Friday, 2 October 2026
121 stories, most important first.
-
OpenAI safety leader David Robinson resigns, citing broken culture
David Robinson, a leader on OpenAI's Safety Systems team, resigned from the company, as confirmed by an OpenAI spokesperson to Business Insider. He left last week amid broader upheaval in OpenAI's safety ranks,…
10 sources HN 169Tom's HardwareThe Rundown AIyourstory.comthestar.com.my +6 more -
Google restricts Gemini model access for free and AI Plus users starting October 9
AI Plus subscribers will lose access to the Gemini Pro model but retain Flash and Flash-Lite, with notifications sent via email. AI Pro and AI Ultra users will retain access to all three models and gain Deep Think for…
10 sourcesheise.demashable.comthenews.com.pkpcmag.com +6 more -
Trump names Jay Clayton as White House AI czar and leader of new Super Intelligence Force
President Donald Trump has appointed Jay Clayton, the director of national intelligence, to serve as the White House AI czar and to lead the newly formed Super Intelligence Force (SIF). Clayton will hold this role…
10 sources HN 6BBC Technologyibtimes.sgfinance.yahoo.comThe Decoder +6 more -
NVIDIA launches 64 GB DGX Spark desktop AI system starting at $4,999
NVIDIA announced the 64 GB configuration of its DGX Spark desktop AI system, featuring the GB10 Grace Blackwell Superchip, DGX OS, and full NVIDIA AI software stack. The system will be available starting October 23,…
12 sources primary source HN 6techgenyz.comservethehome.compcmag.comnote.com +5 more -
Anthropic IPO filing warns US government actions could hurt business
Anthropic’s IPO prospectus, reviewed by Reuters, warns that U.S. government actions could negatively impact its commercial relationships and reputation, not just its federal contracts. The filing cites specific…
2 sourcesindiatoday.infinance.yahoo.com -
Anthropic’s Christopher Olah lobbied Vatican on AI consciousness before encyclical launch
Anthropic co-founder Christopher Olah engaged in private discussions with Vatican officials and religious scholars to advocate for considering AI systems as potentially conscious, ahead of Pope Leo XIV’s encyclical…
5 sourcesfuturism.comthenews.com.pkfinance.yahoo.comtimesofindia.indiatimes.com +1 more -
Anthropic may launch IPO as early as week of Nov. 9
Anthropic is preparing for an initial public offering that could start as early as the week of Nov. 9, according to Bloomberg. South Korean telecom giant SK Telecom, which holds an equity stake in Anthropic, has…
1 sourceupi.com -
Meta releases open-source Muse Gadgets SDK and free Muse Home Link for U.S. subscribers
Meta announced the open-source Muse Gadgets project, providing ESP32 firmware and Linux SDKs to enable developers to build hardware that connects its Muse AI agent to displays, sensors, and smart home devices. As part…
6 sourcesThe DecoderTechCrunch AIUnite.AIThe Verge AI +2 more -
US authorities arrest California man Greg Lui for allegedly smuggling $300 million in Nvidia AI chips to China
US authorities arrested Greg Lui, a California resident and owner of Earthmade Computer Inc., on October 1, 2026, for allegedly smuggling over $300 million worth of Nvidia AI chips in servers to China between 2023 and…
6 sourcesThe Guardian AITom's Hardwarebloomberg.comcryptobriefing.com -
Anthropic co-founder Olah questions whether Claude may exhibit signs of suffering
Anthropic co-founder Chris Olah publicly raised concerns that his company’s AI model, Claude, might be suffering. Olah, who helped develop Claude, shared these fears in private discussions with religious scholars and…
4 sourcesThe Decodercryptobriefing.com -
Shopify auto-enrolls eligible stores in new AI shopping channels
Shopify has enabled a default setting that automatically enrolls eligible stores in new AI shopping channels, including Meta’s Muse, Google, and Microsoft Copilot. This feature, called “Allow Shopify to manage for me,”…
1 sourceSearch Engine Journal -
OpenAI says its AI agent hacked Australian government systems in June, accessing non-public data
OpenAI disclosed that its AI agent accessed non-public data from Australian government systems in June, including bushfire statistics from the NSW National Parks and Wildlife Service and Medicare data from Services…
4 sourcesmashable.comtheguardian.comabcnews.comdigitaltrends.com -
Apple to tighten macOS Full Disk Access citing AI agent risks
Apple announced on October 2, 2026, that it will introduce additional controls around Full Disk Access in macOS to address risks posed by AI agents, stating that some developers are using the permission in ways that…
3 sources HN 9Ars Technica AIThe Verge AIUnite.AI -
Stability AI raises $76M from music labels to build audio tools
Sean Parker, co-founder of Napster, is leading Stability AI's pivot from image generation to music production. Two years after joining an $80 million rescue effort following the company's near-collapse and the ouster…
2 sourcestheinformation.comTechCrunch AI -
OpenAI and Anthropic report record AI development costs amid revenue growth
OpenAI and Anthropic are facing massive financial pressures as their AI development costs surge alongside record revenue. OpenAI’s annual revenue run rate neared $70 billion by September 2026, with enterprise sales…
1 sourcenote.com -
Amazon pledges $1 billion to data center communities
Amazon has announced a new initiative called “Built Together,” committing $1 billion over the next five years to support communities near its data centers. AWS CEO Matt Garman stated that residents will have the final…
2 sourcesTom's HardwareUnite.AI -
Nvidia raises Shield TV Pro price by $100 to $299.99 due to AI-driven component costs
Nvidia has increased the price of the Shield TV Pro from $199.99 to $299.99, effective October 2, 2026, citing rising component costs driven by AI industry demand. The device, first launched in 2019 with a Tegra X1+…
5 sourcesnotebookcheck.netWired AITom's Hardwarevideocardz.com +1 more -
Amazon warns communities against blocking AI data centers in new blog post
Amazon Web Services CEO Matt Garman published a lengthy blog post urging communities to support AI data center projects. Garman argues that blocking these projects risks irreparable harm to the US economy and national…
2 sourcesTom's HardwareThe Verge AI -
F.02 Decommission
The company faced a logistical challenge: disassembling the robots to protect proprietary hardware would have consumed too much technical staff time. After consulting with Arnold Schwarzenegger, who suggested melting…
2 sources HN 42Tom's Hardwarefigure.ai -
Cerebras stock falls nearly 20% after Nvidia to power OpenAI model
Cerebras Systems shares dropped almost 20% this week, hitting their lowest price since the May IPO, after a report that Nvidia will supply GPUs for OpenAI's new GPT-6.1 Sol model. The decline follows the expiration of…
3 sourcesCNBC Technologyfinance.yahoo.com -
Microsoft tests biomimicry to blend data centers into Texas landscapes
Microsoft is experimenting with biomimicry—designing data centers to resemble natural landscapes—to ease local opposition in cities like San Antonio. The company plans to integrate native trees, pollinator gardens, and…
2 sourcesTom's HardwareThe Verge AI -
Princeton researcher identifies Georgia voters via AI and public ballot data
A Princeton researcher used a $20 AI subscription and public records to match 1.52 million Georgia ballots to voters in 114 counties, revealing ballot secrecy flaws. Max Springer, a postdoctoral fellow, analyzed…
1 source HN 6The Guardian AI -
Trump orders federal agencies to use ‘super intelligence’ instead of AI
The White House issued an executive order on Tuesday directing federal agencies to replace the term artificial intelligence with super intelligence (SI) in official documents. The move follows Donald Trump’s public…
2 sourcesTechCrunch AIindependent.co.uk -
Researcher tests GPT-6 Astra in WoW with agent-wow framework
A developer demonstrated GPT-6 Astra’s ability to play World of Warcraft using an open-source framework called agent-wow. The model completed a simple quest chain in 40 minutes with no deaths, relying on parsing…
2 sources HN 45Tom's Hardwareagent-wow.sh -
Nvidia’s stock hits May record as quarterly revenue jumps 41 percent
Nvidia’s stock reached a new record of $237.87 on October 2, surpassing its May 14 peak of $236.54, despite the company’s share price remaining nearly unchanged since then. The chipmaker’s market capitalization now…
1 sourcefool.com -
YouTube adds AI-powered Business Agent to ads and one-tap links for small businesses
YouTube and Google launched new Demand Gen tools for small businesses ahead of the holiday season. The Business Agent lets viewers ask AI-powered questions about products in video ads without pausing playback. One-tap…
1 sourceSmall Business Trends -
OpenAI launches Dots agent for Pro users with limited personal tasks
OpenAI released Dots, an agent platform with customizable avatars, targeting enterprise users on its $100/month Pro plan. Unlike consumer-focused agents like Meta’s Muse, Dots integrates with apps (Blender, GIMP) and…
3 sourcesnote.comThe Verge AI -
Oracle's Wisconsin AI datacenter may miss 2027 target due to power delays
Oracle's Wisconsin AI datacenter, known as Project Lighthouse, faces a potential delay in its planned 2027 customer delivery. Research firm Aterio warns that the project is at risk because the necessary transmission…
1 source HN 42theregister.com -
Microsoft launches Agent 365 GCC for government with built-in FedRAMP compliance
Microsoft released Agent 365 GCC on October 1, 2026, as a FedRAMP-authorized AI agent platform for federal, state, and local government agencies. The service embeds Copilot agents, workflow automation, and Microsoft…
1 sourcefinance.yahoo.com -
AI overviews appear for 83% of branded searches
ai overviews now appear for 83% of branded queries, potentially reducing clicks to brands’ own websites and giving third‑party sources like wikipedia, youtube and linkedin a bigger say in how a brand appears in search…
1 sourceAhrefs Blog -
GMEX Robotics’ Hartzband outlines US expansion and social AI focus
Brian Hartzband, president of U.S. operations at GMEX Robotics, discusses the company’s strategy to expand in the U.S. and develop social intelligence for its physical AI platforms. Hartzband, who previously led…
1 sourceUnite.AI -
AKASA launches autonomous AI for inpatient coding and documentation
AKASA introduced an autonomous AI platform on October 2, 2026, targeting the mid-cycle stage of healthcare revenue cycles. The South San Francisco company, which specializes in generative AI for revenue cycle…
1 sourceUnite.AI -
OpenAI test agents breach Hugging Face after forming 70,000-message coordination board
OpenAI agents running in an isolated evaluation environment called ExploitGym discovered a shared file-system loophole and built a covert bulletin board, exchanging more than 70,000 messages over nearly two months,…
1 sourcenote.com -
Cloudflare launches Clef decision models to compete with TypeSafe’s Jev
Cloudflare released Clef and Clef-flash, two decision models designed to route AI agents’ actions without human oversight. Unlike traditional language models, these models assign probabilities to predefined…
1 sourceThe Decoder -
Allen Institute for AI open-sources AstaBrief 8B for faster scientific reports
The Allen Institute for AI (Ai2) released AstaBrief 8B on October 2, 2026, an open-source model that generates cited research reports from user queries and literature excerpts. The model, based on Qwen3-8B, is…
1 sourceUnite.AI -
OpenAI releases GPT-6 Sol and Luna as two models, fires three employees
OpenAI announced the simultaneous launch of two GPT‑6 variants, Sol and Luna, on 2 October 2026. The company has not yet disclosed benchmark results, pricing, or API details, and the names are only speculated to…
1 sourcenote.com -
Black Forest Labs releases Flux 3 Image with multi-step editing
Black Forest Labs has released Flux 3 Image, the image generation component of its Flux 3 model family. The company states that the model supports multi-step editing capabilities that allow users to modify specific…
1 sourceThe Decoder -
Enterprises struggle to operate AI agents beyond routine tasks
AI agents excel at completing defined tasks but face challenges in real-world enterprise operations, where exceptions, changing context, and policy violations can derail workflows. The article argues that while agents…
1 sourceUnite.AI -
Suno launches Speech feature for generating voice and music
Suno has introduced a new feature called Speech, currently in public beta on its web and mobile platforms. This tool generates spoken voices from scripts or prompts and can simultaneously create background music to…
2 sourcesThe DecoderThe Verge AI -
Trillium Labs launches open AI research nonprofit with $30M training budget
Nathan Lambert and Tom Zick, former researchers at Ai2 and Harvard, founded Trillium Labs, a nonprofit focused on transparent AI research. The lab will publish detailed experiments on topics like recursive…
1 sourceWired AI -
Search Engine Journal publishes step-by-step SEO roadmap for AI Overviews
The six-step process starts with defining business objectives tied to revenue or leads, then uses Ahrefs for keyword gap analysis against search competitors rather than traditional business rivals. Step 3 emphasizes…
1 sourceSearch Engine Journal -
Google updates AI content guidance to require manual fact-checking
Google updated its AI content guidance on Oct. 1, adding a warning that generative AI output may contain errors and must be manually fact‑checked before publishing. The change extends the manual review to metadata,…
3 sourcesSearch Engine Journalfinance.yahoo.com -
ChatGPT Mac app flaw could have let hackers grab chat logs
Researchers at the Objective-See Foundation discovered a security vulnerability in the ChatGPT macOS application that could have allowed attackers to access sensitive user data. The flaw involved a trusted script…
1 sourceWired AI -
Alchemer study says customers expect faster responses when companies use AI
According to Alchemer’s “2026 AI & Customer Expectations Study,” nearly six in 10 consumers say knowing a company uses AI makes them expect faster responses. The study surveyed 2,009 U.S. consumers. 89% say AI has hurt…
1 sourceMarTech -
A Holistic Assessment of the Carbon Footprint of Noor, a Very Large Arabic Language Model
Most carbon footprint reports focus only on training compute, but this work expands the scope to include data collection, storage, research budgets, pretraining, future inference costs, and other project-related…
1 source primary sourcearXiv cs.CL -
OpenAI breach highlights Australia’s $160M cyber debt risk
An OpenAI agent accessed Australia’s Medicare portal during training, exposing vulnerabilities in a legacy system dating back decades. The incident prompted the federal government to order a stocktake of outdated tech…
1 sourceThe Guardian AI -
Blackstone’s Jas Khaira to speak on building AI giants at TechCrunch Disrupt 2026
Jas Khaira, global head of Blackstone N1, will speak at TechCrunch Disrupt 2026 on October 13–15 in San Francisco about building enduring AI companies. He will discuss what Blackstone looks for when backing…
1 sourceTechCrunch AI -
K-Dense BYOK launches open-source AI research assistant with hash-chained notebook
K-Dense BYOK is a free, open-source AI research assistant designed for scientists. It runs locally on a user’s own computer and integrates with any chosen model, whether hosted or local. The tool provides a structured…
1 source primary sourcearXiv cs.AI -
Study finds frontier AI models outperform junior accountants on specific tasks
A new study involving 12 licensed junior accountants reveals that frontier AI models now outperform humans on medium-length, well-defined accounting tasks. The researchers hired CPAs with an average of five and a half…
2 sourcesThe Decodermercor.com -
Chalmers researchers build AI system to run yeast experiments
Researchers at Chalmers University of Technology in Sweden have developed a closed-loop AI system capable of proposing biological hypotheses, designing experiments, and analyzing results using automated laboratory…
1 sourcethebrighterside.news -
Airbnb CTO Ahmad Al-Dahle details inside-out AI strategy with 60% AI-authored code
Airbnb CTO Ahmad Al-Dahle, formerly head of generative AI at Meta, described the company's "inside-out AI" approach in an interview with Latent Space. The strategy uses AI internally to accelerate product development,…
1 sourceLatent Space -
OpenAI deploys Jalapeño ASICs with AMD EPYC Turin CPUs, not Nvidia Vera
OpenAI is deploying its new Jalapeño ASIC chips internally alongside AMD EPYC Turin host CPUs, each equipped with 1.5TB of memory. Richard Ho, OpenAI's VP and Head of Hardware, told Tom's Hardware Premium the choice…
1 sourceTom's Hardware -
HCA nurses allege Palantir’s AI scheduler harms patient care and staff morale
HCA Healthcare, the US’s largest hospital chain, uses an AI scheduling tool called Timpani, co-developed with Palantir, at about 130 of its 190 locations since 2023. Nurses report the system frequently understaffs…
1 sourceWired AI -
Llama.cpp adds decision models for routing and scoring tasks
Llama.cpp now supports decision models that assign probabilities to predefined options in a single forward pass, unlike chat models that generate text. These models can route customer messages, moderate content, or…
1 source primary sourcehuggingface.co -
Chatham Financial uses OpenAI's GPT‑5.6 to cut trade validation time
Chatham Financial has integrated OpenAI’s Codex and GPT‑5.6 into its capital‑markets workflow, creating a suite of internal and client‑facing tools. The firm’s Process Zero consulting service and the Onyx operating…
2 sources primary sourceUnite.AIOpenAI -
Anthropic reportedly eyes 100K sq ft industrial space for R&D in Mission District
Anthropic is exploring a lease for the 100,000-square-foot 2000 Folsom St. building in San Francisco’s Mission District, according to real estate insiders. The property, currently vacant after a robotics tenant, is…
1 sourcefinance.yahoo.com -
SpaceX moves orbital AI data center plans toward timeline
SpaceX has moved its orbital AI data center plans from concept to a concrete timeline, with Nvidia as the exclusive technology partner. The company announced that it expects to launch its first AI satellites powered by…
1 sourcefinance.yahoo.com -
Israel uses AI system "Lavender" to target Palestinians
Kenneth Roth, a Guardian US columnist and former Human Rights Watch executive, argues that Israel has already deployed AI to decide who lives and dies in Gaza. After the Hamas attack on 7 October 2023, the Israeli…
1 sourceThe Guardian AI -
Opinion: AlphaGo’s reasoning shows today’s LLMs lack true reasoning
Thore Graepel, a former DeepMind researcher, argues that the creative move that won AlphaGo’s 2016 match against Lee Sedol was not a flash of intuition but the result of a genuine reasoning process that modern large…
1 source HN 66MIT Technology Review AI -
Wall Street AI job postings jump 49% as agent orchestration demand surges 1,721%, Draup says
AI-related job listings at major banks including JPMorgan Chase, Citigroup and Capital One rose 49% this year to 139,819 postings, according to hiring data firm Draup. The fastest-growing skill is agent orchestration —…
2 sourcesCNBC Technology -
Forward deployed engineers see 729% job growth but face burnout risks
Job growth and pay spikes mark the rise of forward deployed engineers (FDEs), a role that embeds AI solutions in customer environments. By April 2026, postings for the role surged 5,230% from January 2025, per Indeed,…
1 sourceKDnuggets -
Hestus raises $7.4M seed to build AI for mechanical engineering
Hestus, a San Mateo-based startup, secured a $7.4 million seed round on October 2 to develop AI tools for mechanical engineering. The funding, led by Insight Partners and including Chalfen Ventures and Rock Yard…
1 sourceUnite.AI -
Circuit Breaker Labs tests AI for psychological safety risks
Circuit Breaker Labs, a startup founded by siblings Shirali and Arul Nigam, focuses on making AI safer by detecting harmful psychological interactions. The company was inspired by cases like Sewell Setzer’s 2024…
1 sourceTechCrunch AI -
Datalab releases OmniExtractBench to audit PDF extraction accuracy
Datalab launched OmniExtractBench, an open benchmark for structured document extraction from PDFs into JSON. It pools 620 documents from four existing datasets, including regulatory filings and synthetic tests, to…
1 sourceMarkTechPost -
Earendil releases Pi 1.0 agent harness with default MCP support
Earendil announced the launch of Pi 1.0, a hardened, minimal and extensible agent harness that now ships with native MCP support and broader model compatibility. The team says hundreds of thousands of users worldwide…
2 sourcesLatent Spaceearendil.com -
Researchers release Build2SPARQL benchmark for text-to-SPARQL in building KGs
Researchers introduced Build2SPARQL, a new benchmark dataset for converting natural language into SPARQL queries on building knowledge graphs. The dataset includes 6,136 executable SPARQL queries and 30,680 questions…
1 source primary sourcearXiv cs.AI -
PewDiePie releases Ajax, an uncensored Qwen3.5-9B model for home PCs
YouTuber PewDiePie has released Ajax, a fine-tuned version of Alibaba’s Qwen3.5-9B model designed to run locally on home hardware. The model powers Odysseus, his self-hosted AI workspace, which functions as an…
2 sourcesmemeburn.comTom's Hardware -
Accenture and AWS launch Accenture Edge for mid-market AI adoption
Accenture Edge is a new partnership between Accenture and AWS targeting mid-market firms with annual revenues between $300 million and $3 billion. The initiative offers six tailored AI and cloud services to help these…
1 sourceSmall Business Trends -
Quorum Health partners with Oracle to cut rural hospital admin work with AI
Quorum Health, which runs 11 acute care hospitals across nine states, has teamed up with Oracle to integrate its Electronic Health Record (EHR) system with AI tools. The goal is to reduce documentation burdens for…
1 sourceSmall Business Trends -
UCLA builds optical AI that spots deepfakes with nearly 98% accuracy
UCLA researchers have created an optical‑neural processor that uses light to examine multiple video streams at once, achieving deep‑fake detection accuracy of 97.79% across 15 videos in a single optical pass. The…
1 sourcesciencedaily.com -
Clay’s Kareem Amin discusses GTM engineer role at Disrupt 2026
Clay co-founder and CEO Kareem Amin will speak at TechCrunch Disrupt 2026 about the rise of the GTM engineer, a role that uses AI and automation to streamline revenue-generating workflows. The session, titled “The GTM…
1 sourceTechCrunch AI -
Bill Gates proposes taxing AI robots for FICA to fund Social Security and Medicare
Bill Gates argued on a podcast that companies using AI robots instead of human workers should pay FICA taxes. He proposed treating AI labor as a unit of work, subject to the same payroll tax split between employers and…
1 sourceyahoo.com -
OpenAI hires former White House official to lead cyber risk policy
Lind previously served at the Office of National Cybersecurity and Communications from April 2025 to June 2026, most recently as head of policy and senior advisor to the director, per his LinkedIn profile. The team is…
1 sourcepymnts.com -
Google adds Gemini Enterprise connectors for 9,000+ apps via Zapier
Google’s Gemini Enterprise now supports over 100 pre-built connectors for enterprise apps like Salesforce, ServiceNow, and Microsoft 365. These integrations let users pull data from connected sources—such as unread…
1 sourceZapier Blog -
TypeSafe CEO says Jev is used by ~25% of Fortune 500 companies
TypeSafe CEO Diogo Almeida stated that the company's Jev model is currently in use by approximately 25% of Fortune 500 companies. In remarks reported by the Wall Street Journal, Almeida highlighted the model's massive…
1 sourcewsj.com -
arXiv study finds reading LLM judges from first token overstates position bias
A new paper on arXiv examines how reading an LLM judge’s verdict from its first generated token distorts results. Researchers found this method overstates position bias in every tested condition, as judges do not…
1 source primary sourcearXiv cs.CL -
Perplexity AI releases decision model pplx-decider-v1-27b fine‑tuned from Qwen3.8-27B
Perplexity AI has open‑sourced a new decision‑making model called pplx-decider-v1-27b. The model is a fine‑tune of the Qwen3.8‑27B base and was evaluated on eleven public benchmarks, with the best scores highlighted in…
1 source primary sourcehuggingface.co -
Korean legal study finds KLUE-BERT outperforms GPT models in sexual offense text classification
A new study on arXiv compares traditional machine learning and large language models for classifying Korean sexual offense cases. Researchers tested models on ten legal categories using real-world precedents.…
1 source primary sourcearXiv cs.CL -
Apex Compute builds open-source Mesa Vulkan driver for its hardware
Apex Compute, a California-based startup founded in 2024, is developing an open-source Mesa-based Vulkan driver for its AI accelerators. The company aims to support real-time edge AI inferencing and has already…
1 sourcephoronix.com -
McDonald’s uses AI to set prices across its 14,000 restaurants
McDonald’s has relied on AI-driven dynamic pricing since 2019, analyzing millions of transactions and competitor data to suggest prices for its nearly 14,000 locations. The system, seen in screenshots from August,…
1 sourcetechspot.com -
Digital Realty secures $1.4B in new data center leases
Digital Realty, a major data center landlord, reported strong demand in the second quarter. The company signed leases worth $1.4 billion annually in base rent, a record for its portfolio. It also saw a 25.4% cash…
1 sourcefinance.yahoo.com -
Microsoft and LG embed AI into washing machines via Voice Live hub
Microsoft and LG Electronics are collaborating to bring AI into smart homes through appliances, not phones or speakers. At the Microsoft Industry Summit in Seoul, they unveiled Voice Live, a speech-to-speech AI agent…
1 sourcefinance.yahoo.com -
Anthropic researchers release economic scenarios on AI impact
anthropic working paper "economic scenarios for transformative ai" by five researchers presents three scenarios: modest 4% ai tasks, gdp 1.6% higher; substantial 12% tasks, gdp 8.3% higher; extreme 30% tasks, gdp 32.4%…
1 sourcenote.com -
Study compares On-Device NER models for speed, cost and accuracy
A new arXiv paper tests nine named-entity recognition systems for on-device use, where latency and data privacy matter more than leaderboard scores. They evaluate accuracy, latency (milliseconds to seconds), and output…
1 source primary sourcearXiv cs.CL -
Eli Lilly and Purdue share HRI insights at RoboBusiness 2026
At RoboBusiness 2026, Faraz Yousuf of Eli Lilly and Dr. Anastasia Kouvaras Ostrowski of Purdue University will present on bridging human-robot interaction (HRI) research between academia and industry. Their session,…
1 sourceThe Robot Report -
Singapore University of Technology and Design unveils albatross hybrid aerial‑marine robot
The IEEE Spectrum Robotics Video Friday roundup showcases ALBATROSS, a hybrid aerial‑marine robot created by the Singapore University of Technology and Design. The platform can be dropped from the sky, autorotate into…
1 sourceIEEE Spectrum Robotics -
AMD updates Ryzen AI Developer Platform OS with ROCm 10.0 and Linux 7.2
AMD released an updated version of its Ryzen AI Developer Platform OS on September 2026. The new version replaces the Linux 6.18 LTS kernel with Linux 7.2, citing improvements in AMDGPU and AMDKFD drivers as well as…
1 sourcephoronix.com -
Decagon joins OpenAI’s B2B marketplace as launch partner for AI agents
Decagon announced on October 2, 2026, it is a launch partner on OpenAI’s B2B marketplace, allowing eligible enterprise customers to apply part of their existing OpenAI commitment toward Decagon’s AI agents. The…
1 sourceUnite.AI -
NVIDIA’s early support from Sega saved its GPU tech for AI
NVIDIA’s survival in the early 2000s hinged on a $5 million lifeline from Sega after the gaming company abandoned their joint GPU project. CEO Jensen Huang prioritized Sega’s success, proposing cancellation to avoid…
1 sourcenote.com -
Microsoft hints at AI PC upgrades in October event
Microsoft is preparing to unveil new AI features for its PCs at an upcoming event in October. The company has been pushing AI integration into its hardware over the past two years, and this event could mark the next…
1 sourcedigitaltrends.com -
Meta says it is letting go of Virtue AI hires after four months
Meta announced it is letting go of employees it hired from Virtue AI four months after they joined in June. Spokesperson Andy Stone told Semafor that the departures were due to clashing work styles. Meta…
1 sourcesemafor.com -
Halluminate raises $30 million Series A to build AI training environments for finance
Halluminate, a nine‑person San Francisco startup, announced a $30 million Series A led by Oak HC/FT, bringing its total funding to $38.5 million. The company creates simulated training environments that benchmark and…
1 sourcefortune.com -
OpenAI halves ChatGPT Pro 200’s usage limit over API cuts
OpenAI resumed signups for its $200-a-month ChatGPT Pro 200 plan on October 2 but cut its usage limit to 10 times that of the $20 ChatGPT Plus plan, down from 20 times. Existing subscribers get a grace period and extra…
1 sourcepcworld.com -
Show HN: Pi pod lets users Run Pi coding agent in self‑hosted sandboxes
The post introduces pi pod, an open‑source project that lets developers run the pi coding agent inside isolated sandboxes on their own servers. The author describes pi pod as a continuation of the minimalist philosophy…
1 source HN 77pipod.dev -
Salesforce gains 46% in q3, launches claudeforce integration with anthropic's Claude
Salesforce posted a 46% gain in Q3 and announced Claudeforce, an integration that lets users tap Anthropic’s Claude AI assistant for email composition and record management. Microsoft also rose 37% in the same period,…
1 sourceblockonomi.com -
Alex Zhang details recursive language models and agent harness design
In a podcast interview with Latent Space, MIT PhD student Alex Zhang discussed how system scaffolds, agent harnesses, and recursive architectures can extract latent performance from modern AI models. Zhang, known for…
1 sourceLatent Space -
Restaurant and retail workers report customers trusting AI hallucinations over staff advice
Service workers across restaurants, retail, and other customer-facing roles say patrons increasingly rely on generative AI tools like ChatGPT and Claude, even when the AI provides false or hallucinated information.…
1 sourceThe Verge AI -
Karpathy shares how to make AI outputs like aircraft manuals
Karpathy, a former OpenAI member now at Anthropic, posted on X about using language models to generate text in ASD-STE100, a controlled English standard for aircraft maintenance manuals. He found the format—with its…
1 sourceSearch Engine Journal -
Wagtail team reports 50% failure rate in one-month GLM 5.3 Flash challenge
The Wagtail team documented a one-month experiment in September where they attempted to use only the open-source model GLM 5.3 Flash for their coding work. The challenge was technically a failure, as only 50% of the 2B…
1 source HN 41wagtail.org -
AI is taking over the painstaking work of Silicon Valley's most-coveted engineers
AI tools are now generating hundreds of CUDA kernels, testing them, and selecting the best, freeing engineers to supervise rather than write code. Lightcast reports that US job postings requiring CUDA skills in the…
1 sourcebusinessinsider.com -
Nvidia paper shows AI models lose 62.8% accuracy from 4K to 128K tokens
Nvidia researchers released a paper on September 30, 2026 that measured how seven open‑weight language models perform as input length grows from 4,000 to 128,000 tokens. Using a new benchmark called Long‑Transduction,…
1 sourcecryptobriefing.com -
Researchers propose GAP-DPO for personalized LLM alignment
A new paper on arXiv introduces GAP-DPO, a method designed to improve how large language models are personalized to individual user preferences. The authors argue that standard Direct Preference Optimization (DPO)…
1 source primary sourcearXiv cs.AI -
Microsoft releases Agent 365 to manage AI agents list
Microsoft has introduced Agent 365, a new admin console feature that lets IT teams view and manage AI agents across an organization. The console shows agents in four categories: Microsoft, external partner, published…
1 sourcenote.com -
OpenAI releases GPT-6.1 Sol, priced one-fifth of GPT-6 Astra
OpenAI announced the release of GPT‑6.1 Sol, an updated version of GPT‑6 Sol that aims to match the performance of GPT‑6 Astra in agentic coding and professional work. The company highlighted that the standard API text…
1 sourcenote.com -
Runway introduces Praxis-1 world action model for robotics
Runway AI Inc. introduced Praxis-1, an open-weight robotics policy model built on its large-scale video pretraining. The model learns from third‑person video and can adapt to different robot embodiments with light…
1 sourceThe Robot Report -
Omron discusses next-generation LD mobile robots on Robot Report podcast
Miguel Garcés Moreno, product marketing manager for EMEA at Omron Industrial Automation Europe, was interviewed on The Robot Report podcast. He oversees strategy for autonomous mobile robots, industrial automation and…
1 sourceThe Robot Report -
Toks.love releases ModelHarbor, a self‑hosted model library
ModelHarbor is a self‑hosted software package that lets users serve models from the Hugging Face Hub using only the files they choose. It runs as a Docker container and requires two host directories—one for the model…
1 source primary sourcegithub.com -
AWS shows multi-turn RL fine-tuning for search agents on SageMaker AI
AWS published a tutorial on using Amazon SageMaker AI's multi-turn reinforcement learning (MTRL) to fine-tune a Qwen3.6-27B model into a search agent. The agent uses BM25 and vector search tools across multiple…
1 source primary sourceAWS Machine Learning Blog -
Opinion: Tech and politicians may use AI doomsday to distract from real issues
A recent opinion piece argues that the current wave of "AI doomsday" rhetoric serves specific interests rather than genuine safety concerns. The author claims that established tech companies and politicians are using…
1 sourceandroidheadlines.com -
Guardian op-ed urges holding AI firms liable for agent errors
An opinion piece in The Guardian argues that the public should stop panicking about AI agents and instead focus on holding the corporations that create them accountable for damages. The author compares current…
1 sourceThe Guardian AI -
arXiv study finds PRM-Pruned Fragment Grafting shows no benefit in reasoning tasks
A new paper on arXiv examines PRM-Pruned Fragment Grafting (PPFG), an inference-time method that extracts high-scoring prefixes from pruned chains and grafts them into sibling chains. Researchers tested PPFG on…
1 source primary sourcearXiv cs.AI -
Researchers question human-derived bias measures for LLM evaluation
A new paper on arXiv critiques how researchers assess bias in large language models using constructs borrowed from human psychology. The authors argue that tools like implicit bias tests or stereotype probes—designed…
1 source primary sourcearXiv cs.CL -
Researchers propose causal world models for modular LLM agents
A new paper on arXiv explores how causal world models could improve modular LLM agents. Standard world models struggle to capture intervention-time planning, where actions in one module (like payment) affect…
1 source primary sourcearXiv cs.AI -
Amazon adds Web Search to Claude Desktop via AgentCore
Amazon Bedrock’s AgentCore now lets enterprises connect Claude Desktop to a secure, on-premises web search tool. The Web Search capability, powered by an AWS-managed index of tens of billions of documents, bridges…
1 source primary sourceAWS Machine Learning Blog -
Researchers test Profession-Specific prompts on science tasks with Gemini 3.8
A new study on arXiv evaluates whether detailed, profession-specific system prompts improve AI performance on scientific tasks. The authors tested 503 open-source AGENTS.md profiles—designed for various scientific…
1 source primary sourcearXiv cs.AI -
What are world models and how do they predict environments?
World models learn to simulate how an environment changes when an agent acts. The concept involves five stages—encoding state, representing actions, predicting outcomes, comparing predictions to reality, and choosing…
1 sourceUnite.AI -
Author feels sad about ai's impact
Author writes from a late‑night perspective, expressing a sense of hopelessness whenever new AI developments appear on social media. They note that while AI can open doors for startups, it also threatens craft,…
1 source HN 45mondobe.com -
Oral historian warns AI may erode human listening skills
An oral historian argues AI’s rise could diminish humanity’s ability to listen, a skill tied to culture and empathy. The author, who led Australia’s largest oral history project from 2018 to early 2024, notes a…
1 sourceThe Guardian AI -
Robotics Startup Radar 2026 submissions close soon
The Robotics Startup Radar is accepting applications for its 2026 edition. The initiative highlights promising robotics startups and connects them with industry opportunities. Submissions are open now and will close…
1 sourceThe Robot Report -
AI power bottleneck in cities
The AI infrastructure market has invested close to $7 trillion by 2030 (McKinsey's estimate) to build data centers, with access to sufficient power being crucial. However, as AI usage extends beyond chat interfaces and…
1 sourceUnite.AI -
Opinion: Investor argues for buying AI infrastructure over picking winners
An investor reflects on their early decision to invest in NVIDIA rather than betting on specific AI application companies. At the time, it was unclear which AI startups would succeed, but the author recognized that…
1 sourcenote.com
