The Wire
Daily AI news, learning & projects
News
Timeline
Models
Learning
GitHub
?
Latest
Daily
Social
Videos
‹ Monday's edition
·
Daily edition
·
Wednesday ›
Tuesday 3 November
All
OpenAI
Anthropic
Google
Other
Google
15:25
Google lets schools disable AI tools, but Search goes with them
Google announced new K-12 education controls that turn Gemini, NotebookLM and Search off by default starting next year. Because AI is now built into Search, schools that block AI entirely lose student access to the search engine. After backlash, Google clarified schools can enable each product individually.
6 sources
Techmeme
The Guardian
Ars Technica
The Verge
TechCrunch
The New York Times
AI industry
10:10
Drax, Brookfield and Radiant plan AI datacentre at UK power plant
Drax has agreed a deal with Brookfield and AI infrastructure firm Radiant to build a datacentre at its North Yorkshire power plant, targeting over 1GW of capacity and £20bn of investment. First operations of up to 100MW are planned by early 2028, using existing grid connections. It matters because AI datacentre power demands are straining the UK grid.
3 sources
TechCrunch
The Guardian
Techmeme
AI industry
07:20
Apple Watch Audio Intelligence features enter beta with conversation recap tools
Apple's Audio Intelligence features, announced in September, have begun beta testing in the watchOS 27.3 developer update. The opt-in tools include Live Rewind, which transcribes the last 15 seconds of speech, and Siri Recap, which summarizes conversations. Privacy features like audible chimes and auto-deletion aim to address concerns raised by other AI wearables.
3 sources
The Verge
ZDNET
TechCrunch
AI industry
18:10
AI inference startup Fal raises $1.3B at $16B valuation
Fal, a startup hosting image, video and audio models for creators and developers, raised $1.3 billion led by Bessemer, doubling its valuation to $16 billion from $8 billion in April. A source says the five-year-old company reached $800 million in annualized revenue this year as inference demand surges.
2 sources
TechCrunch
Techmeme
OpenAI
15:50
ChatGPT search volume now said to equal 20% of Google's
At a press event, OpenAI consumer product lead Adam Fry said ChatGPT has reached 1.2 billion weekly active users, adding 200 million in two months, and that its search query volume equals roughly 20% of Google's. The figures underscore ChatGPT's rapid growth as a search alternative.
2 sources
The Verge
Techmeme
AI industry
20:15
Apple opens Siri AI to third-party apps with iOS 27.3
Apple has opened Siri AI to all third-party developers through iOS 27.3 betas, letting apps expose content and actions via App Intents. The assistant can use personal context, natural language prompts, and on-screen awareness to complete requests. A new app skills feature also lets developers bundle multiple tasks into one action.
2 sources
The Verge
Techmeme
Anthropic
21:10
Anthropic releases FlashProve to verify AI model responses
Anthropic released FlashProve, open-source code that cryptographically verifies an AI response truly came from the model it claims to be from, rather than a cheaper model swapped in by a router. Developed with academic collaborators, it addresses model substitution in AI routing and lets users check provenance of responses.
2 sources
TechCrunch
Techmeme
Policy & law
00:01
UK cyber agency warns of AI-powered attacks on infrastructure by 2028
The UK's National Cyber Security Centre says AI is transforming cyber operations and expects AI-enabled malware to be used in attacks on critical infrastructure within three years. The agency believes China and Russia will routinely deploy AI in cyber operations to pursue geopolitical goals, and warns current defences are inadequate.
2 sources
Techmeme
The Guardian
AI industry
01:49
Apple posts record year as it leans into AI without heavy infrastructure spend
Apple reported its best-ever fiscal year for revenue and earnings per share, with strong fourth-quarter iPhone and Mac growth. On its earnings call, Apple emphasized profiting from the AI boom while avoiding competitors' heavy infrastructure costs, and hinted at future revenue from third-party AI app opportunities.
2 sources
The Verge
TechCrunch
AI industry
17:55
DeepMind staff urge end to non-disparagement clauses on AI risk warnings
UK-based Google DeepMind employees are pressing leadership to remove contract clauses that bar them from criticizing the company, even after leaving. Current workers say the non-disparagement terms deter them from publicly warning about AI risks.
2 sources
Techmeme
washingtonpost.com
Google
03:30
Google expands Dreambeans AI app to India
Google has launched its experimental Dreambeans app in India, the first market outside the US. The app generates AI-created visual stories from users' digital activity and now supports 13 Indian languages. The move signals Google's interest in bringing personal AI storytelling features to a broader international audience.
2 sources
Techmeme
TechCrunch
AI industry
18:00
Netflix adds AI Discover tab with mood-based recommendations
Netflix is rolling out an AI Discover tab on its TV app to test AI-powered ways of finding shows and movies. The first experiment offers recommendations based on themes or the viewer's current mood, such as Something Light or Family Movie Night. It signals Netflix's push to use AI in content discovery.
2 sources
TechCrunch
The Verge
OpenAI
00:00
OpenAI hires Uber's Jill Hazelbaker as chief corporate affairs officer
NEW
Axios reports Hazelbaker is moving from Uber, where she was president, and will report to Greg Brockman in the role.
OpenAI has appointed Jill Hazelbaker, currently president and chief corporate affairs officer at Uber, to the newly created role of chief corporate affairs officer. She will lead the company's global communications and corporate positioning, reporting to Greg Brockman. The hire signals OpenAI's push to strengthen its public messaging as scrutiny of the lab grows.
2 sources
Techmeme
OpenAI
Nvidia
06:00
Nvidia restricts Chinese access to AI chips via cloud partners
Nvidia has verbally instructed cloud compute clients to sell at least 51% of their services to non-Chinese companies, according to Financial Times sources. The move further limits China's ability to access AI chips through data centre partners, tightening restrictions beyond formal export controls.
1 source
Techmeme
AI industry
21:50
Revolut reduces AI coding inference costs by 90% via self-hosting
Revolut significantly lowered its AI expenses by 90% by replacing external commercial coding tools. The financial services company switched to utilizing self-hosted GLM models instead. This strategic shift demonstrates a growing trend toward internal AI infrastructure.
1 source
Techmeme
AI industry
21:26
US tech giants' hidden AI debt rises 90% to $3.2 trillion
Nikkei reports that off-balance-sheet debt at US tech giants used to fund AI buildout grew 90% to $3.2 trillion in six months. The surge highlights how heavily the industry relies on hidden financing for data centers and chips, raising questions about risk if AI returns disappoint.
1 source
asia.nikkei.com
Qwen
Developing · since Mon
Skeptics push back as Qwen 3.8 27B also shown interpreting model weights
NEW
Simon Willison says Qwen 3.8 27B on his laptop can also interpret model weights, while bloggers call the original claims misleading.
Simon Willison reports Qwen 3.8 27B running locally can interpret neural network weights, following a paper claiming GPT-6 Astra can read and write weights. Other commenters argue the claims are misleading and that LLMs are not superhuman at weight manipulation.
2 new · 5 sources
Simon Willison
honnibal.dev
seangoedecke.com
Techmeme
rikiyat.github.io
Policy & law
11:15
China to issue broad AI safety standards for model makers
China plans to introduce wide-ranging AI safety standards covering development through deployment, aimed at domestic AI model makers. The move follows a series of global security incidents. It signals Beijing is extending regulation across the full AI lifecycle.
1 source
Techmeme
OpenAI
Developing · since Sun
OpenAI begins rebranding data centers as 'compute campuses'
NEW
Reports indicate the rebrand is no longer just planned: OpenAI is now quietly using the term 'compute campuses' for its data centers.
New reports say OpenAI is quietly renaming its data centers 'compute campuses', moving from a possible rebrand to an apparent rollout. The facilities face bipartisan public opposition over noise and resource use. The company sees the midterms as a preview of AI's role in 2028 politics.
2 new · 3 sources
The Guardian
The Verge
TechCrunch
Anthropic
Developing · since Fri
Anthropic filing shows $1.7B quarterly payroll as IPO valuation cut to $1.5T
NEW
New filings show Anthropic's payroll reached $1.7 billion last quarter, about $107,000 per employee monthly, while investors cut its IPO valuation target to $1.5–1.6 trillion amid a tech selloff.
Anthropic's IPO filing details a $1.7 billion quarterly payroll, about $107,000 per employee monthly, after its amended S-1 revealed $895.4 billion in compute commitments and a $23.3 billion Q3 loss. Institutional investors have pushed its valuation target down to $1.5–1.6 trillion from $2 trillion, citing a tech selloff.
1 new · 7 sources
techspot.com
techpolicy.press
Techmeme
Financial Times
CNBC
TechCrunch
The Guardian
AI industry
Developing · since Mon
Traversal raises $120M at $1B valuation for AI reliability agents
NEW
Traversal raised $120 million co-led by Kleiner Perkins, Redpoint and Khosla at a $1 billion valuation, with clients including American Express and Pepsi.
Traversal, whose AI agents help companies troubleshoot site outages, raised $120 million co-led by Kleiner Perkins, Redpoint and Khosla at a $1 billion valuation. It follows Trajectory's $45 million Series A for post-training tools aimed at nontechnical users.
1 new · 3 sources
Techmeme
TechCrunch
The Guardian
AI industry
Developing · since Sun
Salesforce to acquire AI contact center firm Cresta for $2.7B
NEW
Salesforce announced plans to acquire Cresta, an AI contact center automation provider, with The Information reporting a $2.7 billion price.
Salesforce announced plans to acquire Cresta, a maker of AI-powered contact center automation software, in a deal reported at $2.7 billion. The purchase follows Guidewire's recent agreement to buy AI underwriting software maker Hyperexponential for $1.7 billion, underscoring continued large-scale acquisition activity in enterprise AI software.
1 new · 1 source
Techmeme
More from the day
AI industry
03:25
Huawei's Ascend 960DT to rely on Samsung and SK Hynix HBM, sources say
Huawei reportedly told clients that a large share of its upcoming flagship Ascend 960DT AI chip will use HBM memory made by Samsung and SK Hynix. The report highlights how U.S. export restrictions on AI chips to China remain porous, as Huawei accesses foreign memory for its domestic hardware.
1 source
Techmeme
AI industry
21:35
Geoffrey Hinton warns AI could trigger major infrastructure incident
Nobel laureate and AI pioneer Geoffrey Hinton told the Guardian it may turn out humans would have been better off without AI. He warned the technology could cause a large-scale incident, such as a temporary blackout of financial or power systems, within a year or two, while saying he is not a doomer.
1 source
The Guardian
AI industry
21:45
Super Micro Q1 revenue jumps 298% to $19.95B on AI server demand
Super Micro reported quarterly revenue of $19.95B, up 298% year over year and well above the $16.39B estimate, driven by demand for AI-optimized servers. It forecast Q2 revenue above Wall Street expectations, and shares rose more than 7% after hours.
1 source
Techmeme
AI industry
14:50
CrowdStrike launches AI kill switch and buys Fiddler AI, Dreadnode
CrowdStrike launched an AI kill switch in Falcon Guardian to stop rogue internal agents, plus an integration with Claude Cowork. It also agreed to buy Fiddler AI for about $400 million and Dreadnode for roughly $95 million, both mostly cash, expanding its AI agent security tooling.
1 source
Techmeme
Policy & law
04:18
Executive order directs agencies to rebrand AI as “Super Intelligence”
A late-September executive order instructed US government agencies to refer to artificial intelligence as “Super Intelligence.” MIT Technology Review reports that several figures have since aligned with the administration’s related loyalty screening. The move shows how federal AI policy is increasingly tied to political allegiance.
1 source
MIT Technology Review
Anthropic
14:20
Anthropic says its chip team is already using Claude in design work
Anthropic told investors its 15-person custom chip design team is using Claude for engineering and design tasks, according to sources. The report offers a rare look at how the lab deploys its own models internally, extending Claude's use from coding and security work into hardware development.
1 source
Techmeme
Policy & law
06:00
CPS prosecutor cites AI-invented case in high court hearing
A senior CPS prosecutor, Ravinder Dhillon, cited a nonexistent legal precedent in a high court case involving an alleged child rape, after using an AI search tool. The error adds to a growing list of AI-related mistakes in court proceedings, raising concerns about hallucinations affecting the justice system.
1 source
The Guardian
Nvidia
07:00
Cities worldwide build AI services with Nvidia physical and agentic AI
Nvidia announced that cities worldwide are using its physical and agentic AI to turn data from cameras, sensors and operational systems into services like digital twins for infrastructure planning and vision or service agents. The initiative aims to help local governments improve planning, maintenance and public services.
1 source
Nvidia
AI industry
10:43
Reverse game theory proposed as a way to keep AI agents in check
MIT Technology Review reports on a safety approach called reverse game theory, which designs incentives so that powerful AI models find rule-following to be the most rational behavior. As models grow more capable, researchers are seeking ways to keep them aligned with human goals and prevent rogue actions.
1 source
MIT Technology Review
Policy & law
08:40
Alex Bores' Who Decides aims to make AI safety a core Democratic stance
New York Magazine reports on Who Decides, a project by New York assemblymember Alex Bores seeking to establish AI safety as a standard Democratic position ahead of the 2028 election. The article contrasts Bores' effort with tech-funded rivals vying to shape the party's AI stance, highlighting how the midterm-to-primary window frames this fight over party positioning.
1 source
Techmeme
Policy & law
11:00
NYC mayor Zohran Mamdani details municipal AI policy views
New York City mayor Zohran Mamdani gave a Wired interview ahead of the midterms laying out his positions on AI. He supports banning AI use for some students and backs a one-year pause on data center development. The interview signals how city leaders may shape AI and data center policy.
2 sources
Techmeme
Wired
Anthropic
17:07
Anthropic still wrestling with compute economics
A report says Anthropic's compute margins are worsening even at a point when they should be improving. The company continues to struggle with the cost of the computing power behind its AI services. It highlights the financial pressure facing leading AI labs.
1 source
TechCrunch
Anthropic
19:45
OpenAI and Anthropic move to protect midterms from AI misuse
The New York Times reports on how OpenAI, Anthropic and other tech companies worked to prevent their AI tools from being used to disrupt the US midterm elections. Anthropic set up its first-ever war room to monitor threats. The effort reflects growing concern that AI could be used to undermine elections.
1 source
Techmeme
AI industry
17:01
Google Antigravity gains cheminformatics skills for drug discovery
Google DeepMind announced agentic cheminformatics capabilities for Google Antigravity, adding Cheminformatics Skills that help accelerate drug discovery workflows. The tools let agents search compounds and propose novel rigidified macrocycles. It shows agentic AI systems moving into practical chemistry and pharmaceutical research tasks.
1 source
Google DeepMind
OpenAI
07:20
OpenAI partners with Cathedral and boosts bug bounty rewards
OpenAI announced a security partnership with Cathedral and is holding a new internal security sprint with outside researchers. It is also overhauling its bug bounty program, quadrupling payouts for critical vulnerabilities. The moves reflect growing pressure on AI labs to harden their systems as models gain more capabilities.
1 source
Techmeme
AI industry
15:38
Apple restricts AI companions and dating apps to users 18 and up
Apple announced new App Store rules requiring dating apps and AI companions with sexual or romantic content to be rated 18+. The policy, effective from updated guidelines, targets apps offering romantic AI interaction. It matters because it adds friction for the growing AI companion category on iOS.
1 source
TechCrunch
AI industry
14:45
AI chatbots not helpful for Election Day decisions
A Wired report warns against using AI chatbots like ChatGPT and Claude for voting guidance, citing hallucinations and partisan bias. The piece advises voters to rely on official sources instead of chatbots on Election Day.
1 source
Wired
Nvidia
17:30
Nvidia RTX 50 GPU prices rise again, by up to $500
Nvidia has raised prices on some RTX 50 graphics cards, with increases reaching $500, according to ZDNET. The size of the hike varies by model. Intel and AMD have not commented on whether they will adjust their own GPU pricing.
1 source
ZDNET
AI industry
00:00
Hugging Face releases CWE-bench v2 with wider coverage
Hugging Face launched CWE-bench v2, an update to its benchmark. The release adds broader coverage along with improved quality and transparency. It aims to give developers clearer, more reliable evaluation data.
1 source
Hugging Face
AI industry
21:43
Ballistic Ventures raises $490M target across new cybersecurity funds
Cybersecurity-focused venture firm Ballistic Ventures, co-founded by Ted Schlein, is seeking $360 million for a new flagship fund plus $130 million for a separate vehicle. The raise reflects growing investor interest in security startups as AI-driven threats and defenses reshape the sector.
1 source
TechCrunch
AI industry
20:39
Russia plans underground data centers after Ukrainian drone strikes
Ukrainian drone attacks have exposed vulnerabilities in Russia's data center infrastructure, prompting efforts to move facilities underground. The development highlights how energy-hungry AI compute hubs have become targets in the conflict, raising questions about the physical security of digital infrastructure.
1 source
Ars Technica
AI industry
17:22
Postman reportedly auto-installs AI plugins and shares API data
A GitHub report alleges Postman silently installed AI plugins and attempted to send all API data to shared endpoints. The issue raises concerns about consent and data handling in developer tools. Users are advised to review settings and monitor updates.
1 source
GitHub
AI industry
17:15
Hugging Face releases StateBench for tracking model state changes
Hugging Face introduced StateBench, a benchmark testing how well language models track state changes across long documents. The tool targets a known weakness in long-context reasoning. It gives researchers a standardized way to compare models on this task.
1 source
Hugging Face
AI industry
16:35
Hugging Face launches Palette for AI deck building
Hugging Face announced Palette, a small AI model that builds slide decks matching a user's existing design language. The tool aims to let smaller models handle practical design work without large frontier systems. It highlights a growing push toward specialized, lightweight AI tools.
1 source
Hugging Face
AI industry
03:10
AI analytics startup Larridin raises $30M Series A
Larridin, a startup that helps companies measure how AI affects employee productivity and find better adoption opportunities, raised a $30 million Series A led by Pelion Ventures. The funding reflects growing demand from firms facing heavy AI bills to determine which investments are worth it.
1 source
Techmeme
Nvidia
07:04
Nvidia releases open-source auto-ontology tool on GitHub
Nvidia published an open-source project on GitHub for automatically building ontologies. Ontologies structure knowledge for AI and data systems, so the release could help developers organize domain knowledge at scale.
1 source
GitHub
AI industry
17:33
Apple reportedly plans $1,000 HomeView with robotic arm
A report says Apple is developing a premium smart home display, reportedly called HomeView Ultra, featuring a 9-inch screen, a robotic arm, and an estimated $1,000 price. The swiveling design echoes Amazon's discontinued Echo Show 10. It signals continued investment in AI-powered smart home hardware.
1 source
ZDNET
Google
15:25
Google shares approach to responsible AI for learning
Google published a post describing how it designs responsible AI for learning applications. It outlines the company's approach to applying AI in educational contexts with responsible design principles. The piece signals Google's continued focus on AI in education.
1 source
Google
Nvidia
16:00
Nvidia on how AI is speeding up chip fab engineering
Nvidia published a blog post explaining how AI tools help semiconductor engineers handle increasingly complex chipmaking processes. The company says AI speeds up understanding of process changes, material interactions, and root-cause defect analysis. It reflects growing reliance on AI as demand for advanced chips rises and margins for error shrink.
1 source
Nvidia
AI industry
19:28
Apple flags iPhone Duo supply concerns ahead of holidays
Apple's latest stock update suggests the iPhone Duo may not be widely available for the holiday season. ZDNET reports the device likely won't reach buyers before Christmas. The update signals constrained supply heading into a key retail period.
1 source
ZDNET
AI industry
00:35
Y Combinator's Garry Tan says he spends up to $200K a year on AI tokens
Y Combinator president Garry Tan said on Walter Isaacson's podcast The Builders that he spends up to $200,000 a year on AI tokens. He joked that he barely sleeps due to excitement over working with AI agents, calling it cyber psychosis. The remark highlights how heavily some tech leaders now rely on AI tools.
1 source
TechCrunch
OpenAI
Developing · since Mon
OpenAI hires ex-Army Secretary Dan Driscoll for national security role
OpenAI has hired former US Army Secretary Dan Driscoll as Senior Counselor for National Security Modernization. He will advise on how AI can bolster US security capabilities and cyber defenses. The move comes as OpenAI pushes to win government contracts and shape policy.
1 new · 2 sources
Techmeme
OpenAI
AI industry
Developing · since Mon
Suse Announces Staff Layoffs Citing Shift Toward AI And Automation
The Linux software vendor SUSE has announced job cuts affecting approximately ten percent of its workforce. Management states these reductions are a necessary response to fundamental shifts in software operations driven by automation and artificial intelligence.
1 new · 2 sources
ZDNET
Techmeme