Sakhanda Wire
NVDA $155.46 -0.44% MSFT $506.24 -2.27% GOOGL $280.20 -1.24% META $618.87 -1.20% AMZN $257.89 +0.05%

Latest news

Older stories

Ars Technica

DeepMind introduces Genie 4 world model with more realistic simulations

Google DeepMind has introduced Genie 4, an updated world model designed to generate more useful real-time simulations. The company says the system improves on Genie 3 in almost every respect, including the realism of physics and human behavior. Genie 4’s output can resemble a video game, but DeepMind presents the technology as a component of progress toward artificial general intelligence rather than as a game product.

The Verge

Google announces Genie 4 for more realistic interactive AI worlds

Google has announced Genie 4, a world model designed to generate interactive experiences with artificial intelligence. The company says the new version produces more realistic worlds, supports real-time changes made through natural-language prompts, and can create audio to accompany its simulations. It uses Gemini 3.8 Flash to help run those simulations.

The Decoder

Former DeepMind researcher proposes AlphaGo-inspired AI verification

Thore Graepel, a former DeepMind researcher, has founded Metis Reasoning to develop AI systems that record what they know and test conclusions against evidence. The company presented its approach at Nvidia’s GTC conference in Berlin, drawing on the design of AlphaGo, which combined a neural network suggesting moves with systematic searches through possible game sequences.

The Verge

Apple adds AI assistant to Pages, Keynote, and Numbers

Apple is introducing an AI assistant for its productivity apps Pages, Keynote, and Numbers. Built on models from Apple and OpenAI, the assistant can use a document’s context to help design, animate, and refine presentations in Keynote.

TechCrunch

GMO Internet investigates suspected customer data leak

Japanese internet company GMO Internet is investigating an alleged data breach involving its Onamae Desktop Cloud service, which provides virtualized desktops for financial traders. The company said suspected attackers may have posted information connected to customers and the business, but it is still checking whether the material is genuine and has not determined the incident’s full scope. GMO said as many as 27,000 servers could be affected.

TechCrunch

Opera introduces WebAI for running models locally in the browser

Opera has introduced WebAI, a web standard intended to let users run artificial intelligence models directly in the browser rather than through cloud services. The company says the system can work with different models, including Google’s open-weight Gemma and, on Mac computers, Apple’s on-device foundation model. WebAI is not linked to a single model and is designed to support a broad range of open-weight AI systems.

TechCrunch

General Compute raises $100 million as neocloud concerns grow

General Compute has raised $100 million in Series A funding, led by Blackbird Ventures and Abstract Ventures. The round follows a $100 million annual recurring revenue agreement with a large customer and values the company at $800 million post-money, up from a $60 million valuation after its May seed round.

MarkTechPost

Mirendil releases Fern 0.1 with a 10M-token context window

Mirendil has released Fern 0.1, its first AI model, designed for AI research and development rather than general conversation. The coding model was pretrained from scratch and built from Mirendil-Base after a short post-training phase. Its context window reaches 10M tokens, which Mirendil estimates can hold about 23 MB of code, or roughly 500,000 lines. The model is available through hosted early access; no open weights or quantized builds have been announced.

The Decoder

Analysis finds partner content receives priority in ChatGPT

Content from OpenAI’s licensing partners appears to enter ChatGPT through a separate pipeline that gives it preferential treatment over ordinary web pages, according to an analysis by David Konitzny, a GEO researcher at Berlin-based startup Peec AI.

The Decoder

Anthropic renews pledge to keep ads out of Claude

Anthropic has renewed its commitment not to place advertisements in Claude, outlining the decision in a cover letter accompanying its IPO filing. The company says users bring important work and personal questions to the chatbot, so its primary incentive should remain providing useful answers. Anthropic also argues that advertising-based revenue targets can increase pressure to optimize for engagement, a pressure it does not want to introduce into Claude.

MIT Tech Review

Industrial intelligence links manufacturing knowledge to physical action

Manufacturers hold operational knowledge in documents, production data, procedures, and the experience of their workers. The article describes industrial intelligence as the combination of this accumulated expertise and understanding of how a particular business operates. Making that knowledge available across plants and teams could help engineers, maintenance staff, planners, and operators apply experience more consistently.