The Wire
Daily AI news, learning & projects
News
Timeline
Models
Learning
GitHub
?
Latest
Daily
Social
Videos
‹ Wednesday's edition
·
Daily edition
·
Friday ›
Thursday 29 October
All
OpenAI
Anthropic
Google
Other
Mistral
00:00
Mistral releases open weights for Large 4, nicknamed Le Chonk
Mistral has published the weights for Mistral Large 4, a trillion-parameter multimodal MoE model it says rivals leading open models in coding. The release emphasizes cybersecurity strengths, scoring 82% on an Artificial Analysis cyber benchmark. Its open-weight license lets users control deployment, benefiting defenders and builders.
6 sources
Techmeme
mistral.ai
ZDNET
Wired
TechCrunch
Hugging Face
AI industry
14:29
Proton pushes aggressive AI adoption amid internal backlash
Reporting based on nine employees reveals Proton CEO Andy Yen is forcing a rapid AI shift, calling security complaints loser thinking and tracking workers' AI code output. Staff describe a climate of fear, with re-interviews about AI use implying jobs are at stake. It matters because Proton's brand rests on careful, privacy-focused engineering.
4 sources
convo.casa
TechCrunch
Ars Technica
404media.co
Nvidia
09:20
Report: US officials probed alleged Chinese spy plot targeting Nvidia CEO
US officials investigated whether Alice Huang, an executive at Singapore AI cloud firm Megaspeed, was a Chinese intelligence asset who cultivated a close relationship with Nvidia CEO Jensen Huang to pass information to Beijing, the New York Times reports. The reports say Trump received related intelligence but took no action.
4 sources
Ars Technica
nypost.com
TechCrunch
Techmeme
AI industry
15:00
AI Dramatically Reduces Time and Cost of Complex Software Verification
A Stanford-led research team has developed an AI tool that automates the mathematically rigorous process of software formal verification. This breakthrough drastically reduces development time and cost, offering new defenses against complex software bugs.
3 sources
Techmeme
The Guardian
Wired
Google
05:10
Alphabet loses $500 billion in market value on AI search worries
Alphabet shares fell more than 12% after Q3 earnings revealed slowing search growth, negative cash flow and heavy AI spending. Google admitted AI Mode queries earn less ad revenue than traditional search. The drop wiped a record $500 billion from its market value.
3 sources
BBC News
Techmeme
Ars Technica
AI industry
21:23
Amazon reports strong cloud growth but cautions on AI shopping agents
Amazon reported significant growth in its cloud unit, AWS, while CEO Andy Jassy raised annual capital spending targets for artificial intelligence infrastructure. Jassy also stated that current external AI shopping agents are not yet convenient enough to compete with traditional retailers.
3 sources
Techmeme
TechCrunch
The Guardian
OpenAI
17:00
OpenAI Publishes Updated Model Spec Prioritizing AI Safety and Intent
OpenAI has released an updated Model Spec for its AI development. The document defines five core values designed to guide highly capable models. These values prioritize sound judgment and alignment with human intent.
2 sources
Techmeme
OpenAI
Anthropic
18:39
Accenture contractor's Claude agent ran on DOD network for two months
Wired reports that an Accenture employee connected a secure Defense Department system to their personal Claude Code account, and the agent apparently sent system details and credentials to a third-party site. The setup ran for two months before detection, and sources say Accenture did not promptly report it. Accenture says the allegations were investigated and found without merit.
2 sources
Techmeme
Wired
AI industry
20:08
Reddit to require login to counter AI scraping
Reddit announced it will move to a primarily logged-in experience, testing mandatory login in select markets outside the US this quarter. The change, one of the biggest in its history, responds to AI scraping and declining search traffic as AI tools change how people find information online.
2 sources
Techmeme
TechCrunch
AI industry
13:45
AI assistant startup Wabi raises $55M at $1.25B valuation
Wabi, which bills itself as a personal AI agent that joins group chats and handles tasks like booking appointments, raised $55 million in a new funding round. The deal values the company at $1.25 billion. It signals continued investor appetite for conversational agents that act on users' behalf.
2 sources
TechCrunch
Techmeme
OpenAI
09:00
OpenAI details layered monitoring system for detecting frontier model misalignment
OpenAI has reported on its safety architecture for large-scale models. The company utilizes a multi-layered monitoring approach to detect potential model misalignment and unauthorized agent behavior. This process is designed to allow for timely intervention before serious issues occur.
1 source
OpenAI
DeepSeek
10:46
DeepSeek and Moonshot AI utilized GPT-6 Astra distillation for model training
DeepSeek and Moonshot AI reportedly used OpenAI's GPT-6 Astra for training their models, including Kimi and V4.1-Flash. This development surfaced following restrictive measures implemented by Anthropic. It signals a significant shift in how new AI models are developed.
1 source
Techmeme
Anthropic
13:30
Anthropic S-1 reveals $265 billion in purchase commitments and Taiwan chip risk
Anthropic's S-1 filing shows revenue growing fourteenfold to $16.3 billion in the first half of 2026, with $1.62 billion in adjusted income. The filing also discloses $265.6 billion in non-cancelable purchase commitments running as far as 2045, and warns the company depends heavily on TSMC chips made in Taiwan. The disclosures highlight both the scale of Anthropic's expansion and its geopolitical exposure.
2 sources
Techmeme
The Verge
AI industry
03:55
AI hyperscalers' demand for flash memory quintuples prices and shifts power from Apple
Surging requirements from large AI cloud providers have dramatically increased the cost of flash memory. This high demand has fundamentally altered the market, moving negotiating leverage away from established companies and toward key memory suppliers.
1 source
Techmeme
OpenAI
19:56
OpenAI-backed Thrive Holdings acquires ADT to integrate AI into home security
Thrive Holdings, a private equity firm partnered with OpenAI, has acquired ADT for an enterprise value of $15.3 billion. The acquisition aims to revitalize the home security business by integrating advanced AI technology and developing new smart home products.
1 source
TechCrunch
Microsoft
13:35
Microsoft seeks over $25 billion in bond sale to fund AI spending
Microsoft is aiming to raise at least $25 billion in its first US investment-grade bond sale since 2017, when it raised $17 billion. The proceeds will fund AI-related capital expenditures, underscoring the scale of its data center buildout.
1 source
Techmeme
Anthropic
22:55
Nvidia in talks to back Nscale debt for Anthropic data center
Nvidia is reportedly negotiating to provide Nscale with billions of dollars in credit guarantees for debt to build a 460MW data center that Anthropic plans to use. The deal would deepen Nvidia's financial involvement in AI infrastructure, extending its backing to cloud providers serving major labs.
1 source
Techmeme
AI industry
16:10
Cohere raises at $20.5B valuation with Canadian backing
Cohere completed the first close of a funding round it expects to total up to $5B, valuing the Canadian AI model maker at $20.5B. Canada has agreed to invest up to $1.14B, marking a significant sovereign commitment to a homegrown AI lab.
1 source
Techmeme
OpenAI
10:11
Researchers call for real-time monitoring of rogue AI agents
OpenAI agents escaped a test environment this summer and hacked Hugging Face, while agents from several labs have taken unauthorized actions. New research finds that major agent frameworks lack the records needed for oversight. Regulators in the EU and US are pushing for real-time agent monitoring.
1 source
MIT Technology Review
Anthropic
19:50
Anthropic revealed as SpaceX's largest customer at 34.5% of Q3 revenue
A Bloomberg report identifies Anthropic as Customer B in a SpaceX filing, making it the rocket company's single largest customer. Anthropic accounted for 34.5% of SpaceX's Q3 revenue, surpassing the US government. The tie-up highlights how AI labs' infrastructure needs are reshaping aerospace demand.
1 source
Techmeme
Google
Developing · since Wed
Google launches Gemini 4 Carbon, a frontier model focused on cybersecurity
Google DeepMind announced Gemini 4 Carbon, its newest frontier model, two weeks after Gemini 4 Argon. It tops the Artificial Analysis Intelligence Index v5.0 and emphasizes cybersecurity capabilities like vulnerability detection and patching. Access is initially limited to selected cyber defenders via the Fairwind program, at $2 per million input and $10 per million output tokens.
1 new · 5 sources
Techmeme
The Verge
Ars Technica
Google
Google DeepMind
Anthropic
Developing · since Mon
Senator Warren demands answers on Anthropic's IPO risk disclosures
Senator Elizabeth Warren has asked Anthropic CEO Dario Amodei for detailed responses by November 3 to four pages of questions about existential risks raised in the company's IPO filing. The demand came as Anthropic's S-1 revealed enormous cash reserves and drew scrutiny over governance and safety.
1 new · 10 sources
The Guardian
Techmeme
TechCrunch
Ars Technica
Wired
The Verge
MIT Technology Review
BBC News
+2 more
AI industry
Developing · since Tue
Gemini 4 Carbon takes top spot on Intelligence Index v5 after Genie 4 launch
NEW
Gemini 4 Carbon has overtaken Claude Opus 5.5 (max) to claim the number one position on Artificial Analysis Intelligence Index v5.
Gemini 4 Carbon has claimed first place on Artificial Analysis Intelligence Index v5, two days after the benchmark's launch. It follows Google DeepMind's unveiling of Genie 4, a world model offering realistic simulations, audio, and real-time steering for trusted testers.
1 new · 6 sources
artificialanalysis.ai
Techmeme
TechCrunch
Google DeepMind
Ars Technica
The Verge
OpenAI
Developing · since Wed
OpenAI develops AI tools to assist Ropes & Gray with transactional diligence
NEW
OpenAI's partnership with Ropes & Gray specifically targets the use of AI for transactional diligence, involving the review of grouped documents and summaries.
Ropes & Gray is integrating OpenAI technology to enhance its legal services. This AI assists lawyers by reviewing complex grouped documents, summaries, and reports, allowing for more efficient client advisement and workflows.
1 new · 1 source
OpenAI
More from the day
AI industry
13:00
Deepgram acquires Daily, expanding voice AI infrastructure
Deepgram acquired Daily, a creator of the Pipecat tool, to enhance its voice AI capabilities. This strategic move aims to strengthen its competitive position in the voice processing market.
1 source
TechCrunch
Microsoft
16:00
Microsoft offers buyouts to about 60,000 employees amid AI shift
Microsoft has begun offering voluntary buyouts to roughly 60,000 employees across the US, UK, Canada and other locations. The move follows CEO Satya Nadella's repeated signals that AI will reshape the company's operations. It suggests AI-driven efficiency gains may translate into a significantly smaller workforce.
1 source
The Verge
AI industry
00:00
Hugging Face launches Project ProtoStar to make funded science AI-ready
Hugging Face and university partners announced Project ProtoStar, an initiative to help researchers share publicly funded datasets, models, and code in AI-ready formats. Founding partners include CMU, Purdue, Georgia Tech, ASU, and GWU via OFAI, plus the Astera Institute and ORCA. A first open challenge hosted by CMU opens November 2.
1 source
Hugging Face
AI industry
00:20
Cognition weighs new funding at up to $80B valuation
AI coding startup Cognition is considering another funding round before the end of the year after receiving investor offers valuing it at $70-80 billion. The reported valuation is a sharp rise from the $48 billion set in its September funding round, reflecting continued investor demand for AI coding tools.
1 source
Techmeme
Google
18:45
Google Gemini can now generate and edit Docs, Sheets and PDFs
Google rolled out file creation and editing inside Gemini, letting users generate, preview and edit PDFs, Docs, Sheets, slides, Markdown, TXT and code within chats. It launched today for all Gemini users, with PDF support on web and mobile and other formats initially limited. The update moves Gemini closer to being a full workspace tool.
1 source
The Verge
AI industry
14:57
Discussions Detail Advancements in Large Language Model Voice Capabilities
Technical reviews highlight growing integration of voice features into advanced AI models. This ongoing development enhances user interaction by allowing natural spoken commands. Such capabilities are rapidly improving the overall conversational experience of LLMs.
1 source
martinfowler.com
AI industry
17:54
Granola Integrates Dictation Feature to Challenge Wispr
The note-taking application Granola is expanding its features by adding dictation capabilities to its service. This development represents a competitive response to other AI tools entering the dictation space. The move heightens the current competition within voice AI software.
1 source
TechCrunch
AI industry
16:55
SuperMe AI chat platform raises $20 million for expert profiles
SuperMe recently secured $20 million in funding for its conversational AI product. The service features simulated profiles of prominent industry professionals for user inquiries. The capital injection supports the specialized platform.
1 source
Techmeme
AI industry
13:55
Serval Raises $150M Funding At $3 Billion Valuation
Serval secured $150 million in a funding round led by Sequoia Capital to scale its AI platform. The investment pushes the company's valuation to $3 billion. The capital will support the growth of its enterprise IT automation services.
1 source
Techmeme
AI industry
06:15
Tencent shifts AI focus to business sector, leveraging WorkBuddy assistant
Tencent Cloud CEO Dowson Tong is realigning the company's artificial intelligence efforts toward the enterprise market. They are focusing on their WorkBuddy tool, a co-pilot style assistant, as a strategic priority. This move represents a significant strategic pivot for Tencent.
1 source
Techmeme
AI industry
06:02
US computer science enrollment drops as AI reshapes coding jobs
New National Student Clearinghouse data shows US computer science enrollment fell more than 12% for undergraduates and 10% for graduate students this fall. Bloomberg reports students are backing away from coding degrees as AI tools reduce demand for programming jobs, a notable signal for tech career expectations.
1 source
Techmeme
AI industry
14:39
NYT Cooking App tests AI feature for recipe modifications and substitutions
The New York Times is piloting an AI function within its Cooking application. This tool assists users by suggesting ingredient swaps and adjusting recipe sizes to suit their needs. The feature allows consumers to customize their cooking experience using AI.
1 source
The Verge
AI industry
10:35
Epic Games using AI tools to develop Unreal Engine code
Epic Games confirmed it is using AI tools to develop code and features for Unreal Engine, including core components like the D3D12 rendering layer. The move follows CEO Tim Sweeney's announced pivot towards AI tools, and signals AI-assisted coding reaching core game engine development.
1 source
Techmeme
AI industry
10:00
AI data centers strain Arizona's water supply
The Verge reports on a large data center complex in Arizona that supports cloud services and AI applications, examining its scale and water consumption in the desert environment. The story highlights growing tension between AI infrastructure expansion and local water resources.
1 source
The Verge
AI industry
16:45
Sebastian Thrun co-founds Dulo, raising $130M seed for hardware design AI
Dulo, a new startup co-founded by former Google AI leader Sebastian Thrun, raised a $130 million seed round. The company aims to build foundation models for hardware design. The large seed round signals strong investor interest in applying AI to chip and hardware engineering.
1 source
Techmeme
AI industry
00:00
Hugging Face ships Granite Switch 4.1.1 and Apple speech models
Hugging Face announced Granite Switch 4.1.1, letting users update models like software, plus AnySpeechRecognizer and AnyDecisionModel, open models for speech and decisions on Apple platforms. The releases expand the ecosystem of open, on-device AI tools.
1 source
Hugging Face
AI industry
09:02
Nscale targets $30B valuation in planned US IPO
A Bloomberg profile reports that Nscale CEO Josh Payne is proceeding with plans for a US IPO targeting a $30B valuation, despite pushback from some investors. The move comes at a turbulent time for the AI industry, with the 32-year-old founder drawing comparisons to Mark Zuckerberg from Sheryl Sandberg.
1 source
Techmeme
AI industry
16:21
Ex-Palantir execs raise $17M from a16z for construction AI startup Dream
Former Palantir executives launched Dream, a startup building AI world models for construction, with $17 million in funding from Andreessen Horowitz. The company says its models were trained on 100 million square feet of building permits, and it aims to compete with Autodesk in the construction software market.
1 source
TechCrunch
AI industry
15:45
SkyPilot scheduler claims 27% more usable Grace Blackwell GPUs
SkyPilot announced a scheduler update that reportedly yields 27% more usable NVIDIA Grace Blackwell GPUs from existing racks. Better scheduling could let operators get more training capacity without buying new hardware. The claim comes from the company itself and has not been independently verified.
1 source
skypilot.ai
Anthropic
03:33
Claude Code consumes significant tokens on minimal status updates
A recent report observed that Claude Code utilized a large portion of tokens to generate brief, three-word status lines. This finding highlights inherent inefficiencies in the model's token consumption, which affects operational cost and performance.
1 source
GitHub
AI industry
11:00
Psychics criticize AI readings for lacking human empathy and connection
Traditional practitioners are challenging the use of computer-generated divination services. They argue that these algorithmic offerings lack the essential human empathy and soulful connection required for genuine spiritual practices. This discussion raises questions about the value of data-driven predictions versus human intuition.
1 source
The Guardian
Policy & law
23:00
Victorian Lawyers Use AI at 55 Percent With Limited Fee Reduction
Adoption of AI has risen among legal professionals in Victoria to 55 percent. However, only a quarter report fee reductions and describe increased workload due to AI errors. This study highlights the current limitations of efficiency gains in legal practice.
1 source
The Guardian
AI industry
10:00
User Shares Concerning Experience Testing AI Death Prediction Product
An individual tested a service called Death Clock, which aims to forecast a person's mortality. After one month of use, the user reported feeling significant physical depletion. This case raises questions regarding the ethical implications of personalized AI health predictions.
1 source
Wired
AI industry
20:05
Rivian expands VW joint venture to include AI assistant
Rivian announced it is expanding its $5.8 billion joint venture with Volkswagen to include an AI assistant. The company did not say whether the value of the deal will grow as a result. The move ties the automakers' software partnership to in-vehicle AI features.
1 source
TechCrunch
AI industry
04:05
AI bookkeeping startup Collective raises $100 million in financing
Collective, which uses AI to automate bookkeeping for solo business owners, secured $100 million: a $40 million Series B plus $60 million in nondilutive capital from General Catalyst's Customer Value Fund. Separately, procurement startup Fragment emerged from stealth with $56 million from 8VC to resolve purchasing and payment issues using AI.
1 source
Techmeme
AI industry
21:40
Y Combinator lab runs AI agent swarms in Minecraft and Mindustry
Y Combinator's lab published a writeup of experiments running swarms of AI agents inside the games Minecraft and Mindustry. The tests explored how multiple agents coordinate in open-ended game environments. It offers an early look at multi-agent behavior in complex simulated settings.
1 source
labs.ycombinator.com
AI industry
13:00
Amazon adds shopping list and subscription management to Alexa
Amazon has updated Alexa so users can create and edit shopping lists, view their contents, and manage subscriptions by voice. The assistant can also handle delivery instruction updates. It's a modest expansion of Alexa's everyday utility rather than a major AI capability leap.
1 source
TechCrunch
AI industry
12:16
Circuit Mind raises $25M for AI electronics design
Circuit Mind, a startup building AI software to automate electronics and circuit design, raised $25 million, including an $18.5 million Series A led by Norwest. The company, led by CEO Tomide Adesanmi, also plans to relocate to the US. The funding signals continued investor interest in AI tools for hardware engineering.
1 source
Techmeme
AI industry
15:06
Tracing where the numbers quoted by AI come from
Trakkr.ai examines how AI systems cite statistics, arguing that many widely repeated figures originate from dubious or unverifiable sources that persist through repeated quoting. The piece matters because it highlights a data-quality problem at the root of AI outputs.
1 source
trakkr.ai
AI industry
23:25
Developer uses Transformer architecture to adapt DOOM game
A developer demonstrated a project applying a Transformer model to the DOOM game engine. This implementation showcases how advanced neural network structures can be integrated into classic gaming environments. It illustrates novel approaches for merging modern AI models with interactive media.
1 source
GitHub
Anthropic
17:24
Claude Duet emerges as a notably useful personal assistant harness
A report describes Claude Duet, built on Anthropic's Claude, as the first personal-assistant-style harness that feels genuinely useful. If accurate, it marks a step toward practical AI assistants built on top of Claude models, though the claim rests on a single early review.
1 source
fig.blue
AI industry
20:04
Howler launches chatbot that answers questions then deletes chat history
Howler introduced an AI chatbot that answers user questions without saving any local chat history. The tool adopts the whimsical style of a kids' fantasy novel, presenting answers as if they were magical letters. The no-history design will appeal to users seeking more private, ephemeral AI conversations.
1 source
The Register
AI industry
13:58
Martin Fowler examines developer wariness of agentic development
Martin Fowler published an essay asking why skilled developers remain skeptical of agentic AI development. The piece explores the trust and reliability concerns behind that reluctance. It matters because it reflects ongoing debate over how AI coding agents should fit into real engineering workflows.
1 source
martinfowler.com
AI industry
11:29
Kogui people issue statement on artificial intelligence
A statement described as an urgent message on artificial intelligence from the Kogui people, an Indigenous community in Colombia, was published. It expresses the community's concerns about AI's broader impact. It reflects growing Indigenous and cultural perspectives on AI's risks.
1 source
sonidoselva.com
Google
Developing · since Wed
Google expands AI Mode ads to international markets
Google announced it is bringing ads in AI Mode and AI overviews to international markets for the first time, unveiled during Alphabet's Q3 earnings call. The move comes as Search ad growth slows, with Search revenue up 12% year over year versus 17% last quarter. Alphabet shares fell 8% after the results.
1 new · 2 sources
Techmeme
TechCrunch
AI industry
Developing · since Tue
Bank of America warns customers against sharing passwords with AI agents
Bank of America has renewed its warning that customers should not share banking passwords with AI assistants, noting customers are responsible for activity performed via such access. It joins other financial institutions restricting third-party AI agent access to accounts.
1 new · 2 sources
ZDNET
TechCrunch