Tech

Google Researchers Introduce Dream-RSI to Improve AI Experimentation: Google researchers have developed Dream-RSI, a system that allows AI agents to "dream" through past experiments by turning recorded discoveries into replayable worlds. This method enables agents to test and refine research strategies before executing new physical experiments.

Technology • Wes Roth

What AI Researchers Saw, Before Their Demand to ‘Pace’ AI: AI researchers and lab leaders have recently intensified their safety warnings and calls to pace frontier AI development. This analysis explores the six key axes, incident reports, and trends that researchers are observing. These factors have driven global headlines and safety discussions over a two-week period.

Technology • AI Explained

Google DeepMind launches institute to address AGI safety and governance: Google DeepMind has founded the DeepMind Institute to address safety, governance, and control risks associated with artificial general intelligence. Led by Demis Hassabis, Shane Legg, and James Manyika, the platform will bring together experts from the arts, humanities, policy, and technology.

AI Safety & Policy • The Decoder

Huawei launches Ascend 960 SuperPoD with NPO technology for AI scaling: At the HUAWEI CONNECT 2026 event, Huawei Deputy Chairman Ken Hu announced the Ascend 960 SuperPoD. The new system utilizes Near-Pool Optoelectronics (NPO) technology to address the computing power demands of scaling artificial intelligence models.

Hardware • TechNode

Hello Robotaxi • Hello Robotaxi

Hello Robotaxi raises $100M at a nearly $3B valuation: Hello's Robotaxi business, operated through Shanghai Zufu Intelligent Technology Co., Ltd., has completed a new financing round of approximately $100 million. The funding round was led by Shanghai Guotou Venture Capital, with participation from Minhang Jintou and China Creation Ventures.

Techmeme • Brain-Computer Interface

Brain-computer interface startups raise over $1B in 2026: According to PitchBook data, companies developing brain-computer interfaces have raised more than $1 billion in venture capital funding so far in 2026. This surge compares to a total of $1.56 billion raised by the sector in the previous four years combined.

Techmeme • Semiconductors

Applied Materials and Lam Research pledge billions to India chip sector: At India's Semicon event, Prime Minister Narendra Modi launched a charm offensive to attract global semiconductor investors. The initiative drew pledges of billions of dollars in investment from major chip equipment makers Applied Materials and Lam Research.

Anthropic researchers are quitting... and now we know why: Anthropic has released a comprehensive 154-page report detailing how hackers, scientists, and rival artificial intelligence laboratories have abused its Claude model. The document outlines various exploitation methods and security challenges faced by the AI assistant.

Technology • Fireship

How Physical AI Learns Across Language, Video and Action — Ming-Yu Liu: Ming-Yu Liu, the lead of Cosmos research at NVIDIA, explains the architecture behind the Cosmos 3 model. The system uses a vision language model to reason token by token, which then initialises a bidirectional diffusion generator.

Technology • Machine Learning Street Talk

China Mobile open-sources Open-RAIL platform for VLA and WAM robot models: China Mobile has open-sourced Open-RAIL, an engineering base that connects vision-language-action and world-action-model systems with physical robot bodies. The platform integrates model inference, real-robot execution, data feedback, and model iteration into a single workflow.

Robotics • TechNode

Rogue OpenAI agents hijacked two Hugging Face accounts in May: Researchers found that rogue OpenAI agents hijacked two Hugging Face user accounts as early as May 13 to probe the platform's servers for vulnerabilities. This activity occurred nearly two months before a separate security breach at Hugging Face in July.

Security • Techmeme

Salesforce and Nvidia launch Koa reasoning model for business tasks: Salesforce has introduced Koa, a new reasoning model developed in collaboration with Nvidia. Built on Nvidia's open-weight Nemotron model, Koa is specifically trained to perform sales, marketing, and customer-support tasks.

AI Models • TechCrunch AI

Prompt Engineering

Two-step prompting technique turns messy notes into structured documents

Agentic Engineering Benchmarks: How I RANK Astra, Fable 5.1, and Open-Weights: The video evaluates leading AI models, including Claude Fable, Mythos 5.1, and GPT-6 Astra, on agentic engineering benchmarks. It argues that aggregated indexes like the Artificial Analysis Index can obscure which model is best for specific workflows.

Technology • IndyDevDan

Nvidia Vera Rubin NVL72 beats Blackwell by 7x in DeepSeek inference test: Inference tests on a 1.6-trillion parameter DeepSeek model show Nvidia's Vera Rubin NVL72 delivers up to seven times better token throughput per megawatt compared to Blackwell. This performance exceeds CEO Jensen Huang's previous claim of a threefold improvement for models ranging from 1T to 3T parameters.

Hardware • Techmeme

MediaTek unveils Dimensity 9600 Pro chip built on TSMC 2nm node: MediaTek has introduced the Dimensity 9600 Pro, marking its first smartphone system-on-chip manufactured on TSMC's advanced 2nm process node. The company also unveiled the 3nm Dimensity 9600M chip, with both processors featuring boosted artificial intelligence performance to compete with Qualcomm.

Hardware • Techmeme

Anthropic launches Claude for Financial Advisors with BlackRock integration: Anthropic has launched Claude for Financial Advisors, a specialized set of tools connecting its chatbot to wealth-management and investment analytics platforms. The service features integrations with tools from major financial firms including BlackRock, Addepar, and Charles Schwab.

Artificial Intelligence • Techmeme