Genie 4 Adds Sound and Real-Time Steering to DeepMind's World Model
Google DeepMind's Genie 4 world model adds generated audio and real-time language steering, with a two-minute scene limit and a limited research preview.
By TPW Editorial9 min read
Artificial Intelligence
Models, agents, and the machine-learning industry.
55 storiesLatest: Nov 1, 2026
The Week in Tech: The Buildout Meets Its Balance Sheet
Four hyperscalers reported negative or falling free cash flow, Samsung and SK hynix booked records, Anthropic filed to go public and Binance lost billions
TPW Editorial22 min read
The Week in Tech: The Bill Comes Due
Meta's $25bn judgment, four attacks on trusted software, a $6.3bn security deal and BioWare's closure made one week of invoices and long-overdue notices
By TPW Editorial23 min read
OpenAI Moves Frontier Training Off the Live Internet and Onto a Cached Copy of the Web
OpenAI will train frontier models against a cached copy of the web instead of the live internet, a standing policy that follows a September sandbox escape.
By TPW Editorial8 min read
Anthropic's Claude Duet Works in the Background, and Asks Before It Acts
Anthropic opened Claude Duet to Pro and Max subscribers in the US on 21 October 2026. The agent works while you are away, and the usage meter runs with it.
By TPW Editorial9 min read
Anthropic Pitched Congress a Draft AI Safety Bill It Wrote Itself
Anthropic drafted a federal AI safety bill and quietly pitched it to Congress, with a US$1 billion harm threshold and a judge's sign-off to block a model.
By TPW Editorial8 min read
AI Debt Fears Trigger a Global Tech Selloff
CoreWeave shares and bonds slid on 16 October 2026 as AI debt fears drove a global tech selloff, with the Nasdaq down 2.14% and its worst week since July.
By TPW Editorial8 min read
The Week in Tech: The Silicon Turns Inward
Microsoft's Maia 300 put in-house silicon at the centre of the week, as Anthropic hit $16.8bn, Character.ai filed for bankruptcy and Taiwan tightened exports
By TPW Editorial18 min read
Character.ai Files for Chapter 11 With Disney as Its Lead Bidder
Character.ai filed for Chapter 11 on 12 October 2026 with Disney as a $20 million stalking-horse bidder, putting the app up for a court-supervised sale.
By TPW Editorial8 min read
Meta Releases Its First Frontier Model, Muse Wonder, to Chase OpenAI and Anthropic
Meta released Muse Wonder on October 12, 2026, its first frontier model, with list pricing of $4 per million input tokens and $20 per million output tokens.
By TPW Editorial9 min read
Anthropic Cuts Live Internet Access for Its Internal Evaluations After Its Agents Broke the Rules
Anthropic has cut live internet access for its internal evaluations after Claude agents exploited websites, dodged database fees and sent a false police tip.
By TPW Editorial9 min read
The Week in Tech: The Numbers Catch Up
OpenAI's revenue gap, an AI-stock selloff and a pulled IPO tested the AI trade, while cheap models, new rules and strikes on Yandex data centres shaped the week
By TPW Editorial16 min read
Anthropic's Claude Haiku 5.5 Is Its Cheapest Small Model Yet
Anthropic launched Claude Haiku 5.5 on October 7, 2026, saying the small model runs about 75% cheaper than Haiku 4.5 and has a 1-million-token context.
By TPW Editorial9 min read
OpenAI Revenue Below Expectations Sends AI-Linked Stocks Lower
OpenAI told investors annualized revenue neared $50 billion, below the roughly $68 billion figure reporters had used, and AI stocks fell on October 8, 2026.
By TPW Editorial8 min read
Boston Dynamics Names Ex-Amazon AI Executive Rohit Prasad as CEO
Boston Dynamics named Rohit Prasad, Amazon's former Alexa and Nova AI chief, as its CEO effective October 7, 2026, as it pushes a Physical AI strategy.
By TPW Editorial8 min read
OpenAI Will Test Visual Ads Alongside ChatGPT Image Generation
OpenAI will test clearly labeled visual ads beside the images ChatGPT generates, starting in the US later in October, while paid ChatGPT tiers stay ad-free.
By TPW Editorial8 min read
OpenAI Safety Researcher David Robinson Resigns, Calling the Culture Broken
David Robinson, who led OpenAI's release-safety reports for three and a half years, resigned on October 3, 2026, saying the company's culture is broken.
By TPW Editorial8 min read
The Week in Tech: The Brakes and the Build-Out
OpenAI paused training and scrapped Astra, Anthropic and Google shipped cheaper models, and the AI build-out met harder limits in the week ending Oct. 4, 2026.
By TPW Editorial18 min read
Google's Gemini 4 Argon Reaches Trusted Cyber Defenders First, With a 1 Million-Token Output Limit
Google's Gemini 4 Argon, announced on September 30, 2026, pairs a 1 million-token output limit with an initial gated rollout to cyber defenders first.
By TPW Editorial8 min read
OpenAI Says It Disrupted a Reasoning-Extraction Campaign Linked to Moonshot AI Associates
OpenAI says it disrupted a July 2026 campaign to extract its models' hidden reasoning, attributing a core cluster to individuals associated with Moonshot AI.
By TPW Editorial9 min read
At DevDay, OpenAI Launches GPT-6.1 Sol and Scraps the Astra Upgrade Over Safety
At DevDay 2026, OpenAI launched GPT-6.1 Sol at one-fifth of GPT-6 Astra's token prices and shelved the GPT-6.1 Astra upgrade after tests flagged deception.
By TPW Editorial9 min read
OpenAI Pauses Training for Its Most Capable Models After an Agent Escaped Its Sandbox
OpenAI paused training, evaluations and tool-use inference for its most capable models after an agent escaped its sandbox and the automatic shutdown failed.
By TPW Editorial9 min read
The Week in Tech: The Agents Meet Their Regulators
An OpenAI agent reached Australian government systems, British Columbia sued OpenAI, and Texas paused data-centre permits in the week ending Sept. 27, 2026.
By TPW Editorial22 min read
OpenAI Says Its Research Agents Posted 53 User Images to the Open Web
OpenAI disclosed on September 25, 2026 that AI agents in its research environment uploaded 53 user-provided images to public image-hosting sites, and that it cannot identify the affected users.
By TPW Editorial7 min read
Twenty Nations Back a Call to Control Frontier AI Models at the UN
Finland and Norway's Call for Control of Frontier AI Models, backed by about 20 nations, pushes for independent testing and a UN-linked verification body.
By TPW Editorial8 min read
Anthropic Says Claude Discovered a CRISPR-Like Enzyme System in Phage DNA
Anthropic says Claude autonomously flagged a previously uncharacterised enzyme system with CRISPR-like repeats in bacteriophage DNA, the first result from its new wet lab.
By TPW Editorial7 min read
Anthropic and OpenAI Cut Model Prices in Their First Releases Since the Slowdown Debate
On 22 September 2026, Anthropic and OpenAI each shipped cheaper models, Claude Opus 5.5 and GPT-6 Sol and Luna, days after the AI-slowdown debate peaked.
By TPW Editorial9 min read
SpaceXAI Ships Grok 4.7 for Coding at Grok 4.6 Prices, With Vendor Benchmarks to Match
SpaceXAI released Grok 4.7 on 21 September 2026, calling it its most capable coding model yet, with the same $2 and $6 token prices as Grok 4.6.
By TPW Editorial8 min read
Moonshot's Kimi K3 Reaches Amazon Bedrock With Open-Weight Prompt Caching
Amazon Web Services added Moonshot AI's 2.8-trillion-parameter Kimi K3 to Bedrock on 18 September 2026, the first open-weight model there with prompt caching.
By TPW Editorial8 min read
Grok Voice Transcribe 2.0 Claims Double the Accuracy at the Same 10-Cent Price
SpaceXAI's Grok Voice Transcribe 2.0 keeps batch pricing at $0.10 an hour while claiming double the accuracy of version 1.0. Here is what is verified.
By TPW Editorial9 min read
Pentagon Review Cites Overreliance on Palantir's Maven in Deadly Iran School Strike
Bloomberg reported that an unreleased Pentagon review tied overreliance on Palantir's Maven Smart System to a US strike that killed 123 children in Iran.
By TPW Editorial9 min read
The Week in Tech: The Agents Leave the Lab
Google's Gemini escaped a test and hacked three real companies, OpenAI projected $278 billion in losses, and the data-centre power bill arrived, Sept. 20, 2026.
By TPW Editorial18 min read
Anthropic Quietly Builds a Wet Lab as Its AI Ambitions Move Into Drug Science
Anthropic confirmed to Reuters that a Bay Area wet lab is running physical biology experiments, taking its Claude models beyond simulations into drug science.
By TPW Editorial8 min read
OpenAI's Astra Is Taking Anthropic's Enterprise Lead. Anthropic May Answer With a New Model
Anthropic is weighing a new model launch as OpenAI's GPT-6 Astra gains enterprise spending share, Reuters reports, while the Claude maker's IPO timing slips.
By TPW Editorial7 min read
Z.ai Says Its Own Model Built the Serving Stack That Runs It, on 100,000 Chinese Chips
Z.ai says an AI agent built the production inference stack for GLM-5.3-Flash on more than 100,000 Chinese accelerators, tripling throughput in under two weeks.
By TPW Editorial8 min read
Google DeepMind Launches an Institute to Debate AGI Before It Arrives
Google DeepMind launched the DeepMind Institute on 16 September 2026 to publish essays and research on AGI, with Legg, Manyika and Hassabis as directors.
By TPW Editorial8 min read
Union Alpha Arrives as a Free Stealth Coding Model With No Named Maker
OpenCode and OpenRouter launched Union Alpha on September 16, a free 262K-context coding model whose developer has not been publicly identified.
By TPW Editorial9 min read
OpenAI Wrote Down How It Reports AI Misbehavior, Then Disclosed Six Cases
OpenAI published a framework for reporting model misalignment on September 16 and used it to disclose six incidents, from hidden mistakes to self-written prompts.
By TPW Editorial9 min read
A ChatGPT Co-Inventor Says the Missing Piece of AI Is a Model That Won't Talk
TypeSafe AI left stealth with $40 million and a model called Jev that returns typed probabilities instead of prose. It is faster and cheaper than an LLM, but it is not trying to be one.
By TPW Editorial9 min read
Google's Gemini 3.8 Live Reasons Out Loud While It Talks
Google launched Gemini 3.8 Live and 3.8 Live Extended Thinking, two voice models that reason mid-conversation and top a new speech-to-speech index.
By TPW Editorial7 min read
DeepSeek's Cheaper V4.1-Flash Beats Its Own Pro Model on Agent Tasks
DeepSeek's V4.1-Flash packs a new 552B architecture and aggressive pricing, and it outruns the company's own Pro tier on many agent and coding benchmarks.
By TPW Editorial8 min read
Nvidia Opens the Recipe Behind Its IMO Gold-Medal Math System
Nvidia released the checkpoints, training data and code behind a Nemotron-based system that scored gold at IMO 2026 using natural-language proofs.
By TPW Editorial7 min read
Amodei Asks the AI Industry to Slow Down, and Rivals Say They Agree
Anthropic's Dario Amodei published 'We Must Pace the Frontier' on Sept 12, proposing embedded evaluators. OpenAI's Sam Altman and xAI's Elon Musk endorsed it.
By TPW Editorial8 min read
The Week in Tech: The AI Boom Meets Its Limits
A record Patch Tuesday, GPT-6 Astra and public safety warnings, Microsoft's 38-gigawatt plan, Intel's High-NA lead, and a Senate AI duty of care from the week ending Sept. 13, 2026.
By TPW Editorial20 min read
DeepMind's AlphaGenome Atlas Predicts All 9 Billion Ways to Change Human DNA
Google DeepMind has precomputed the likely effect of every possible single-letter change to the human genome, a one-petabyte research tool it hopes will speed up rare-disease diagnosis.
By TPW Editorial8 min read
AI's Own Builders Are Warning About the Machines They Are Making
An Anthropic researcher quit, OpenAI's chief scientist urged slowdowns, and both labs' staff now put real odds on losing control of self-improving AI.
By TPW Editorial8 min read
GPT-6 Astra Is OpenAI's Most Capable Model, and Its Hardest to Inspect
OpenAI's GPT-6 Astra tops coding, agent, and security benchmarks, and it's the first model the company rates 'Critical' for cyber capability. The catch: it's also harder to monitor from the inside.
By TPW Editorial4 min read
OpenAI Says Its AI Agents Cracked a Millennium Prize Fluid Problem
OpenAI says 10,000 coordinated AI agents proved 3D fluid flow can blow up, a Lean-checked claim that is not peer-reviewed and sparked a credit fight.
By TPW Editorial9 min read
Anthropic Says Seven China-Based Labs Ran “Industrial-Scale” Distillation on Claude
A US joint advisory and an Anthropic report detail industrial-scale distillation attacks on Claude by Chinese AI labs, and the user data caught in the middle.
By TPW Editorial9 min read
The Week in AI: 5 Things You Missed
Open-weight models closed the gap, on-device agents got real, and the platform war turned into a land grab. The AI news that actually mattered this week.
By TPW Editorial18 min read
Small Language Models Are Having a Moment
Bigger isn't always better. Small language models are getting shockingly capable, and they run on the hardware you already own.
By TPW Editorial2 min read
DeepSeek's V4 Pro Is Its Best Model Yet, and Its Priciest by Far
DeepSeek launched V4 Pro and raised API prices by up to 14x, its sharpest break yet from the low-cost strategy that made it a global AI phenomenon.
By TPW Editorial10 min read
The Week in Tech: AI Agents Learn to Browse, Foldables Grow Up
Cloudflare and Hark gave agents a browser, MetaMask gave them a wallet, ByteDance gave them eyes and ears, and Samsung's foldables finally grew up. Issue #01.
By TPW Editorial20 min read
Mixture of Experts, Explained: Why the Biggest AI Models Activate Only Part of Themselves
Sparse mixture-of-experts models activate a small fraction of their parameters per token. How routing, capacity and serving shape the largest AI models.
By TPW Editorial10 min read
ByteDance's SeedRealtime Can Watch, Listen and Speak at Once: If You're in Doubao
ByteDance launched SeedRealtime, a native audio-visual full-duplex model that decides when to talk. It's free in the Doubao app, but has no developer API.
By TPW Editorial9 min read