Iron Device to build on-device AI module for robotic hands
Fabless chip firm Iron Device joined a Korean government-backed project to develop an on-device AI computing module for robotic hands.
Iron Device
Latest news rss
Arm proposes rack-level CPU benchmarks for agentic AI
Arm introduced a CPU benchmarking framework for agentic AI backed by Fujitsu, Meta and OpenAI, and claimed twice AMD's performance in its own testing.
Arm
Ex-Fadu CEO launches AI startup amid inflated-IPO trial
Former Fadu CEO Lee Ji-hyo, who resigned amid an alleged IPO valuation inflation controversy, has founded an AI startup six months later, raising questions about his Fadu stake.
Fadu
SK hynix unveils 3D FeNAND compute-in-memory
SK hynix detailed analog compute-in-memory research using ferroelectric 3D NAND that performs AI computations directly inside flash memory.
SK hynix
Korean battery makers expand ESS push for AI data centers
Korean battery makers are accelerating a shift to energy storage systems as North American demand grows from power-hungry AI data centers.
Rokid opens self-service AI glasses store in Hong Kong
Rokid opened what it calls the world's first self-service AI glasses experience store, at Hong Kong's Harbour City, letting visitors try and buy smart glasses.
RokidClosedOn sale
More: PR Newswire UK
PayPal and Glance to integrate agentic commerce services
Why it matters: Extends agentic commerce and agent payment rails into a major consumer platform.
PayPal and Glance announced they will integrate agentic commerce services, bringing AI agent-driven shopping and payments onto Glance's platform.
PayPal
GPT-6.1 Sol found errors in 6% of nearly 800 math papers
Why it matters: Shows frontier models can systematically audit the mathematical literature.
A Reddit report says GPT-6.1 Sol examined nearly 800 mathematics papers, flagging major errors in 6% and minor mistakes in 30%, with researchers estimating 90% of major findings correct.
OpenAIClosed
Woman fined in Russia over ChatGPT-generated image panic
A woman was fined after generating a ChatGPT image that triggered panic in Russia, according to UNILAD Tech.
FLock.io launches open-source Qwen-based edge decision model
FLock.io released an open-source, Qwen-powered decision model built for local edge deployment, letting developers run decision-making AI on-device.
FLock.ioOpenOpen weightsFree
Apple's first touchscreen MacBook reportedly days away
ZDNET reports Apple has set a date to unveil new touchscreen MacBooks, with an additional OLED surprise expected.
AppleClosed
NTT deploys Nvidia GPU test cluster at Munich 2 data center
NTT set up an Nvidia GPU test cluster at its Munich 2 data center, letting customers trial accelerated AI workloads before committing to deployments.
NTT
Workday eyes usage-based billing as AI agents take on HR work
Workday's CEO wants a slice of customers' labour budgets that its AI agents could replace, moving the HR software vendor toward usage-based billing beyond traditional IT spending.
Workday
MetaMask launches AI agent wallet on Grok bot
MetaMask released an AI agent wallet inside a Grok bot, letting users trade crypto autonomously with the wallet handling secure transactions.
MetaMaskClosedApp
Anthropic PAC spent over $250K in third quarter
Anthropic's political action committee spent more than $250,000 in the third quarter, splitting contributions between the two parties.
Anthropic
Construction robotics startup Origin raises $20M led by Accel
Origin, a construction robotics startup, raised $20 million in funding led by Accel and Zacua Ventures to scale its building-site robots.
Origin
Delta highlights Nvidia 800 VDC power and cooling at OCP
Delta detailed its collaboration with Nvidia on energy-efficient 800 VDC power and cooling for AI data centers at OCP Global Summit 2026.
Delta Electronics
FT examines what happens when AI can read minds
The Financial Times reports brain-computer interfaces are already treating Parkinson's disease and letting paralysed patients operate computers, raising questions as AI decoding advances.
Professor says OpenAI, Anthropic AI lead is 'very perishable'
A professor argued that the current lead held by OpenAI and Anthropic in AI is 'very perishable,' warning investors against assuming durable dominance.
Stripe president says agentic commerce went vertical
Stripe president John Collison said the company's agentic commerce activity went 'vertical' last month, expanding work on AI agents that handle payments and checkout.
Stripe
Cerebras says inference speed is key for agentic workflows
Cerebras argued that agentic workflows make inference speed a requirement, highlighting its speed advantage over GPU-based serving for multi-step agent tasks.
Cerebras
Reuters: market selloff and rising rates test AI data center deal boom
Why it matters: Rising rates threaten the debt-fueled data center buildout underpinning AI compute expansion.
Reuters reports that a broader market selloff and rising interest rates are testing the boom in AI data center deals, raising questions about financing for planned builds.
Bank of America assigns 4,000 people to fix AI-found vulnerabilities
Why it matters: Shows AI-driven vulnerability discovery forcing large-scale remediation in enterprises.
Bank of America brought in 1,000 outside contractors and committed 4,000 people total to patch software flaws surfaced by AI models including Anthropic's Mythos.
Bank of America
GPT-6-Astra solves 49 of 50 physics problems on TheoPhysEval
Why it matters: Shows frontier models approaching expert-level theoretical physics reasoning at low cost.
On the TheoPhysEval benchmark GPT-6-Astra solved 49 of 50 unpublished theoretical physics calculations at about $2.40 each, matching specialists; some reference solutions had errors.
OpenAIClosed
Study finds Korean-language gaps in AI safety tests
A study reports that AI safety evaluations miss or underperform on Korean-language prompts, leaving Korean users less protected.
Report warns federal Anthropic restrictions raise competition concerns
A legal analysis argues federal restrictions on Anthropic could distort AI market competition and disadvantage the company relative to rivals.
Anthropic
Humanoid robot kicks drone across room at conference
Video from a robotics conference shows a humanoid robot kicking a drone across the room during a demo.
Elevatry raises €3M to automate procurement workflows
German startup Elevatry secured €3 million to build AI that automates corporate procurement workflows.
Elevatry
Multiverse Computing launches Comet 60B with Nvidia
Why it matters: Adds a parameter-efficient multimodal competitor tuned for Nvidia hardware.
Multiverse Computing released Comet 60B, a multimodal model built with Nvidia that it says matches larger models at half the parameters.
Multiverse ComputingClosed
More: TradingView
Ubtech Robotics pursues three growth paths at differing maturity
Ubtech Robotics is pursuing three distinct growth paths, each at a different stage of maturity, according to a report on the humanoid and service robotics company.
Ubtech Robotics
Microsoft Maia 300 compared with Nvidia Rubin on specs
TechRepublic compares Microsoft's Maia 300 accelerator with Nvidia's Rubin on specifications, not benchmarks.
Microsoft
Rivian to bring AI agents to its cars
TechCrunch reports AI agents are coming to Rivian vehicles, with more details expected at the company's 'AI & Autonomy Day' event in December.
Rivian
Mindbeam partners with Nvidia on open-source AI speed
Mindbeam announced an Nvidia collaboration aimed at making open-source AI faster and more efficient.
MindbeamOpen
More: Yahoo Finance
Simons Foundation launches $15M AI math initiative with Anthropic, OpenAI
Why it matters: Major philanthropic funding aims to shape how AI tools are adopted in pure mathematics research.
The Simons Foundation launched a $15 million effort to help mathematicians navigate the AI era, partnering with Anthropic and OpenAI on the initiative.
Simons Foundation
More: citybiz
Germany plans Nvidia AI partnership; Intel expands Korea physical AI
Why it matters: National AI partnerships signal governments embedding Nvidia compute and physical AI into industry.
Germany is set to unveil an AI cooperation deal with Nvidia covering automation and robots, while Intel expanded physical AI partnerships in South Korea and Scale AI partnered with Nvidia on post-training.
Nvidia
Apple's Joswiak says Apple Intelligence targets mainstream users
In a TIME interview, Apple marketing chief Greg Joswiak said Apple Intelligence is built for 'everyday people' rather than the 'lunatic fringe,' and said Apple does not feel threatened by OpenAI's hardware plans.
AppleClosed
More: 9to5Mac
Apache Ranger creators launch AI agent security guide series
The creators of Apache Ranger published a guide series for CISOs on securing AI agents, covering governance, access control and risk management for enterprise deployments.
Economist brief: Ukraine oil strikes resume, AI chip sell-off continues
The Economist's world-in-brief roundup notes Ukraine resuming strikes on Russian oil infrastructure while the AI chip stock selloff continues.
Ramp reaches $60B valuation as revenue tops $2B
Why it matters: Signals investor appetite for AI-embedded fintech at near-decacorn scale.
Fintech Ramp hit a $60B valuation as annualized revenue crossed $2B with 85,000 customers, pushing deeper into AI-powered finance tools.
Ramp
Artificial Analysis releases AA-Robotics v0.1 benchmark
Why it matters: First benchmark ranking LLMs for embodied robotics tasks, exposing slow progress.
Artificial Analysis launched AA-Robotics v0.1, a benchmark ranking LLMs on robotics tasks. Few models are on it; several complete most tasks but slowly, with GPT-6.1 Sol best overall.
Artificial AnalysisFree
