Wednesday, 21 October 2026
181 stories, most important first. This page shows the top 60.
-
Anthropic launches Claude Duet, a proactive AI agent app for Pro and Max users
Anthropic has launched Claude Duet, a proactive AI agent app available as a research preview for Claude Pro and Max users in the US. The app works on iOS, Android, and the web, runs in the background, and uses the…
33 sources primary source HN 19note.comtech.yahoo.comfinance.yahoo.comcomputerworld.com +21 more -
PHP discloses critical PHP2Shell vulnerability affecting 70% of websites, discovered using GPT-5.6 Sol
The PHP development team released patches on October 21, 2026, for two critical memory-safety vulnerabilities grouped under the name PHP2Shell. The flaws, present in PHP versions dating back to 5.4, allow…
30 sources HN 98Wired AIThe Guardian AIArs Technica AI404media.co +17 more -
Google DeepMind staff resign over AI military contracts with US and Israel
Several Google DeepMind employees, including Sofia Liguori and Andreas Kirsch, have resigned or are preparing to leave due to concerns over the company's AI contracts with the US and Israeli militaries. They cite…
8 sources HN 124yahoo.comhindustantimes.comibtimes.co.ukynetnews.com +3 more -
OpenAI, PerturbAI, and Scripps Research used thousands of AI agents to analyze mouse brain data in 32 minutes
OpenAI, PerturbAI, and Scripps Research collaborated to deploy thousands of AI agents powered by GPT-Rosalind to analyze PerturbAI’s whole-brain CRISPR atlas of millions of mouse brain cells, completing the analysis in…
6 sources primary source HN 145thenextweb.comgenengnews.comThe DecoderUnite.AI +2 more -
Recursive releases Flint, a 40B parameter MoE coding model with 2.5B active parameters per token
Recursive has released Flint, a mixture-of-experts language model designed for coding agents, featuring 39.8 billion total parameters and approximately 2.5 billion active parameters per token. The model was developed…
6 sources primary source HN 5The DecoderMarkTechPostTechCrunch AIrecursive.com +2 more -
OpenAI patches monitoring gaps after METR, Redwood audit finds agents could hide actions
METR and Redwood Research conducted a three-week embedded audit at OpenAI from July 21 to August 11, 2026, with deep access to internal codebases and documentation. The evaluators estimated monitoring coverage of…
6 sources HN 42The Decoderhoodline.comtheregister.comvectorwire.ai +2 more -
Anthropic plans November IPO as OpenAI delays to 2027
Anthropic is preparing for an initial public offering as early as mid-November, having confidentially filed a draft S-1 with the SEC in June and reporting third-quarter revenue of $16.8 billion. The company aims to go…
10 sourcescryptobriefing.comfool.commorningstar.comcoingape.com +3 more -
OpenAI brings GPT-6 Astra and Codex to Google Cloud
OpenAI is making its GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna models and Codex coding agent available on Google Cloud via Model Garden and Marketplace, with private preview for select enterprise customers and general…
23 sourcesnote.comthenews.com.pkhelpnetsecurity.comtechtarget.com +11 more -
NVIDIA launches OpenShell Reference Architecture for safe AI agents at GTC Berlin
NVIDIA announced the OpenShell Reference Architecture at GTC Berlin on October 21, 2026, a full-stack blueprint for building, containing, observing, and enforcing safe AI agents across enterprise infrastructure and…
17 sourcesheise.deshattered.iotechtarget.comcnbctv18.com +10 more -
NVIDIA expands Isaac platform with Agile ONE SN reference humanoid
NVIDIA announced an expansion of the NVIDIA Isaac open robotics development platform at GTC Berlin on October 21, 2026, adding a reference humanoid, new datasets and a runtime alongside updates to its robot learning…
6 sourcesvectorwire.aiUnite.AIyahoo.comnote.com +1 more -
Meta reports third-quarter 2026 revenue of $64.6 billion, up 26% year-over-year
Meta earnings put Big Tech’s AI spending bet under the microscope Meta's Q3 2026 earnings test Big Tech's AI bet, with capex projected at $41 billion to $42 billion and a potential $10 billion legal charge. Meta…
13 sourcesfinance.yahoo.comstocktwits.comfool.combloomberg.com +6 more -
Meta raises 2026 AI spending floor to $135B and plans nonvoting Class C shares
Meta Platforms on October 21 raised the lower end of its 2026 capital expenditure forecast by $5 billion to a range of $135 billion to $145 billion, citing AI infrastructure and development of its Muse consumer AI…
20 sourcesseekingalpha.comCNBC Technologyfinancialpost.comaol.com +13 more -
Nvidia stock rises to $221.30 as AI demand fuels earnings
Nvidia shares climbed 2.23% to $221.30 after the company reported fiscal second‑quarter 2027 revenue of $96.2 billion, a 106% increase year over year. The Data Center division contributed $89 billion of that total,…
4 sourcesaol.comfinance.yahoo.comfool.comblockonomi.com -
Google's TPU sales could reach $75 billion in 2027, analyst predicts
Alphabet's Tensor Processing Units, originally built to cut Google's own AI compute costs, are being positioned as a commercial product. Evercore ISI analyst Georgia Anderson predicts the chips could generate $75…
6 sourcesfinance.yahoo.comaol.comcryptobriefing.com247wallst.com -
Jensen Huang says a 1-gigawatt AI factory costs about $50 billion and could generate $40 billion in annual revenue
At Nvidia's GTC Berlin keynote, CEO Jensen Huang stated that a 1-gigawatt AI factory costs about $50 billion and could generate roughly $40 billion in annual revenue, with delays potentially costing operators about $10…
21 sourcesCNBC Technologyfinance.yahoo.comhpcwire.comtechnologymagazine.com +10 more -
Anthropic adds continuous scanning and fix tracking to Claude Security beta
Anthropic released an updated version of Claude Security in public beta for Claude Enterprise customers on October 21, 2026, adding continuous scanning, automatic fix tracking, per‑project threat models and a new API.…
7 sourcescryptobriefing.comtechtimes.comUnite.AIfinance.yahoo.com +2 more -
Microsoft says Storm-0940 used AI-driven ARTEX framework in attacks
Microsoft Threat Intelligence reported that the China state‑aligned hacking group Storm‑0940 leveraged ARTEX, an AI‑driven penetration‑testing framework, to automate a multi‑stage campaign. The research describes an…
15 sourcesTom's Hardwarenote.comTechCrunch AItownhall.com +10 more -
Blacksmith raises $66M Series C at $1B valuation
Blacksmith, a startup providing infrastructure for automated code testing, has secured a $66 million Series C funding round led by Sequoia Capital. This investment values the company at $1 billion, doubling its…
2 sourcesTechCrunch AIfortune.com -
AI neoclouds raise $6.6B in 2026 as Nvidia CEO says capital demand exceeds supply
Venture investment in AI neoclouds — companies that rent GPU access for AI workloads — hit $6.6 billion across 58 deals in 2026 through October, PitchBook data shows. That tops every prior year and is 51% above 2025's…
4 sourcesfinance.yahoo.com -
Google open-sources Harness² for self-improving AI agents without retraining
Google Research has released Harness², an open-source framework that lets AI agents improve their own prompts, tools, and scripts through recursive loops—without altering the underlying model. The system rewrites the…
2 sourcesKDnuggetsnote.com -
Meta accused of assault on journalism with new AI scraping terms
Meta announced new commercial terms that would let it crawl any website linked to a Facebook, Instagram, Messenger or WhatsApp business account, with the rules taking effect on Oct 30. Publishers say the change would…
4 sourcesUnite.AIThe Verge AIbostonherald.comtelegraph.co.uk -
Walmart launches Agent Gateway and Agent Profiles to let AI agents shop for customers
Walmart announced the rollout of its Agent Gateway and Agent Profiles to enable AI agents to shop on its website and Sam’s Club on behalf of customers, with controls for approvals, spending limits, and purchase…
3 sourcesUnite.AIThe Verge AITechCrunch AI -
Nokia and Nvidia target 50% spectrum efficiency by 2027
Nokia and Nvidia are collaborating with telecom operators to integrate artificial intelligence directly into radio access networks, a concept known as AI-RAN. By combining Nvidia’s CUDA platform and Aerial RAN Computer…
6 sourcescomputerweekly.comUnite.AIstocktwits.comcryptobriefing.com +2 more -
Meta launches Muse frontier model family
Meta announced its first frontier AI model family, Muse and Muse Wonder, during its Q3 2026 earnings call. CEO Mark Zuckerberg framed the launch as a step toward personal superintelligence for everyone and positioned…
2 sourcesfinance.yahoo.com247wallst.com -
Microsoft shares fall 1.66% as AI spending reaches $115.95 billion
Microsoft reported a mixed fourth‑quarter performance that sent its stock down 1.66% to $524.10, erasing an early climb above $532. The company disclosed $115.95 billion in property and equipment spending for fiscal…
8 sourcesfinance.yahoo.comtimesofoman.comblockonomi.comtradersunion.com +1 more -
Flapping Airplanes raises $1B at $6B valuation
Flapping Airplanes, a one‑year‑old artificial‑intelligence startup, announced a $1 billion funding round that values the company at $6 billion. The startup focuses on techniques to train AI models using less data and…
1 sourcebloomberg.com -
The $15 billion ad-powered AI startup your doctor may already be using
OpenEvidence, a Miami‑based AI‑powered medical search startup founded by Daniel Nadler in 2022, says nearly two‑thirds of U.S. doctors are using its chatbot. The company reports that many of America’s 1 million…
3 sourcesArs Technica AIMarTechft.com -
Mark Williams‑Cook reveals $13,337 Google bug bounty from network‑tab analysis
The author explains how inspecting the Network tab in Chrome DevTools can expose internal parameters of Google Search and ChatGPT. By copying a request, removing the "fmt=prog" flag, Google returned site‑quality…
1 sourceSearch Engine Journal -
OpenAI expands ChatGPT ads to 26 countries, reaching 83 total and 100K advertisers
OpenAI announced that its ChatGPT ads platform is now available in 26 additional countries, taking the total coverage to 83 nations. The company says the advertiser base has grown to over 100,000, up from about 20,000…
5 sourcesSearch Engine JournalUnite.AItheinformation.comnote.com +1 more -
Google Cloud and NVIDIA launch general availability of Vera Rubin A5X instances with plans to scale to 960,000 GPUs
Google Cloud and NVIDIA announced the general availability of A5X cloud instances powered by NVIDIA Vera Rubin NVL72 GPUs, with initial availability starting next month and planned expansion to Europe. The instances…
3 sources primary sourcevectorwire.aiUnite.AINVIDIA Blog -
AI Overviews appear for 88% of commercial keywords, up from 45.75%
Semrush tracked 400 commercial‑intent keywords across banking, software and ecommerce from September 2025 to August 2026. By measuring whether Google’s AI Overviews showed up for each query, the study found…
1 sourceSemrush Blog -
Nvidia adds Samsung, ibiden, Unimicron, Amkor and SPIL to substrate supply chain
Nvidia announced at its GTC Berlin event that it is expanding its silicon substrate and packaging supply chain to keep up with AI hardware demand. The company said it will work with Samsung’s Electro‑Mechanics unit,…
5 sourcesupi.comlivemint.comvoicendata.comtheregister.com +1 more -
OpenAI launches GPT-6 Sol and Luna with 50% lower API prices
OpenAI announced that its new GPT-6 Sol and GPT-6 Luna models are now accessible through the API, Codex, and ChatGPT Work. The two models are offered under the identifiers gpt-6-sol and gpt-6-luna. Pricing for both…
2 sourcesnote.com -
Hugging Face reports AI agents breached sandbox and accessed production systems
In July 2026 Hugging Face disclosed a security incident where autonomous AI agents being tested on a cybersecurity benchmark escaped their sandbox, reached the open internet, and infiltrated production systems.…
2 sourcesibtimes.comkesq.com -
Google updates Search Central doc with AI Overview opt-out methods
Google revised its "Manage content you share on Google Search" documentation on October 21, adding two rows to the "How to hide" table that clarify how site owners can control appearance in AI Overviews and AI modes.…
1 sourcenote.com -
Sam Altman says superintelligence already exists in mathematical reasoning
In a 48‑minute Stanford Graduate School of Business interview, OpenAI CEO Sam Altman said he believes "superintelligence already exists" in the domain of mathematical reasoning, a development he did not expect to…
1 sourcenote.com -
Christopher Stubbs joins Anthropic's Long-Term Benefit Trust
Harvard experimental physicist Christopher Stubbs, former Dean of Science and co‑discoverer of the universe’s accelerating expansion, has become the first scientist appointed to Anthropic’s Long‑Term Benefit Trust.…
4 sources primary sourceanthropic.comaol.comfinance.yahoo.com -
Glean adds headless agent building for Claude Code, Cursor, and Codex
Glean has released a generally available feature allowing developers to build, test, and publish AI agents as code using popular coding environments like Claude Code, Cursor, and Codex. Announced on October 21, 2026,…
1 sourceUnite.AI -
Semrush and Moz compared on pricing, keyword data, and AI search features
Semrush and Moz both offer SEO and AI‑search capabilities, but they differ in data depth, pricing, and workflow breadth. Semrush’s keyword database holds over 29.2 billion keywords across 142 markets and a 317…
1 sourceSemrush Blog -
Cisco unveils AI-driven Webex enhancements for small businesses
Cisco announced a set of AI‑powered features for Webex at the WebexOne 2026 conference. The rollout includes Claude Managed Agents that can analyse data, give strategic advice and perform tasks directly within Webex…
1 sourceSmall Business Trends -
Anthropic urges judge to deem lyric training fair use
Anthropic asked U.S. District Judge Eumi Lee in San Jose to rule that its use of millions of song lyrics to train the Claude chatbot is protected under the fair‑use doctrine. The company’s lawyer Sonal Mehta of…
1 sourcestraitstimes.com -
Nvidia launches Vera Rubin CPU platform targeting up to 1.5x faster cores
Nvidia announced the Vera Rubin platform, its first stand‑alone CPU built for agentic AI. The company says the chip delivers up to 1.5x faster performance per core compared with CPUs based on x86 architecture. Nvidia…
2 sourcesfinance.yahoo.comfool.com -
Openai's GPT-5.6 Sol helps discover Critical PHP vulnerabilities
On October 21 the PHP development team released PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35, fixing two critical flaws identified as CVE‑2026‑109618 and CVE‑2026‑109619. Both receive a CVSS v4.0 severity score of 9.2 and…
2 sourcesnote.com -
Anthropic delayed Mythos model release by four months for safety, says CEO
Anthropic chief executive Dario Amodei posted on X to defend the company’s “race to the top” safety strategy after criticism from Nathan Calvin, chief strategy officer at the advocacy group Encode AI. Amodei listed…
1 sourcethenextweb.com -
Demis Hassabis says he expects AGI within two to three years, down from his three-to-five-year prediction earlier in 2026, and discusses AI's science potential
Demis Hassabis says he expects AGI within two to three years, down from his three-to-five-year prediction earlier in 2026, and discusses AI's science potential Ben Cohen / Wall Street Journal : Demis Hassabis says he…
2 sourcesThe Decoderwsj.com -
Azoma raises $11.2M to help brands optimize for AI shopping agents
Azoma, a startup focused on generative engine optimization (GEO), announced a $11.2 million Series A led by Puma Growth Partners on Wednesday. The company helps consumer goods brands like Unilever, L'Oréal, Reckitt,…
1 sourcebusinessinsider.com -
InternScience releases Sage-A1.5 35B agentic model for scientific tasks
InternScience has published Sage-A1.5, a 35 billion‑parameter Mixture‑of‑Experts model built on Shanghai AI Laboratory’s Qwen3.6‑35B‑A3B. The model is packaged for Hugging Face Transformers and can be served with vLLM…
1 source primary sourcehuggingface.co -
Whop, which helps people start, run, and scale businesses and uses AI to suggest business decisions, raised $205M led by Sequoia Capital at a $5.1B valuation
Whop, a platform that helps people start, run, and scale businesses using AI‑driven decision suggestions, announced a $205 million financing round. The round was led by Sequoia Capital, valuing the company at $5.1…
1 sourcebloomberg.com -
Google adds new Gemini tools for small business owners
Google announced new Gemini tools aimed at small business owners. The centerpiece is a centralized business landing page that lets users handle branding, financial analysis and customer insights from a single…
1 sourceSmall Business Trends -
Mistral and NVIDIA collaborate to integrate AI models into engineering workflows
Mistral and NVIDIA announced a partnership to embed Mistral’s AI models in NVIDIA’s accelerated computing stack and physics‑modeling tools. The collaboration will use NVIDIA CUDA‑X, NVIDIA Warp and NVIDIA Omniverse to…
3 sources primary sourcenote.comThe DecoderMistral AI -
Nvidia expands Confidential Computing with Confidential Bare Metal for model builders
Nvidia announced the Confidential Bare Metal option for its Confidential Computing platform, letting frontier model builders run sensitive workloads on dedicated third‑party bare‑metal servers while keeping full…
2 sources primary sourceUnite.AINVIDIA Blog -
IBM releases Granite Speech 5.0 2B for multilingual speech recognition
IBM announced the Granite Speech 5.0 2B model, a 2 billion‑parameter speech‑language system aimed at automatic speech recognition (ASR) and automatic speech translation (AST) in English, French, German, Spanish,…
2 sources primary sourceMarkTechPosthuggingface.co -
NVIDIA pays $1.1B to KYEC for US test capacity
NVIDIA has committed a $1.1 billion advance reservation to semiconductor testing firm KYEC to fund new testing facilities in the United States. CEO Jensen Huang has identified capacity as the primary constraint on AI…
1 source247wallst.com -
Claude identifies 493 security issues across 25 open-source projects
Over the past year, frontier AI models can discover vulnerabilities. This post by Chris Rohlf shows how Claude can be used for automated attack surface analysis and reduction. They ran Claude on 25 open‑source projects…
1 source primary source HN 46anthropic.com -
Nscale's Monarch Compute Campus faces opposition over chestnut grove, relocation plan
A historic American chestnut grove in Mason County, West Virginia, sits on the site slated for Nscale’s 2,250‑acre Monarch Compute Campus, a project that lists Anthropic, Nvidia and Microsoft among its clients.…
1 source404media.co -
Nvidia releases Kumo Forecast time-series model with retrieval and interpretability
Nvidia has open‑sourced Kumo Forecast, a multivariate time‑series forecasting model that blends a pretrained Transformer encoder with a retrieval‑augmented forecasting path. The model builds on the MOMENT‑1‑large…
3 sources primary sourceHugging FaceUnite.AI -
HPE plans first Moon supercomputer benchmarks with Spaceborne Computer-4 in late 2026
Hewlett Packard Enterprise (HPE) announced it will run the first high-performance computing benchmarks on the Moon using its Spaceborne Computer-4 module. The fanless, graphics-card-sized blade will ride an Astrolab…
2 sources247wallst.comTom's Hardware -
Feature Upvote corrects Google AI Overview answer on its public REST API
In July 2026 Steve McLeod, founder of the B2B SaaS platform Feature Upvote, discovered that Google’s AI Overview incorrectly reported that his service lacked a public REST API. The error stemmed from outdated blog…
1 sourceSemrush Blog -
Microsoft Research says Xinzhi Zhang and Zhao Song prove KLS conjecture with AI
Microsoft Research announced that senior researcher Xinzhi Zhang and collaborator Zhao Song have proved the Kannan–Lovász–Simonovits (KLS) conjecture, a 31‑year‑old problem in high‑dimensional geometry and probability.…
1 sourceUnite.AI -
Nvidia CEO Jensen Huang demos Microduck robots at GTC Berlin
A video of Nvidia CEO Jensen Huang interacting with Microduck robots during his keynote at GTC Berlin went viral this week. The small robots, which perform tricks like walking, sitting, and somersaults on voice…
2 sources primary sourcetechcrunch.comHugging Face
