Nvidia announces native Rust support for GPU programming, a consequential shift for CUDA developers.
OpenAI expands ChatGPT advertising with Sponsored Agents, introducing ads into agent interactions.
US and Chinese experts propose nuclear-style AI safeguards, including a dedicated military AI hotline.
Reuters : US and Chinese experts propose nuclear-style AI safeguards, including a dedicated AI military hotline, as part of a dialogue ahead of planned bilateral talks — An AI system that interferes with a nuclear command network or launches a military cyber operation could leave Washington or Beijing …
Huawei details 2027 Ascend AI chips and UnifiedBus interconnect plans for next-generation AI systems.
Reuters : Huawei plans to launch the Ascend 960DT AI chip in Q1 2027 and the 960PR in Q3 2027, and says its UnifiedBus tech is key to its next-generation large AI systems — Huawei Technologies (HWT.UL) will launch two new artificial-intelligence semiconductors in 2027, rotating chairman David Wang said on Thursday …
Crusoe raises $3.9B at a $30.9B valuation to build factory-made AI data centers.
Anissa Gardizy / Wall Street Journal : Data center developer Crusoe raised $3.9B co-led by Atreides, Valor, and Mubadala at a ~$30.9B post-money valuation, as it bets on factory-built data centers — Crusoe raised $3.9 billion at a valuation of nearly $31 billion, as it bets on factory-built data centers
Google, Nvidia, and Emerald AI form an alliance for grid-responsive AI data-center power use.
Amy Harder / Axios : Google, Nvidia, and Emerald AI launch the AI Energy Management Alliance to advance data centers that dynamically adjust electricity use based on grid conditions — A new coalition launched Wednesday is seeking a triple win for a data center industry that lately has struggled to notch any wins at all.
Xiaomi publishes a live post-training dashboard for Mimo 2.6, exposing its model optimization process.
Paper reports ternary LLM results beyond the 1.58-bit barrier, advancing ultra-low-bit model design.
HarnessTax measures how much coding-agent results depend on the evaluation harness rather than the model.
A 4B model is trained to generate query plans claimed to run 81% faster than PostgreSQL's.
Financial Times examines how AI-assisted target generation is accelerating warfare and amplifying error risks.
Financial Times : A look at the rapid integration of AI into warfare, as the greater speed and scale of AI-assisted target generation processes increase the risk of errors — Autonomous systems are being rapidly integrated on the battlefield, as models advance in ways even their creators cannot fully predict.
Researchers document dating scams that used Claude and other LLMs to generate replies at scale.
Yael Grauer / The Verge : Anthropic and other researchers detail how thousands of people were catfished by dating scam apps using LLM-generated replies from Claude and other models — Security researcher Matthew “Zigula” Gore-Kormanik was analyzing a fraudulent dating app called Dora when he got a pop-up message saying he was receiving a call from Jennifer.
Former DeepMind researchers' new startup Emulate is reportedly discussing a $700M raise at $3.7B.
Financial Times : Sources: Emulate, a month-old UK AI startup founded by former Google DeepMind researchers, is in advanced talks to raise as much as $700M at a $3.7B valuation — Researchers leaving Big Tech group's London lab close to raising $700mn for new start-up Emulate
Manus reportedly nears a $500M round at $4B after Beijing unwound Meta's proposed acquisition.
Bloomberg : Sources: Manus is set to soon close a $500M funding round at a $4B valuation, signaling growing confidence after Beijing unwound its $2B buyout by Meta — Chinese-founded AI startup Manus is set to double its valuation to $4 billion in its first fundraising since Beijing ordered it to split …
BCI companies have raised over $1B in 2026 alone, exceeding the prior four years combined.
Financial Times : PitchBook: companies developing brain-computer interfaces have already raised more than $1B in 2026, compared with $1.56B in the previous four years combined — Start-ups developing thought-controlled technology are drawing venture capital funding from Silicon Valley to Beijing.
PlayStation's Linux lead quits, blaming LLM-assisted work that contributors cannot understand.
Enclave reports that DeepSeek V4.1 Flash is now its strongest model for hacking tasks.
Paper benchmarks frontier models' physics capabilities, directly probing scientific-reasoning limits.
Dream-RSI proposes recursive self-improvement through agents evolving their own training worlds.
Paper develops accurate models of AMD matrix cores, useful for AI-kernel performance engineering.
A reverse-engineered Jev-like model offers an inspectable implementation of a browser agent.
Jev Ultrafast introduces a browser agent with a dynamic indexed action space.
OpenSpec offers a lightweight, configurable specification framework for AI-assisted development.
Amazon Science explains developing provably correct Rust programs with the Verus verifier.
Microsoft details performance improvements landing in .NET 11.
Google's vectorized, performance-portable Quicksort is a practical study in cross-platform optimization.
Researcher recounts recovering signing keys used in US driver's-license barcode systems.
WatchTowr analyzes a 32-year-old Telnet-server bug and its modern security implications.
Wired reports hackers accessed a Flock camera, highlighting risks in networked surveillance infrastructure.
WSJ reports AWS cannot restore some data from Middle East facilities struck by Iran.
Wall Street Journal : A profile of Jacob Coxon, the British researcher who quit Anthropic after a series of security escalations and says the explosive response surprised him — Long-simmering worries about the technology's potential dangers have exploded into global consciousness since Jacob Coxon's dire warning
Andrew Cunningham / Ars Technica : macOS 27 Golden Gate review: addresses many Tahoe-era Liquid Glass issues and has useful new AI features, but no Intel support and uses a lot of storage space — It's both a “Snow Leopard update” and a major leap for Apple Intelligence. — Apple Intelligence is back, baby!
Alice Lassman / Bloomberg : How beauty companies like Qoves are commercializing facial-analysis algorithms that evaluate geometric proportions, balding, and more to prescribe treatments — AI beauty tools can measure your jawline, grade your “harmony” and prescribe treatments for improving your face. Does any of it work?
Boone Ashworth / Wired : Snap is pitching its $2,195 Specs as an enterprise product, inking Salesforce, Amazon, and Nvidia deals to enable visual overlays, alongside consumer features — Snap thinks its chunky, pricey smart glasses are the future of human computing, powered by a new AI app and features that do everything …
Sarah Perez / TechCrunch : Docs: Automattic interim CEO Mark Davies and CLO Andy Missan signed each other's severance packages worth $8.15M during Matt Mullenweg's brief ouster as CEO — Last week, on September 9, Automattic's board voted to put CEO Matt Mullenweg on paid leave, a decision the board still hasn't explained publicly.
Bloomberg : Sam Altman and Jensen Huang are among business leaders attending a White House state dinner for Xi Jinping next week; a source says Tim Cook will also attend — Apple Inc. Executive Chairman Tim Cook and OpenAI Chief Executive Officer Sam Altman are poised to attend a White House state dinner …