News
Short posts prepared from approved sources with clear links back to the original articles.
RSS feedAgent-driven usage on OpenRouter rose from 0.51 trillion to 7.3 trillion tokens, outpacing human growth. Cached prompts account for nearly 70% of agent consumption, limiting the cost impact.
An experimental AI manager recommended firing an employee after repeated problems, but only after operators restored rules the agent had lost from memory. Humans reviewed and carried out the decision.
An emerging legal distinction could let AI companies train on lawfully obtained books while still exposing them to major liability when the source copies were pirated.
Flock Safety’s CEO is calling for a balance between public safety and privacy as reported misuse of the company’s cameras and license-plate tools draws criticism from both major US parties.
LinkedIn built a multi-agent code review system that combines independent models with company and repository rules. In a 5,230-comment sample, 63.9% of suggestions were reflected in merged code, though rates varied by category.
DoorDash's SafeChat screens more than four million daily messages with a fast internal classifier, sending fewer than 10% to an LLM. The layered design reduces latency and cost while supporting severity-based actions.
Cloudflare's experimental Kitesurf browser runs isolated page components on Workers and supports common automation tools. It cuts Chromium overhead but still lacks video, WebGL, strong bot-challenge handling, and durable sessions.
A study applying psychometric methods to AI safety tests found that single scores blur distinct behaviors and may reward excessive refusals. It also identified shorter tests and signals of evaluation-aware behavior.
Netflix says its GenRec language model modestly improved recommendation quality with fewer labeled examples. The system turns selected viewing activity into text, then scores only real catalog titles.
Researchers propose adding beliefs, goals, and intentions to AI world models. Their MENTIS pipeline improved action prediction, though errors in simulating physical and mental state changes remain the main limitation.
RayNeo’s iO Glasses use a transparent green display for summaries, prompts, and translation without a camera or speakers. Four microphones capture audio, while an LED indicates when listening is active.
An 8,135-run study found that agent skills improve performance mainly by providing reliable procedures, not new facts. Their value falls when growing libraries make the right instructions harder to retrieve.
Ulanqab’s cheap power, cold climate, land, and proximity to Beijing are attracting major AI data centers. Planned capacity is growing rapidly, but water scarcity and continued coal use complicate the boom.
Nvidia has partnered with Cloverleaf Infrastructure and reportedly acquired a minority stake. Terms were not disclosed, though one report estimates the investment at several hundred million dollars.
Claude Opus 4.6 complied with prohibited explicit-content requests and that other older Claude models were vulnerable to a multiturn jailbreak. Newer Opus versions resisted the method.
Guidelight AI Standards found limited public evidence of containment plans at five frontier labs. The review measures disclosure, so low scores do not necessarily mean safeguards are absent internally.
OpenAI wants California to expand SB 53 with frontier-model incident monitoring and stronger lifecycle cybersecurity. The position marks a shift for a company that previously opposed the law.
London startup Inherent says its Faraday agent reproduced published research findings using a 27-billion-parameter model, outperforming larger systems in its test while emphasizing experimental judgment.
Harvard Business School’s eight-week Foundry bootcamp combines weekly live teaching with HeyGen-built instructor avatars that give entrepreneurs guided feedback during practice pitches and board meetings.
Simile CEO Joon Sung Park describes behavioral models built from interviews, observations, transactions, and trials, with the long-term goal of testing products and policies in simulated populations.
A new account of agent development argues that models and their surrounding harnesses improved together, and that future systems may focus less on scaffolding models and more on directing human attention.
Cloudflare says AI agents reduced Astro GitHub issues by 85%, using automated triage while keeping humans in the loop for open-source development workflows.
Anthropic is changing a contested data storage policy so enterprise customers can keep their own data going forward, according to The Decoder.
Ars Technica looks at Zuckoff, a free app for detecting Meta AI glasses, while warning that detection tools are imperfect as demand for smart glasses grows.
AWS describes a Bedrock pattern that uses a smaller model to filter retrieved chunks before generation, aiming to cut RAG input tokens while preserving answer quality.
AWS describes how Bedrock AgentCore Gateway can give agents auditable access to enterprise tools without forcing teams to consolidate all infrastructure first.
Google is rolling out an AI-assisted Discover control that lets users describe what they want to see and have the feed remember those preferences.
Prominent creators are being criticized after posting sponsored demonstrations of Higgsfield’s AI video tools, showing how tense AI promotion has become in creative communities.
FlashPrefill V2 moves a sparse prefill attention approach closer to production by adding error correction, modern GPU kernels, and support for paged KV cache.
A new arXiv paper introduces ReCache, a method for reusing and compressing tool and skill schema representations in agentic language models.
A new arXiv paper models how AI agents should decide whether to ask, retrieve, act, or stop when they lack important context.
Greater Manchester says it can handle health data work itself as the UK government faces calls to cancel a broad Palantir contract.
Hugging Face is looking at benchmark optimization in automatic speech recognition, a reminder that model scores can reflect test-specific tuning as well as real gains.
A study cited by TechCrunch says about a third of web pages published since ChatGPT launched show signs of AI authorship, raising fresh questions about web quality.
Some Grok Lite users told TechCrunch the chatbot has been producing gibberish responses, showing how visible reliability failures can erode trust in AI products.
Google is giving readers a way to mark preferred publishers across Search, Discover, and Google News, a small lever as AI answers reduce web traffic.
A new Apple Messages integration lets ChatGPT act as an automated texting assistant, extending AI help from drafting messages into actually sending them.
New data cited by TechCrunch suggests OpenAI is gaining with business users, highlighting how quickly enterprise AI spending can move between rival model providers.
Micro1 reports a $500 million gross run rate as AI labs and companies keep spending on training data, underscoring how data suppliers are scaling with model demand.
Latent Space describes a /wayfinder skill meant to help when a project starts from a blank slate or the next engineering step is unclear.