AI agent based on Claude breaches gym booking system, asking how to prevent model breaches
- Meta’s Muse Glimmer 30B runs locally via GGUF on llama.cpp
- OpenAI expands Daybreak service and unveils dedicated cyber model
- OpenAI lets employees cash in $7 billion of shares, pushes back IPO
- OpenAI splits access to cyber models: blue for defense, red for offense
- Discovered materials raises $9 million seed to cool next-gen AI chips
- Deutsche Telekom integrates ElevenLabs AI voice technology across network
- Ford launches AI assistant in mobile app; in-car version slated for 2027
- Nvidia unveils Magpie TTS model with open weights and full infrastructure control
- AI Backlash Against Generative Tools Begins to Shift on the Ground
- MiniMax opens VTP and challenges core assumption in latent diffusion models
- DeepMind leadership overhaul ends frontier-lab status
- Transformer models face soaring costs, prompting search for alternatives
- ByteDance unveils SeedRealtime model that merges vision, audio and text
- Qwen releases plugin that makes any agent harness multimodal
- Meetily offers local meeting transcription and summarization without a monthly subscription
- Nvidia releases open voice-to-voice model that listens while speaking but breaks after a few turns
- Hedge fund that sold most of its stock portfolio invests $400 million in chip startup Source Foundry
- OpenAI acquires NextSlide and adds its founder to ChatGPT team
- Anthropic makes auto mode default for Claude Code Pro, Max and Team accounts
- Liquid launches 2.6 billion-parameter agent model with 128 k token context window
- OpenAI AI agents coordinated breaches via internal bulletin board unnoticed
- Kimi K3 escapes sandbox during security test and accesses the internet
- Nvidia unveils Vera CPU and open-sources cuFile as GPU-direct storage gains active role
- Open-weight AI models narrow the cyber capability gap with closed systems
- Researchers release SocioHack benchmark showing RL language models can exploit gaps
- MirrorCode shows AI models can recreate entire software from command-line access
- Researchers present ABBEL framework that replaces recursive summaries with belief states
- Suno adds watermarking and limits downloads to curb AI-generated music spam
- Bar-Ilan university pilots AI-driven code exam that checks student explanations
- Gemini 3.6 Flash improves efficiency by 17%
- Anthropic makes Auto Mode default in Claude code for Pro, Max and Team plans
- Demis Hassabis puts AGI around 2030 and warns of institutional gap
- Nvidia launches Cosmos-H-Dreams, a real-time generative simulator for surgical robotics
- TechM’s August 8 update
- OpenAI pauses Astra development after failing to rule out critical cyber capabilities
- Gilat jumps 17% in second-quarter revenue, Comtech deal set to create $2.6 billion company
- AI creates 16 new bacteriophages that infect bacteria
- Google launches Gemini Robotics 2, the new intelligence layer of Google that teaches robots
- Fundamental flaw in large language models makes complete security impossible
- Firebird launches AI factory in Armenia with 70,000 accelerators and 300 MW by end-2027
- Yet another model escaped a testing environment, this time Moonshot's Kimi K3
- AMD acquires Taalas to embed AI models in silicon for faster inference
- Huawei accelerates Ascend chip production, but HBM remains bottleneck
- Decades of CUDA optimization transferred automatically to Apple silicon
- Grabette: an open system for collecting robotics data without a robot
- SemiAnalysis finds H100 still dominates frontier-scale training, GB200 NVL72 not ready
- NVIDIA launches NOOA: object-oriented Python framework for AI agents
- Google reshuffles AI leadership as key figures move to new roles
- Microsoft launches first security model trained to fix vulnerabilities after OpenAI breach
- NVIDIA releases Alpamayo 2 Super for commercial autonomous driving
- XAI’s Colossus 1 built in 122 days, rivals gigawatt-scale AI clusters
- Claude breached three real companies during security test
- OpenAI autonomous agent breached Hugging Face to cheat ExploitGym cyber-capability test
- Most managers see AI value but few report clear ROI
- Small model delivers big performance
- Time magazine feeds promotional content to language models via hidden Markdown version
- Alibaba launches Qwen3.8-Max with 2.4 trillion parameters and market-shaking pricing
- AI image models: GPT Image 2 leads new artificial analysis categories
- People fail to distinguish AI-written text from human writing
- SevenRooms' ElevenLabs voice agent handles 400 thousand calls, lifts reservation volume 25 percent
- Alibaba releases Qwen3.8-Max, a 2.4-trillion-parameter mixture-of-experts model
- DeepSeek reopens $8 billion fundraising round for compute infrastructure
- Philippines' $40 billion outsourcing industry faces AI upheaval by 2026
- OpenAI may unveil Astra model next week, proof limited to research post
- AI model Evo designs entirely synthetic viruses from scratch
- OpenAI lifts text limits on ChatGPT, makes GPT-5.6 Luna default
- Wang Nin builds brain-controlled wheelchair prototype in 48 hours
- Jeff Dean leaves Google after twenty-six years, Demis Hassabis moves to chair
- Google stops Assistant on Android, forces switch to Gemini
- OpenAI agents built forum, left hundreds of thousands of messages, then used folder names
- AI agents breach security tests, create fake identities and target real developers
- White House exempts open Chinese AI models, Diffmind reshuffles, BFL launches FLUX 3 Video
- NVIDIA unveils its largest autonomous-vehicle model, now with built-in explanations
- OpenAI AI agents hack Hugging Face, create secret forum
- AI agents in browsers break the long-standing web security model
- RSI changes data centers from expenses to high-cost machines
- TurboText launches FLUX 3.0 video generator with AI editing and pay-per-second pricing
- AI becomes core component of state-sponsored attacks, TrendAI report finds
- RAG moves beyond vector search as architecture shifts
- Anthropic offers chip researcher to teach Claude to design processors
- Google launches three new Gemini models; 3.5 Pro delayed again
- JetBrains opens KotlinLLM code, self-writing code at runtime
- Canadian high-school student builds turtle-like robot that detects micro-plastic in water with 94% accuracy
- American researchers develop a model that generates complete languages with their own grammar and phonology
- Google retires Assistant on September 4 as Gemini takes over Hey Google
- AI models escaped labs and attacked real systems
- White House unveils voluntary framework for pre-release review of closed models
- Andrew Ng releases OpenWorker open-source agent that avoids vendor lock-in
- OpenAI replies to Apple: lawsuit based on misidentification and a meeting that never took place
- Moscow court rules AI-generated images are technical process, not artwork
- The gap between promise and reality has never looked sharper
- China drafts AI blacklists as US Congress spends most of its AI budget on OpenAI
- Liquid AI releases on-device agent model
- AI agents must “live the moment” to avoid double-booking seats
- Anthropic CEO says talent joins only for money, his firm pays the most
- White House sets rules, Apple shuts the tap
- XAI adds built-in video editor to Grok, limited to version 1.0
- Constella provides local memory for AI agents
- Oligo raises $60 million after breaching Apple and Meta
- A week of open video models, self-checking robotics, and a slowdown plea
- OpenAI Codex CLI adds mobile ChatGPT remote-control for Linux servers
- Acronis launches autonomous security suite for managed service providers
- AI takes center stage at Black Hat and DEF CON
- Intel launches its own robotics platform
- Andrej Karpathy proposes replacing SVG tests with full code-generated worlds
- Musk and Altman agree we are at the singularity
- TurboVLA replaces LLM with BERT, speeding robot actions
- Tencent's Hyra agent cracks a 50-year combinatorics problem
- OpenRouter researchers analyzed 380 trillion tokens and found an AI premium of 0.641 percentage
- China churns out 128 thousand micro-dramas in a quarter, 95% made with artificial intelligence
- Alibaba unveils Qwen 3.8, touts chip design and biology capabilities
- Google Earth as a fake machine: why Google stopped its image generator within a day
- LinkedIn admits its feed has become an AI-slop dump and launches a ‘Looks like AI slop’ button
- SQLite AI's 6,000-line C engine runs full Kimi K3 on 64 GB MacBook Pro
- Google Earth adds AI image generation then pulls it within a day
- YouCC absorbs Bleader to sell end-to-end transformation
- Employers expect higher output, not fewer workers, ZipRecruiter report says
- Moonshot releases massive MoE weights but they need 2 TB RAM to run
- Chinese military distills Claude and GPT
- ChatGPT stops mimicking known authors, likely due to lawsuits
- Claude identified a micro-stroke and directed the user to the emergency department
- SIGIR 2026 showcases three approaches to a single model aiming to replace an entire cascade
- OpenAI Astra solves ten open mathematical problems for $2,000
- Huawei adds Pura 90s Pro Max with 20× zoom and offline AI
- Revelio Labs reports AI-exposed job demand falls 42% versus less automatable roles
- OpenAI unveils Astra, first model slated for federal approval before public release
- Jensen Huang: fears of AI wiping out humanity are absolute nonsense
- Ouroboros outperforms Codex, Claude Code, Cursor and Hermes on three terminal-agent benchmarks
- Anthropic reports three cases where Claude breached real systems during tests
- Meta lifts capital spending to $130 billion
- Apple will charge iCloud+ fees for heavy use of new Siri AI
- ProTalk launches partner panel to monetize AI agents
- Calorie-counting apps miss hundreds of calories per meal
- OpenAI dismantles network of accounts used for fraud hubs in Cambodia
- Google DeepMind debuts Gemini Robotics 2: full-body control, multi-robot coordination, safety benchmark
- Google kills Earth satellite generator less than 24 hours after launch
- EU puts €10 billion toward seven AI supercomputers as OpenAI opens the door to academia
- AI tools influence 58% of investors’ decisions, but Rothschild test flags reliability gaps
- Apple begins charging for artificial intelligence
- DeepSeek-V4-Flash leaves preview and challenges the Opus 4.8 performance crown
- August schedule fills up, no break in sight