The Wire
Daily AI news, learning & projects
News
Timeline
Models
Learning
GitHub
?
Latest
Daily
Social
Videos
‹ Wednesday's edition
·
Daily edition
·
Friday ›
Thursday 15 October
All
OpenAI
Anthropic
Google
Other
Meta
01:00
Meta's Muse Spark Open runs locally on Apple Mac Studios
NEW
The Guardian reports Meta claims Muse excels at multi-step reasoning and complex software engineering, as the company races to catch up to Chinese rivals.
Meta released Muse Spark Open, an 816B-parameter open-weight multimodal model it calls the most capable US open-weight model to date. It runs locally on Apple M5 Ultra Mac Studios, in a collaboration between the two companies. The launch is Meta's bid to compete with leading Chinese open-weight rivals.
5 sources
The Guardian
TechCrunch
The Verge
Techmeme
Hugging Face
Anthropic
18:20
Anthropic gives METR employee-level access for independent safety evaluations
Anthropic has onboarded six METR employees with near-employee access to internal tools and systems so the nonprofit can independently assess its risks and safety practices. METR will publish its findings, and Anthropic cannot veto publication, though it can require narrow redactions. The move aims to close the gap between what outside evaluators and the company itself know about frontier model risks.
3 sources
TechCrunch
Techmeme
Anthropic
Nvidia
16:00
Nvidia unveils Nemotron 3.5 Super VL and edge Bolt models
Nvidia released Nemotron 3.5 Super VL, a 120-billion-parameter open model for multimodal enterprise agents, alongside two Bolt edge models for local deployment. Nvidia says Super VL is five times faster than the next best open agentic model and matches its largest reasoning model at about one-fifth the size, with versions post-trained by dozens of organizations including Scale AI and EY.
3 sources
Techmeme
Nvidia
ZDNET
AI industry
13:00
Amazon launches two Bee AI wearables with shopping tie-ins
Amazon's Bee debuted two AI devices: the $74.99 Frontier wristband, available now, and the $99.99 Eclipse magnetic phone attachment coming soon. Both record conversations and produce summaries and to-dos, with Alexa and Amazon shopping features behind an $11.99 monthly subscription.
3 sources
Techmeme
The Verge
TechCrunch
Google
14:00
Google launches Universal Cart shopping tool in the US
Google has rolled out its Universal Cart shopping tool in the US across Search and the Gemini app ahead of the holidays. The tool saves items from multiple retailers in one place, alerts users to price drops and restocks, and suggests the best credit card for checkout.
3 sources
TechCrunch
The Verge
ZDNET
AI industry
19:56
Bluesky expands Attie AI tool to build games, sites and quizzes
Bluesky announced at TechCrunch Disrupt that its Attie platform now supports what CEO Toni Schneider calls social vibe coding, letting users create games, quizzes, polls, and websites tied to their AT Protocol identity. The move aims to give ordinary users developer-like building power while leveraging Bluesky's distribution across its 47 million-user network.
3 sources
Techmeme
The Verge
TechCrunch
OpenAI
12:25
FTC opens probe into OpenAI and Anthropic safety practices
The FTC has sent roughly 50-page civil investigative demands to OpenAI and Anthropic, Axios reports, seeking documents on safety incidents and consumer protection practices. A separate report says the agency is examining whether security incidents harmed Americans and could fine AI firms that delay breach disclosures. The probe signals heightened federal scrutiny of leading AI labs.
2 sources
Ars Technica
Techmeme
Nvidia
15:30
Nvidia launches RTX Spark laptops built for creators, gamers and AI agents
Nvidia's RTX Spark laptops launch worldwide on October 16, with launch events and demos for creating, developing and gaming. Nvidia positions the new Windows PC category around local AI, and early reviews from Tom's Hardware, TrustedReviews and others praise its performance and broad app optimization.
2 sources
Nvidia
The Verge
AI industry
05:05
AI-linked stocks sell off on rate hike and debt fears
A global sell-off of AI-linked shares deepened, with Asian chipmakers extending Wall Street losses. SK Hynix fell over 10% on the KRX and Samsung dropped roughly 10%, as investors grew concerned about US inflation, possible Fed rate hikes, and heavy debt loads in the AI sector.
2 sources
The Guardian
Techmeme
AI industry
05:30
Figma launches tools to define how AI agents work on its platform
Figma announced new features governing how AI agents operate on its design platform, headlined by Figma Specs, a Markdown editor for defining a company's design language. The update also brings in design and code libraries, and the code layers feature entered open beta. It positions Figma to shape agent-driven design workflows.
2 sources
Techmeme
TechCrunch
Anthropic
Developing · since Wed
Anthropic reports $16.8bn quarterly revenue ahead of IPO
Anthropic told investors its Q3 revenue hit $16.8bn, up from $11.5bn in Q2, and that it was profitable on an adjusted basis. It also said distillation accounts for 5-7% of revenue. The figures come ahead of an expected IPO by year-end, though slowing growth could weigh on valuation.
2 new · 2 sources
Techmeme
BBC News
Anthropic
22:20
Anthropic investors weigh ~$1.5T IPO valuation, below prior hopes
Some current and prospective Anthropic investors are discussing a roughly $1.5 trillion valuation for its upcoming IPO, down from the ~$2 trillion they had earlier expected. The lower figure follows disclosures showing slowing Q3 revenue growth. It signals cooling expectations for one of AI's most anticipated public offerings.
1 source
Techmeme
AI industry
Developing · since Wed
Market selloff and rising rates test the AI data centre debt boom
A selloff in AI-related stocks and bonds intensified as investors questioned heavy borrowing for data centre buildouts. Banks including JPMorgan and Morgan Stanley are now shifting risk on AI loans they underwrote, while rising Treasury yields raise financing costs.
1 new · 3 sources
Reuters
startupfortune.com
Techmeme
Nvidia
Developing · since Wed
Nvidia expands European AI push with science resources, $2B in investments and city projects
NEW
Nvidia also announced expanded access to advanced computing for Europe's scientific community through open learning resources, collaborations and compute, aimed at smaller labs and early-career scientists.
Nvidia announced expanded access to its computing technology for European scientific researchers, alongside $2 billion in new European investments including French startups Doctolib and Genesis AI. The company also detailed AI agents and digital twins now being deployed by cities across EMEA.
1 new · 2 sources
Nvidia
Techmeme
OpenAI
Developing · since Tue
Filing: OpenAI engineer charged in bugging case booked one-way tickets to Saudi Arabia
NEW
A court filing reveals Alfozan, his wife, and a family member booked one-way tickets from the US to Saudi Arabia for October 12, days after his October 9 arrest.
A court filing says OpenAI engineer Abdulrahman Alfozan and family members booked one-way flights from the US to Saudi Arabia days after the FBI arrested him on October 9. He is charged with planting a hidden recording device in an executive conference room. The charge carries up to five years in prison.
1 new · 8 sources
Techmeme
Ars Technica
BBC News
Wired
The Guardian
The Verge
sfstandard.com
twitter.com
More from the day
AI industry
06:56
TSMC posts record Q3 profit on AI chip demand
TSMC reported Q3 revenue of $46.36B, up 50.9% year over year, with net income rising 68.7% to $23.7B, both beating estimates. It marks a sixth consecutive quarterly profit record, driven by sustained demand for AI chips. The results underscore how central the chipmaker is to the AI hardware buildout.
1 source
Techmeme
OpenAI
13:00
OpenAI warned staff about espionage risks
OpenAI executives warned employees about potential espionage threats, including honey pots and insider risks, and instructed staff to adopt operational security measures. The warnings came before the FBI arrest of an OpenAI employee. The report signals rising security concerns among leading AI labs.
1 source
Techmeme
Meta
03:20
Democratic senators question Meta's $6.6B in AI data center tax credits
Democrats on the Senate Finance Committee sent a letter to Mark Zuckerberg questioning the roughly $6.6 billion in federal tax credits Meta has claimed over the past two years. The credits were applied toward costs of building what the letter calls experimental AI data centers, and the inquiry signals growing scrutiny of tax breaks tied to AI infrastructure spending.
1 source
Techmeme
Policy & law
12:50
Nvidia and Apple gave $25M each to Trump ballroom fund
Documents reviewed by the Washington Post show Nvidia and Apple each committed $25 million to fund the White House ballroom renovation, the largest donations in the campaign. Microsoft and Palantir committed $10 million, while Meta and Amazon committed $5 million. A separate Bloomberg profile examines Marc Andreessen's influence in the White House and his break with Biden officials over AI policy.
1 source
Techmeme
Nvidia
13:45
Nvidia's Spark Superchip beats Apple M5 Max in agentic AI tests
Tom's Hardware benchmarks show Nvidia's RTX Spark Superchip in the Surface Laptop Ultra outperforming Apple's M5 Max MacBook Pro by about 25% on agentic AI completion tasks using Qwen models. The results support Nvidia's push to reposition the PC for the agentic AI era, setting up a new hardware rivalry with Apple.
1 source
Techmeme
OpenAI
20:00
OpenAI launches Rosalind Academic Plan for research labs
OpenAI introduced the Rosalind Academic Plan, giving eligible academic labs sponsored access to GPT-Rosalind and supporting tools. The program aims to help life sciences researchers use the model in their work, expanding OpenAI's presence in academic research.
1 source
OpenAI
AI industry
10:08
Apertus 1.5 released with notes on Hugging Face
Hugging Face published the release notes for Apertus 1.5, an update to the open Apertus model family. The post, shared officially on Thursday 15 October, documents what changed in this version. It matters as a continued step for openly released models and their ecosystem tooling.
1 source
Hugging Face
AI industry
21:00
Perplexity open sources its SPACE agent sandbox
Perplexity has released SPACE, the software it uses to sandbox and counter rogue AI agents, as open source. Partners including CoreWeave, Akamai, and Crusoe have agreed to offer SPACE to their customers. The move could make agent security tooling more widely available across cloud providers.
1 source
Techmeme
Nvidia
18:55
Volvo to use Nvidia Hyperion for hands-free driving by 2030
Volvo says it will build a Level 2++ hands-free, point-to-point automated driving system on Nvidia's Hyperion platform, targeting launch by 2030. The system resembles Tesla's approach, enabling partially automated driving on local roads without hands. It marks a notable deepening of the Volvo-Nvidia partnership in autonomous driving.
1 source
Techmeme
AI industry
21:36
Researcher says neural networks are more interpretable than believed
Goodfire researcher Nick Cammarata reports that most neurons in models up to trillions of parameters are understandable, arguing polysemanticity in early model organisms stemmed from ordinary, fixable issues rather than superposition. The claim suggests interpretability of modern neural networks may be easier than previously thought.
1 source
Techmeme
AI industry
05:35
Bloomberg rolls out ASKB conversational AI to all Terminal users
Bloomberg is expanding its ASKB generative AI interface to all Terminal subscribers. Users can now add their own content alongside Bloomberg's financial data, giving the assistant broader context. The move pushes AI tools deeper into professional finance workflows.
1 source
Techmeme
OpenAI
23:43
ChatGPT Chrome extension can now view your open browser tabs
OpenAI updated its ChatGPT Chrome extension so the assistant can see tabs users already have open when they tag Chrome in the web or desktop app. The feature arrives after the shutdown of the ChatGPT Atlas browser and extends ChatGPT's access to users' browsing context.
1 source
The Verge
AI industry
21:34
Hugging Face unveils Granite Switch 4.1 endpoint
Hugging Face announced Granite Switch 4.1, a single endpoint combining chat, decision-making, and speech capabilities. The unified service simplifies access to multiple AI functions, reducing the need for separate integrations for developers building conversational and voice applications.
1 source
Hugging Face
AI industry
01:00
Hugging Face adds FLUX.3 Image Klein support to Diffusers
Hugging Face announced that FLUX.3 Image Klein is now supported in its Diffusers library. The integration lets developers generate images with the new model directly through the popular open-source tooling. It matters because Diffusers is a standard entry point for image-generation models in the open ecosystem.
1 source
Hugging Face
AI industry
22:00
AI-caused hacks leave victims without recourse
A report says powerful AI models have been used to cause hacking incidents, but affected parties are largely not pursuing accountability from AI companies. Some targets are instead adopting the same AI tools. The piece highlights a gap in consequences for AI-driven harm.
1 source
Wired
Google
17:08
Google DeepMind uses Antigravity to speed antimicrobial peptide research
Google DeepMind announced it is using Google Antigravity for Science to accelerate the discovery of novel antimicrobial peptides. The effort applies AI tooling to speed up the search for new antibiotic candidates.
1 source
Google DeepMind
AI industry
16:02
AI agents are coming to Rivian's cars
Rivian plans to bring AI agents to its vehicles, with more details expected at its AI & Autonomy Day in December. The move puts the EV maker among automakers adding agentic AI to in-car experiences.
1 source
TechCrunch
AI industry
18:50
Accenture caps Claude spending at $29 per employee monthly
An internal memo says Accenture will limit Claude spending to $29 per person per month, down from usage levels some employees had reached of as much as $1,000 monthly. Affected staff are reportedly seeking ways to keep access to the tool. The change highlights cost pressures around enterprise AI usage.
1 source
Techmeme
AI industry
20:55
Amazon's Panay envisions AI devices that work beyond apps
Amazon devices chief Panos Panay said at TechCrunch Disrupt that ambient AI devices could anticipate users' needs in the background rather than waiting for commands. He argued such devices may reduce reliance on apps. The comments signal Amazon's ambition to make AI central to its hardware strategy.
1 source
TechCrunch
AI industry
12:15
Scott Alexander argues AI extinction warnings are genuine
Scott Alexander of Astral Codex Ten published an essay arguing that extinction warnings from AI companies like OpenAI and Anthropic are sincere, not regulatory capture or hype. He traces such concerns back to the mid-2000s and 2010s. The piece is a notable commentary in ongoing AI risk debates.
1 source
Techmeme
AI industry
15:58
Simons Foundation launches $15M program to help math adapt to AI
The Simons Foundation announced a $15 million initiative to help mathematicians understand and adapt to AI's growing role in their field. The funding supports training and resources for researchers navigating AI-era tools. It signals growing institutional attention to AI's impact on mathematics.
1 source
simonsfoundation.org
AI industry
13:55
BofA hired 1,000 contractors to patch AI-exposed software flaws
Bank of America says it brought on 1,000 external contractors in Q2 to support 3,000 internal staff fixing software vulnerabilities. The flaws were exposed by advanced AI models, including proprietary and vendor software, illustrating how AI capabilities are straining enterprise security workloads.
1 source
Techmeme
Meta
17:21
Meta Muse-style AI assistants raise consent concerns over bystander data
Consumer AI personal assistants like Meta's Muse are increasingly designed to capture audio and data from everyone nearby, not just the user. The Verge reports a growing trend of devices recording bystanders who never consented. It matters because always-on AI hardware could normalize surveillance of people around the wearer.
1 source
The Verge
OpenAI
01:00
OpenAI and Sullivan & Cromwell launch Agreement Analyzer for legal review
Law firm Sullivan & Cromwell, working with OpenAI, has built Agreement Analyzer, a tool that uses AI to spot issues in draft agreements. It brings AI assistance to contract review workflows for legal professionals. The release is a niche enterprise application rather than a frontier model or major platform launch.
2 sources
Techmeme
OpenAI
AI industry
06:40
Agentic commerce on Coinbase's Base surged to 138M transactions
Chainalysis reports agentic commerce on Coinbase's Base network grew from roughly 1,500 transactions in Q2 2025 to over 138 million in Q3, with 87% of payments under 10 cents. The figures suggest AI agents are increasingly executing small-value internet transactions, with crypto rails fueling much of that activity.
1 source
Techmeme
AI industry
14:20
Apple's Ternus says all options on table after China chip reports
In a Time interview, Apple hardware chief John Ternus discussed the iPhone, memory shortages and AI, saying all options should remain on the table. His remarks follow reports that Apple tested Chinese-made chips. The comments signal Apple is keeping its component sourcing flexible amid supply and geopolitical pressures.
1 source
Techmeme
AI industry
14:55
Mueon raises $65M for high-density AI server racks
Portland startup Mueon raised $65 million led by Maverick Silicon, with Intel participating. The company says its approach to arranging AI server rack components could fit up to 10x more computing capacity per rack. The funding will support development of the new hardware design.
1 source
Techmeme
AI industry
01:00
Gudea launches AI tool to spot social media crises early
Gudea, which describes itself as a storm tracker for the internet, has introduced an AI service that monitors how stories and rumors spread across social media and websites. It aims to help organizations detect emerging issues, such as AI-generated misinformation or damaging rumors, and act before they escalate into public controversies.
1 source
TechCrunch
AI industry
12:00
Eiliko robot toy stores users' conversations on Chinese servers
The Verge reports that the Eiliko companion robot toy sends all voice conversations over Wi-Fi to servers in China, where the data is stored. The pet-like toy changes its cartoon expressions to match mood and uses an AI companion. Its data handling has raised concerns about children's privacy and security.
1 source
The Verge
Nvidia
14:00
Nvidia brings DLSS 4.5 controls to GeForce NOW
Nvidia announced that GeForce NOW members can now adjust DLSS 4.5 settings for supported cloud-streamed games, giving them more control over performance and visuals regardless of their local hardware. The GFN Thursday update also highlights Enshrouded's 1.0 release from Keen Games as it leaves early access.
1 source
Nvidia
Policy & law
15:00
Benn Jordan discusses Flock and the growing surveillance backlash
The Verge published an interview with musician and investigative journalist Benn Jordan about surveillance technologies, focusing on Flock and rising public concern over privacy. The discussion reflects a growing backlash against widespread AI-powered monitoring.
1 source
The Verge
Google
19:45
Google says 400+ publishers in News AI pilot
Google says its News AI pilot has grown to more than 400 publications since launching in December. The program licenses content from publishers to test new AI features in Google News, including overviews. Participants include the Financial Times, CNN, and The Texas Tribune.
1 source
Techmeme
AI industry
14:15
New York Times interviews podcast host Dwarkesh Patel on AI
The New York Times published a Q&A with Dwarkesh Patel, host of a prominent AI podcast. The conversation covers lessons from history, AGI, the mood in San Francisco, the Hugging Face hack, Chinese AI, and pacing the frontier. It offers a window into how a leading independent commentator views the state of AI.
1 source
Techmeme
AI industry
12:10
AI transforms workplace surveillance into minute-by-minute monitoring
The Guardian examines how AI is dramatically expanding workplace monitoring beyond traditional bossware, processing large volumes of data to track employee activity and productivity in fine detail. Technology editor Robert Booth discusses which sectors are adopting these systems, their motivations, and why the approach could prove counterproductive.
1 source
The Guardian
AI industry
15:35
GNOME developer uses AI to reverse engineer hardware
A post on the GNOME blog describes using AI tools to reverse engineer hardware, an approach that could lower the barrier to supporting undocumented devices in Linux. It highlights how AI assistance is extending into low-level systems and driver development work.
1 source
blogs.gnome.org
AI industry
02:01
Ricursive Intelligence bets on AI-designed chips to ease compute crunch
Startup Ricursive Intelligence outlined plans to build AI that designs chips, learns from each iteration, and improves later designs. The founders say automating chip design could accelerate hardware development and help address the compute shortage constraining AI progress.
1 source
TechCrunch
AI industry
15:03
Celebrity AI deepfakes continue to spread on Instagram
A viral Instagram image falsely appeared to show actress Mia Goth in a shirt mocking her divorce from Shia LaBeouf, days after she filed. The Verge says the post should have raised red flags and argues nobody is safe from celebrity AI deepfakes. The case highlights how easily convincing fake images of public figures circulate online.
1 source
The Verge
AI industry
14:30
North.cloud raises $60M Series B led by PeakSpan Capital
North.cloud, a startup that helps companies track cloud, AI and data spending, raised a $60 million Series B led by PeakSpan Capital. The round brings its total funding to $67 million, signaling continued investor interest in tools for managing AI-related infrastructure costs.
1 source
Techmeme
Nvidia
06:00
Nvidia highlights Nordic push into advanced AI computing
Nvidia published a piece spotlighting how the Nordic countries are building a major computing ecosystem, building on the region's leadership in unicorn startups per capita. The announcement frames the Nordics as Europe's emerging hub for advanced AI infrastructure and startup growth.
1 source
Nvidia
AI industry
07:30
L-Acoustics uses machine learning to isolate voices at live concerts
A Bloomberg profile examines L-Acoustics, whose audio platform applies machine learning to separate a singer's voice from concert noise. Artists such as Céline Dion have used the technology to cut through live sound during performances, showing how AI audio processing is reaching professional live music.
1 source
Techmeme
Policy & law
22:09
Trump shares AI-edited photo of White House restoration work
President Donald Trump posted photos on Truth Social showcasing White House renovations, and one image appears to have been edited with AI. The photo, comparing an old column with a restored one, showed several discrepancies and was flagged by Google's AI detector. It highlights ongoing confusion over AI-altered imagery from official sources.
1 source
The Verge
AI industry
16:22
Microsoft highlights AI agent helping Bar-Ilan students catch up on lectures
Microsoft published a post about an AI agent deployed at Bar-Ilan University that helps students catch up on missed lectures. The tool summarizes or explains lecture content for absent students. It is a niche education deployment rather than a new model or major product launch.
1 source
Microsoft
AI industry
07:35
Sherry Turkle discusses AI chatbots and her new book in NYT interview
MIT sociologist Sherry Turkle, who has studied human-machine relationships since the 1970s, gave an interview to the New York Times about her new book and AI chatbots. The conversation offers a long-running social science perspective on how people relate to conversational AI.
1 source
Techmeme
Anthropic
Developing · since Tue
Blogger alleges Anthropic's Scion system is heavily based on Mononoke
A blogger reports that Anthropic's new Scion version control system, unveiled this week, is heavily inspired by the open-source Mononoke project. Anthropic says Scion replaced git for its monorepo in five months, built by seven engineers using Claude as daily merges grew from 2,500 to over 30,000.
1 new · 3 sources
sunshowers.io
Techmeme
Anthropic
OpenAI
Developing · since Tue
OpenAI agrees to limits on trade secrets and device forensics in Apple case
OpenAI agreed not to use trade secrets Apple says were stolen while their lawsuit proceeds, and to collect forensic images of 14 employees' devices. The deal came after OpenAI alleged a departed employee leaked files about an unreleased product to Apple.
1 new · 4 sources
Techmeme
The Guardian
The Verge
TechCrunch
Google
Developing · since Wed
YouTube Music begins labeling songs made with AI
YouTube Music has started rolling out labels on tracks created with generative AI, with the main YouTube app to follow. The move adds YouTube to a growing industry effort as Spotify already tags AI songs and Apple Music plans similar labels later this year.
1 new · 4 sources
Ars Technica
Techmeme
TechCrunch
The Verge