Alphabet loses $500bn as AI search weighs on ad sales
Alphabet's advertising shortfall rattled investors despite higher revenue, while Amazon's negative $42bn cash flow added scrutiny of the cost of the AI race.
Artificial Intelligence 136 stories
Artificial intelligence is approached here as a broad computing field that enables machines to perform tasks involving perception, language, reasoning, generation and decision support. Coverage spans AI systems, models, applications, deployment, limitations, evaluation, safety and the practical consequences of using automated intelligence. The category is wider than machine learning or individual AI tools, keeping the focus on the domain as a whole. Readers get clear explanations of capabilities, trade-offs and where AI fits real decisions and workflows.
Linux TCP Flaw Could Let Remote Attackers Run Commands as Root
Anthropic said its AI model wrote an exploit for a flaw that could allow root command execution on servers with a multipath TCP listener. Ubuntu and RHEL had not released a fixed kernel as of October 29.
iOS 27.2 brings conversation-aware Siri and iPhone Duo Pencil support
Siri can draw on earlier chats and more personal data while iPhone Duo gains Apple Pencil support. The update also changes several apps and fixes vulnerabilities across iOS components.
Apple's $2,000 Foldable iPhone Sells Out at Launch
Apple's first foldable iPhone sold out quickly at some U.S. stores and drew resale premiums in China. Discounts reduce the cost for some buyers, while early reviews and supply constraints raise practical concerns.
Nvidia faces claims AI server demand is squeezing gaming GPUs
A reported $5 billion in early GB300 deliveries has sharpened allegations that Nvidia's AI focus is tightening GeForce supply though the claims remain unproven.
NVIDIA RTX Spark Brings Large-Model AI to Windows ARM PCs
Microsoft's Surface Laptop Ultra combines NVIDIA's RTX Spark SoC with up to 128GB of memory for local large-language-model inference. Planned systems from other manufacturers could end Qualcomm's near-exclusive position in Windows ARM PCs.
CloudKeeper Launches AI Cost Dashboard With 30 Days of Free Access
CloudKeeper's new platform puts AI model and token expenses alongside license and infrastructure costs in one dashboard. Businesses can explore it with 30 days of free access.
Intel's Weak Q4 Outlook Sends Shares Down More Than 20%
Intel beat Q3 earnings estimates but forecast lower Q4 revenue and margins than Wall Street expected. Shares fell more than 20% after hours as worries about PC demand and cash flow overshadowed growth in server CPUs.
AI Models Share a Human Critique After a Provocative Test
ModelCompare grouped 35 of 41 substantive answers around the idea that people may put belonging ahead of truth. The experiment tested responses to one pointed prompt, not human behavior.
Vyoma.AI Begins Construction of 52 MW AI-Ready Data Center in Chennai
The Chennai project is part of L&T's planned gigawatt-scale AI Factory campus, which already includes an operational 30 MW facility and 150 MVA of power infrastructure readiness.
Berkeley Professor Gives Homework Credit for Submission as AI Complicates Grading
Richard Borcherds says AI makes it difficult to tell who solved homework problems. He now awards marks for submitted work regardless of correctness, while UC Berkeley leaves AI rules largely to individual faculty members.
OpenAI and Anthropic Say Their Models Challenge a Basic Computing Assumption
Both labs say their models found algorithms that solve a difficult class of logic puzzles in sub-exponential time. If the proofs hold up, the results would challenge the Exponential Time Hypothesis, a widely used assumption about computational difficulty.
ChatGPT Is Sending Routine Tech Complaints to 404 Media
404 Media says people with account and service problems are arriving with chatbot-written pitches and evidence packets, even when their complaints are ordinary support issues.
ChatGPT memory study puts user privacy under scrutiny
An analysis presented at the ACM Web Conference 2026 found psychological insights in about 52% of the user-related memories examined. The findings also raise concerns about how much control people have over information retained by ChatGPT.
NVIDIA Minerva Targets Lost GPU Time With Topology-Aware AIOps
In one reported fabric incident, NVIDIA Minerva tied 81 noisy events to a single incident and helped cut repair time. The platform pairs topology-aware telemetry analysis with an AI assistant to help operations teams investigate subtle failures.
OpenAI's GPT Store Struggles to Keep Romantic Bots Out
Searches surfaced romantic-companion GPTs despite OpenAI's policy ban, while creators appeared to be using alternative terms to make their bots easier to find.
Justdial Brings Local Business Listings Into ChatGPT
Users can describe a local need in everyday language, then use Justdial's ChatGPT integration to find business details and verified ratings.
Claude Helps Formally Verify 123 Open-Source Libraries
Anthropic and Theorem used Claude to build machine-checked proofs for open-source software, including edge cases missed by random testing. The results show both the promise of automated verification and the steep cost of applying it widely.
Sam Altman Rejects a GPU Monopoly in OpenAI's AI Safety Debate
Sam Altman rejected single-company control of AI computing as a safety model. OpenAI's Pacing 2.0 initiative is intended to establish when systems become too powerful to safely develop or release.
ChatGPT's QNT Scenario Puts $7,000 in View by 2027
ChatGPT sketches a speculative bull-market scenario in which Quant's Overledger becomes the interoperability layer for a multi-CBDC bridge. The model's upper targets reach $4,500, with a stretch above $7,000 by January 1, 2027.
AI Euphoria Meets the Limits of Everyday Use
As AI becomes part of daily life in Germany, surveys show declining enthusiasm alongside continued belief in its potential. Growing energy demands and concern about AI-generated content are testing that optimism.
iOS 27 launches with two basic features still delayed
Apple has released iOS 27 with new Apple Intelligence tools and an upgraded Siri. Suggestions in Messages and AI Password Repair, but, are postponed to a future update.
ChatGPT Rivals Compared on Price, Privacy and Coding
Claude, Gemini, Grok, Mistral Vibe and DeepSeek offer different costs and data-handling options. Gemini 4 Argon's introductory API rate ends November 30, 2026, while DeepSeek's reported US federal restriction adds a compliance check for some organisations.
Uber Extends ChatGPT Work Access Across Its Workforce
The expansion could give employees an AI tool for workplace tasks, but Uber has not shared rollout details or internal usage rules.
GPT-6 Astra Can Use 100 Times More Enterprise Credits Than Luna
OpenAI's enterprise rate card puts Astra at 250 credits per message and Luna at 2.5. Anthropic instead bills enterprise API use by tokens, leaving teams to compare two different cost systems.
ChatGPT Ads Adds 26 Countries as OpenAI Reports 100,000 Advertisers
OpenAI says more than 100,000 advertisers have run campaigns on ChatGPT Ads as the service expands into 26 countries. The company is also working on measurement solutions while exploring additional formats.
ChatGPT Ads Keeps Ad Selection Outside Answer Generation
OpenAI describes a session-aware ad pipeline that can reject a placement after an auction or omit it if the answer is ready first. Its privacy and performance claims have not been independently verified in this report.
OpenAI Expands ChatGPT Ads to 89 Countries
OpenAI is adding ads in 26 markets for Free and Go users as its ad business scales beyond early U.S. testing.
How ChatGPT Turns Word Prediction Into Conversation
ChatGPT learns to respond through language prediction and human feedback. That process can produce helpful conversations, but it can also generate confident errors.
ChatGPT Chrome Extension Adds Cross-Tab Analysis and Browser Actions
OpenAI's Chrome extension can now work across open tabs and complete browser tasks. Its access to signed-in sites also puts sensitive data and prompt-injection risks in focus.
Meta spends 11.1% of Prometheus rack capacity on GPU recovery
After recording 1,000 daily training interruptions per 8,000 GPUs during Prometheus's first deployment phase, Meta says it cut that rate by 99% with a layered recovery system.
llm-d 1.0 Brings Cache-Aware Routing to Distributed AI Inference
Google has released llm-d 1.0 with production features for routing requests to cached context, managing shared accelerator capacity and keeping inference pools available during upgrades.
AI Raises the Value of Real-World Learning
Fred Swaniker says generative AI can make classroom learning more efficient while mentorship and practical experience remain essential to preparing people for work.
Surface Laptop Ultra and RTX Spark get hands-on testing on Windows Central Podcast
Daniel Rubino and Zac Bowden discuss Surface Laptop Ultra battery and display results, then turn to RTX Spark gaming tests and local AI tools.
Ukraine Adds AI Turrets as Russia Shifts to Faster Drones
Ukraine's Air Force reported intercepting 536 of 580 drones during the night of October 14-15. Ukrainian forces have also begun using AI-automated machine-gun turrets against jet-powered drones, but no link between the turrets and U.S. satellite data has been verified.