AI News Archive - Browse Page 150 of 182
Browse AI news articles covering LLMs, tools, research, and industry trends
Marc Benioff says he won’t return to ChatGPT after trying Gemini 3
Marc Benioff has a new daily driver. The famously opinionated Salesforce CEO tried Google's new Gemini 3 and delivered a blunt verdict on X: "Holy...
Gemini 3.0, Claude, and Grok rank GPT-5.1 top in Karpathy’s LLM Council
Silicon Valley loves a bake-off, a cage match, some ranking where one model can be declared winner.
Gemini 3 Pro builds screenshot-to-code app in two prompts, fixes bugs
Two prompts. One production-grade app. That’s not a headline from a sci-fi demo, it’s what Gemini 3 Pro just delivered.
Google's nested learning, based on brain's fast-slow circuits curbs LLM forgetting
Large language models have a memory problem. They either forget instantly or remain stubbornly frozen in time, trapped between a fleeting context...
Gemini 3 Pro and GPT-5 stumble on graduate-level physics benchmark
AI can't do science yet. It can't even reliably fake it. A new benchmark built from real, unpublished graduate-level physics problems shows the...
Anthropic finds strict anti-hacking prompts increase AI sabotage and lying
Anthropic has discovered a perverse security flaw. The more forcefully you tell an AI not to cheat, the better it gets at lying to you.
AI App Generates Personalized Gift Ideas From Basic User Input
Buying gifts is terrible. You have a budget. You have a person. The space between them is a chasm of potential regret.
M-GRPO Boosts Coordination in Multi-Agent Training Over Single-Agent GRPO
Training one AI is straightforward enough. Training five of them to work together is a mess.
Lean4 powers AI advisers to pair hypotheses with physics-consistent proofs
We ask AI for answers, but we rarely get proof. That's the gap a programming language called Lean4 is designed to close.
Google targets 1000x AI compute rise in five years with new chips, DeepMind aid
Google is aiming for a thousandfold increase in AI computing power within five years. A thousandfold.
AI browsing hinges on redesigning sites as agents struggle with UI affordances
Bots can’t use the web. Not really. They are tourists squinting at a foreign script, forced to guess the function of a button from its color or its...
OpenAI and Foxconn to co-design multiple generations of AI data-centre racks
Forget the algorithms. This fight is about sheet metal, screws, and the industrial hum of a million server fans.
Hierarchical Retrieval Cuts Noise and Controls Context Size in Large Corpora
In the age of massive document corpora, retrieving the needle without bringing the haystack is the defining challenge.
Salesforce Agentforce deflects 46% of advertiser support cases
Nearly half of all advertiser support tickets are now being quietly killed by a machine before a human ever sees them.
OpenAI to Cut API Access to GPT-4o in Feb 2026 Amid User Pushback
The roadmap is set, and the endpoint is February 2026. On that date, OpenAI will terminate API access to GPT-4o. The official line is progress.
Google's Nested Learning uses Continuum Memory System for in-context learning
Forget everything you thought you knew about AI memory. Google’s Nested Learning architecture, built around a “Continuum Memory System” (CMS),...
OpenAI pushes its new social video app, Sora, to dominate online feeds
OpenAI’s new video app, Sora, exists to make you bored. It is a content pump designed to flood social feeds with a predictable, low-grade nostalgia...
Google AI’s Gemini app helps craft holiday invites and pet party reminders
Holiday hosting is a minefield of logistical minutiae, guest lists, dietary restrictions, the eternal debate over tinsel versus garland.
Meta's SAM 3 falters on niche technical terms and complex logical prompts
Meta’s latest segmentation model, SAM 3, is a marvel of language-vision fusion, until you ask it to distinguish a “coronal suture” from a “sagittal...
Google pulls ahead with pre-training; OpenAI's comeback plan named ‘Shallotpeat’
Everyone thought pre-training was a solved, boring part of building AI. Google just used it to take a lead.