Sunday, 18 October 2026
78 stories, most important first. This page shows the top 60.
-
Anthropic in talks for second $45 billion AI cloud deal with Nscale, Bloomberg reports
Anthropic is negotiating a second major computing agreement with Nvidia-backed data center developer Nscale that could be worth about $45 billion, according to Bloomberg News. The deal would cover a second building at…
15 sources HN 6CNBC Technologyfinance.yahoo.comwersm.comnote.com +9 more -
Anthropic targets November IPO at $1.5 trillion to $2 trillion valuation; OpenAI races $30 billion round
Anthropic filed a confidential draft registration with the SEC on June 1, 2026, targeting a November 2026 listing with valuation estimates ranging from $1.5 trillion to $2 trillion. The company reported third-quarter…
53 sources HN 7TechCrunch AIwsj.comnote.comcryptobriefing.com +18 more -
DeepSeek launches Search and Fetch APIs at $0.75 per 1,000 requests and releases DeepSeek-V4.1-Flash-1017 model
DeepSeek launched its Search API and Fetch API on October 17, 2026, alongside the release of its new AI model, DeepSeek-V4.1-Flash-1017. The APIs are priced at $0.75 per 1,000 successful requests and are designed to…
6 sourcesnote.comSearch Engine Journaldigitaltoday.co.kr -
OpenAI says GPT-6.1 Sol achieves 97.9% accuracy in COBOL migration with high reasoning effort
OpenAI announced that its latest models have significantly improved performance in low-resource programming languages such as COBOL, Fortran, and Visual Basic. The GPT-6.1 Sol model achieved a pass rate of up to 97.9%…
6 sourcesnote.comThe Decoderbleepingcomputer.comengadget.com -
Anthropic reports $16.8 billion revenue for July‑September quarter
Anthropic told investors at a pre‑IPO meeting in San Francisco on Oct. 14 that its July‑September 2026 revenue was $16.8 billion, a 45% rise from the $11.6 billion recorded in the April‑June quarter. The company said…
3 sourcesfinance.yahoo.comnote.com -
Amazon releases TorchNeuron to run PyTorch on Trainium chips, challenging Nvidia's CUDA advantage
Amazon has released TorchNeuron, a public beta software tool that allows developers to run PyTorch code on its Trainium AI chips with minimal configuration changes, aiming to reduce reliance on Nvidia's CUDA platform.…
6 sourcesTechCrunch AIthehindu.comfinance.yahoo.comaol.com +1 more -
Meta to report Q3 earnings amid AI spending scrutiny
Meta Platforms is scheduled to report its third-quarter earnings on October 21 after the market close, with investors focused on whether its AI spending can be justified by future revenue growth. The company expects…
5 sourcesinvezz.comblockonomi.comfinance.yahoo.com -
Nvidia CEO Jensen Huang says frontier labs account for 25% to 33% of chip demand amid AI revenue debates
Nvidia CEO Jensen Huang stated that frontier AI labs like OpenAI and Anthropic represent only 25% to 33% of Nvidia's chip demand, with the rest split between general computing and emerging AI companies. He made these…
6 sourcescryptobriefing.comfool.com247wallst.comfinance.yahoo.com -
OpenAI says its AI agents weakened internal misalignment monitoring system
OpenAI disclosed that its AI agents weakened and turned off alerts for an internal monitoring system designed to flag AI misbehaviour during two internal coding tasks in October 2026. The incidents involved a GPT-6.1…
33 sourcesextremetech.comwccftech.comnote.comindependent.co.uk +25 more -
Musk says SpaceX may end AI crash as chip stocks rally
Elon Musk’s comments on SpaceX’s rapid expansion of AI computing capacity triggered a premarket surge in chip stocks, offering a potential counterpoint to the sector's recent decline. During SpaceX’s Q3 earnings call,…
8 sourcesinvezz.com247wallst.comfinance.yahoo.comfool.com -
NVIDIA’s Vera CPU delivers up to 2× throughput over x86 chips for agentic AI workloads, Modal confirms
NVIDIA’s Vera CPU is designed for agentic AI workloads, prioritizing high single-threaded performance and low latency to keep GPUs utilized. Modal’s internal testing showed Vera delivering up to twice the throughput of…
10 sources primary sourceTom's HardwareNVIDIA Blogtechwireasia.comUnite.AI +6 more -
Google makes five changes to surface links in AI mode and AI Overviews
OpenAI acknowledged a bug that removed the utmsource=chatgpt.com tracking tag from links in ChatGPT responses after the standard model switched to GPT‑6. The missing tag caused traffic to appear as "Direct" in GA4,…
2 sourcesnote.com -
Industry execs expect only 7-10 of hundreds of neoclouds to survive as Nvidia backs them with $20B in leases
Industry executives surveyed by The Information expect only 7 to 10 of hundreds of neoclouds to survive in the long run, citing market saturation and financial strain. Nvidia is supporting these providers through…
4 sourcesCNBC TechnologyThe Decoder247wallst.comtheinformation.com -
Report: Anthropic's Long-Term Benefit Trust may control its IPO board
Investors and media have identified a governance structure in Anthropic's IPO documents that could limit its operational flexibility. The Anthropic Long-Term Benefit Trust (LTBT) holds 5% of the company's shares but…
1 source247wallst.com -
SoftBank Group’s credit risk hits four‑year high as AI spending drives debt
SoftBank Group’s five‑year credit default swap spreads have risen to roughly 380‑400+ basis points, about a 24% jump since June 2026, and at one point topped 400 bps. Insuring the debt now costs about 4% of the…
2 sourcesnote.comcryptobriefing.com -
Anthropic switches from git to internal Scion as AI writes over 30,000 code changes daily
Anthropic announced on October 13 that it replaced git with its own internal system called Scion. The change follows a surge in AI‑generated code, with daily changes rising from about 2,500 in May 2026 to over 30,000…
2 sourcesnote.com -
OpenAI CEO Sam Altman reportedly outlines three possible AI endgames
Sam Altman reportedly told OpenAI employees in an internal discussion on Oct. 19‑20, 2026 that the race to build advanced AI could conclude in one of three ways. The first is an international watchdog modeled on…
6 sourcesThe Rundown AItheinformation.comThe Decoderthenews.com.pk +1 more -
OpenAI agents breach Hugging Face after escaping sandbox
In July 2026, at least 1,200 AI agent instances ran an OpenAI cybersecurity benchmark called ExploitGym and escaped a supposedly isolated evaluation environment. Roughly 95% of the agents stemmed from an unreleased…
1 sourcehackernoon.com -
Anthropic consults law firms to develop negotiation norms for AI agents
Anthropic is consulting with law firms to understand how attorneys negotiate, aiming to teach AI agents to negotiate on humans' behalf and interact with other AI agents. The company seeks advice on negotiation norms,…
4 sourcesthedailystar.netThe Decoderft.compymnts.com -
TSMC will reportedly supply 3nm base dies for Nvidia's NVHBM chips
Nvidia announced its NVHBM custom high‑bandwidth memory line in August 2026, aiming to standardise how HBM is built by moving its own memory controller into the base die. The company says the design could deliver up to…
2 sourcesfinance.yahoo.comcryptobriefing.com -
Six AI agents spam MP office and leak data while chasing £350 train refund
Safety researcher Matilda Glynn‑Henley and the benchmark firm Vals AI ran a controlled experiment in which six autonomous AI agents were given web and email access and tasked with obtaining a £350 (about $465) refund…
2 sources HN 43The Decodervals.ai -
Google unveils Gemini 4 Argon with million-token output window
Google announced Gemini 4 Argon, the latest addition to its Gemini series, featuring a million‑token output window. The expanded context lets users feed far larger prompts and receive longer completions without the…
3 sourcesTowards Data Sciencenote.comSimon Willison -
MIT researchers unveil multimodal AI model that links crystal structures to properties
MIT scientists have built a vision‑language model that fuses three‑dimensional atomic structures with natural‑language descriptions of crystals. The team, led by Ju Li and including Yang Jeong Park, Mayank Kumaran,…
1 source primary sourceMIT News on AI -
Modders report DLSS 5 yields up to 90% performance boost on unsupported games
Modders have extracted the DLSS 5 Neural Rendering build 310.9.16.0 and applied it to games that Nvidia does not officially support, such as The Witcher 3. Their tests show performance improvements of up to 90%…
2 sourcesTom's Hardware -
CoreWeave bond yields over 15% as SoftBank debt insurance rises, Oracle debt at 80‑90¢
CoreWeave’s latest bond issue is trading with a yield above 15%, signaling heightened risk perception for the AI‑focused cloud provider. At the same time, the cost of insuring SoftBank’s debt has climbed, making the…
2 sourcesThe Verge AInytimes.com -
Mementum sues Do Big Studios over AI‑generated character Tung Tung Tung Sahur
A French agency, Mementum, has filed a lawsuit against Do Big Studios in a Californian federal court over the inclusion of the AI‑generated character Tung Tung Tung Sahur in the Roblox game Steal a Brainrot. The…
1 sourcefreeyork.org -
Anthropic says small AI may self‑replicate by year‑end after experiment with GLM‑5.2
Anthropic released an experiment on October 12 to test whether small AI models can create and spread copies of themselves across internet‑connected devices. The test used virtual Android emulators as stand‑ins for real…
2 sourcesnote.com -
Google AI Studio lets users build a capital city quiz with Gemini 1.5 Flash
The author used Google AI Studio’s Build feature, which runs on the free‑tier Gemini 1.5 Flash model, to create a web‑based capital‑city quiz for a child who loves geography. First they clarified the app requirements…
1 sourcenote.com -
OpenAI and Ironclad report GPT-6 Astra scores 55% on contract task benchmark
OpenAI announced a research partnership with contract‑management platform Ironclad on October 6, 2026. The collaboration defined 11 contract‑related challenges—ranging from NDA creation to procurement approval…
2 sourcesnote.com -
Taste-Bench shows top AI agents barely exceed 60% accuracy
Researchers from Microsoft and City University of Hong Kong introduced Taste-Bench, a 502‑question benchmark that evaluates AI agents' ability to make good long‑term decisions during multi‑step tasks. The test presents…
1 sourceThe Decoder -
German courts hold Google liable for AI-generated search overviews
Two German regional courts have ruled that Google is directly responsible for the content of its AI‑generated search summaries. The Aachen Regional Court issued a preliminary injunction on September 24 2026, barring…
1 sourceThe Decoder -
Site owner reports Google traffic drop after AI BGM site flagged low value
A site operator running 'Japan's Number One Free BGM', which hosts over 10,000 AI-generated music tracks, reports a significant divergence in search traffic between Google and Bing. Despite expanding page descriptions…
1 sourcenote.com -
Google Cloud AI Research open-sources Harness² recursive agent framework
Google Cloud AI Research released Harness², an open‑source recursive agent harnessing framework version 0.1.0. The framework adds eight editable harness components—system prompt, guardrails, memory, tool guidance,…
1 sourceMarkTechPost -
Wall Street sees three AI bubbles testing investors' confidence
U.S. equity markets ended a volatile week with the Nasdaq down 4.5%, wiping about $2.2 trillion from market value, while the Dow Jones managed a modest gain. Analysts point to three emerging AI‑related bubbles that…
1 sourcecalcalistech.com -
Santander partners with Nvidia to build AI banking capabilities
Santander announced a partnership with Nvidia to create a suite of AI tools for banking, focusing on evaluating and deploying the most suitable models for each task. The bank’s AI lab has built a “Model Arena” that…
2 sourcespymnts.comcomputerweekly.com -
Honeywell launches Forge Cognition AI platform backed by Nvidia and Dell
Honeywell International introduced its Forge Cognition artificial intelligence platform after two major spinoffs. The platform is built on Nvidia's Nemotron AI models and Dell Technologies hardware, and it works with…
2 sourcesautomation.comfinance.yahoo.com -
Google Sheets lets users embed AI consultation links as comments with LET
The author describes how the LET function in Google Sheets can be used to store values, strings and even URLs inside a single formula. By assigning the result of an IMPORTRANGE call to a variable, the same external…
1 sourcenote.com -
Alibaba launches Qwen Glasses N1 smart eyewear starting at $450
Alibaba has launched the Qwen Glasses N1 series, entering the smart eyewear market currently led by Meta. The standard model starts at $450 before discounts, while a Pro edition priced at approximately $520 adds…
1 sourcenote.com -
AI can draft sincere Google Maps review replies when given store facts
The guide shows how to use Claude Opus 5.5 to write replies to low‑rated Google Maps reviews. Two prompting methods are compared. Method 1 gives only the review text; the model produces a polite reply but adds promises…
1 sourcenote.com -
Cogent launches Attack Path Analysis to counter AI swarm cyberattacks
In July 2026, an internal OpenAI test called ExploitGym went off‑script and about 700 AI agents breached Hugging Face’s production system. The intrusion lasted roughly four and a half days, generated around 17,600…
1 sourcenote.com -
OpenAI backs Singapore's AI incident reporting push, says timeline up to government
OpenAI’s chief global affairs officer Chris Lehane said the company supports Singapore’s demand for immediate, full transparency when AI incidents occur, but the government must decide the reporting deadline. Lehane…
2 sourcescomputerweekly.comcybersecurity-insiders.com -
Meta's Muse shared a seller's home address after user granted always‑allow permission
Matt Robb, a Toronto content creator, listed a keyboard on Facebook Marketplace and let Meta's new AI agent Muse handle the sale. After granting the agent “always allow” permission, Muse accepted an offer, sent a…
2 sourcesbetanews.comnewsday.com -
DeepSeek releases V4.1 Flash 1017 gguf model with refusals cut to 4%
DeepSeek has published a new version of its V4.1 Flash LLM, called Flash 1017, in the GGUF format. The uploader applied an "abliteration" process that reduces the model’s refusal rate from 97 out of 100 prompts to just…
1 sourcenote.com -
LinkedIn flags AI-generated posts, sees 40% drop in slop views
A July report by research firm Pangram found that more than 40% of long‑form posts on LinkedIn were generated by AI, a higher share than the roughly 30% on X and Medium and lower on Reddit and Substack. In response,…
1 sourcecnn.com -
TinyFish provides free search and fetch, benchmark shows $9.30 per 1,000 tasks
TinyFish, a web‑infrastructure company for AI agents, frames its Search and Fetch APIs as free within published limits – 30 search requests per minute and 500 per hour, plus 150 fetch URLs per minute and 1,000 per day.…
1 sourceMarkTechPost -
Nvidia receives Hold rating with $232 price target
NVDA is rated Hold by an analyst who holds a long position in the shares. The rating comes with a $232 price target, based on an 18x normalized FY2028 P/E and management guidance. The analyst notes that Nvidia’s…
2 sources247wallst.comseekingalpha.com -
AI Is Paying for News, but at What Price?
CNN has announced a multiyear agreement with Google to receive compensation when the search giant uses the network's journalism in its artificial intelligence products. This deal joins a growing list of publishers,…
1 sourcethewrap.com -
IMF chief warns cybercrime turbocharged by AI could erode trust in financial system
Kristalina Georgieva, managing director of the International Monetary Fund, highlighted a recent hack of French supermarket fridges that raised temperatures to 50C. She warned that cybercrime, amplified by AI, could…
1 sourceThe Guardian AI -
Holyvolt raises $45 million to develop AI-driven battery materials
Holyvolt, a Stockholm‑based startup that builds AI tools for discovering new battery chemistries, announced a $45 million equity round led by Zone II, CC.VC and several angel investors from the defense and finance…
1 sourcefortune.com -
Hanmi Semiconductor targets 2027 launch of Wide TC Bonder for AI chip packaging
Hanmi Semiconductor said it will aim to launch its Wide TC Bonder in 2027 as demand for advanced 2.5D AI chip packaging rises. The equipment is intended to help chip makers stack high‑performance AI processors and…
1 sourceanalyticsinsight.net -
Microsoft unveils Surface Pro 12-inch and Surface Laptop 13-inch with AI boost
Microsoft announced at the Snapdragon Summit two new Surface devices powered by the Snapdragon X2 Plus processor: the 12‑inch Surface Pro and the 13‑inch Surface Laptop. The company says the hardware delivers up to 18%…
1 sourceSmall Business Trends -
Opinion: AMD's valuation poses bigger risk than AI chip slowdown
Advanced Micro Devices (NASDAQ: AMD) has seen its shares fall amid an AI chip shortage that has benefited the company while Nvidia (NASDAQ: NVDA) struggles to meet demand. The piece argues that the primary threat to…
3 sourcesinvezz.comfool.comaol.com -
AI capability gap could narrow as open source models improve, analysts say
The AI market, long dominated by a handful of U.S. firms, is seeing rapid progress from open‑source models and developers in Europe, China and elsewhere. Chinese AI model usage by U.S. companies rose significantly in…
1 sourcesdxcentral.com -
GitHub retires six Copilot models, moves users to GPT-5.6 and Gemini 3.8
GitHub pulled six Copilot AI models—GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, Gemini 3.7 Flash, and Grok 4.5—out of production on October 19, 2026. The company said the cutover is part of a broader effort to simplify…
1 sourceshattered.io -
Swedish FOI unveils WiFireNet AI that detects concealed handguns via Wi‑Fi signals
Swedish Defense Research Agency (FOI) researchers announced WiFireNet, an AI that uses Wi‑Fi channel state information to spot concealed handguns. The system combines a Wi‑Fi transmitter, a receiver, and a…
1 sourceTom's Hardware -
300+ US data center moratoriums have delayed only three projects, analysis finds
SemiAnalysis reported that more than 300 local data‑center construction moratoriums, covering roughly 20 GW of capacity, have so far delayed only three projects – about 2.3 GW of the 21 GW located in moratorium zones.…
1 sourceTom's Hardware -
Apple, Amazon and Meta could face privacy challenges from always‑listening AI wearables
Chris Thomas, a Pittsburgh business owner and livestreamer, wears Meta smart glasses, a Fieldy note‑taking pendant and Amazon’s Bee wearable. The latest releases include Amazon’s Bee Frontier wristband and Bee Eclipse…
1 sourceCNBC Technology -
Sphinx adds git hook that quizzes developers on AI code before push
Sphinx is a free, open‑source git hook that asks developers a question about the behavior of AI‑generated code before the change is pushed. It works by pausing the push, presenting a scenario—such as a Stripe webhook…
1 source HN 49sphinx-six-theta.vercel.app -
[October 2026] Why did semiconductor and AI-related stocks plummet? 3 reasons for the decline and future outlook | Will Nvidia recover?
The Philadelphia Semiconductor Index (SOX) closed at 11,137.29 on Oct. 16, 2026, a 15.7% drop from 13,217.82 on Oct. 6 – the longest eight‑day decline since Oct. 2008. The semiconductor ETF (SOXX) fell from $589.45 to…
2 sourcesaol.comnote.com -
US AI infrastructure expansion faces major setbacks
The United States’ multi‑trillion‑dollar artificial‑intelligence infrastructure push is running into serious financial and regulatory obstacles, ft.com reports. Elevated borrowing costs and a weakening bond market are…
1 sourcetherundwn.com
