Google announced a unified Gemini agent that can plan goals and complete tasks using company systems and tools, rather than merely answering instructions; it will launch first for businesses. The company said it can select a suitable model by default, users can also choose Anthropic’s Claude models, and open-source models will be added later.
Google said it will offer its Gemini agent to businesses first; the agent can plan goals and act within company systems.
02INDUSTRY
OpenAI’s annualized revenue nears $50 billion
The Financial Times reported that OpenAI told investors its annualized revenue was nearing $50 billion. A previously discussed figure of $70 billion was based on investors’ effort to compare it directly with Anthropic. But the companies calculate revenue differently: Anthropic includes sales by cloud partners, while OpenAI does not.
According to the Financial Times, OpenAI told investors its annualized revenue was approaching $50 billion.
03POLICY & LAW
USA Today and local papers sue OpenAI over copyright
In the lawsuit, the publisher argues that the use of its content is causing ongoing harm to its organizations. Newspapers including The Tennessean and the Indy Star are also part of the suit; OpenAI reportedly did not immediately respond to a request for comment.
USA Today Co. and affiliated local newspapers are seeking more than $250 million from OpenAI, alleging that hundreds of thousands of articles were copied without permission for training.
04SECURITY
Nvidia expands Halos safety system to robots
Nvidia expanded its Halos safety system, launched for autonomous vehicles in 2025, into robotics in June 2026. It offers hardware and software tools, including processors dedicated to safety workloads, such as IGX Thor; the source provides no measurements of the system’s safety outcomes.
Nvidia is bringing its Halos safety architecture, developed for autonomous vehicles, to robots, aiming to support the safe use of warehouse, factory, and surgical robots.
05INDUSTRY
$1.8 billion committed to AI biology data
Biohub, the U.S. Department of Energy (DOE), NIH, and partners announced a coordinated $1.8 billion commitment to create an open data resource for AI-powered biology models. DOE will invest more than $500 million over five years; NIH will coordinate datasets developed through earlier federal investments. Google DeepMind, Isomorphic Labs, and Meta are allocating $300 million to the initiative.
Biohub, government agencies, and technology companies launched a coordinated effort to create an open data resource for advancing biology models.
06MODELS
StepFun’s Step 5 Preview appears on OpenRouter
StepFun says it developed Step 5 Preview for tool-using tasks that refine results over multiple steps. The model uses a sparse Mixture-of-Experts architecture and is presented as having 600 billion total parameters, with 27 billion active; it is said to be strong in finance.
StepFun’s Step 5 Preview model is designed for multistep tasks involving large codebases and documents.
07MODELS
Whistle: a 16.9 MB on-device speech recognition model
Whistle is a 16.9 MB speech recognition model released for mobile and embedded devices. It runs on a CPU without dependencies and supports seven languages. In the published comparison, Whistle leads on some tests, while Whisper base performs better on the average across TED-LIUM, AMI, and MLS.
Whistle is a 16.9 MB model released to transcribe speech on-device in seven languages.
08MODELS
Writer touts DeepSeek 4.1 Flash as a low-cost coding model
The writer says that, after using DeepSeek 4.1 Flash for a month on about a dozen projects, it produced work comparable to Opus, while noting this assessment is based on personal experience. They say the expected cost per session rarely exceeds $1 and that the KV cache is about 437 times smaller than V1’s.
A writer argues that DeepSeek 4.1 Flash performs close to expensive leading models on coding tasks, and that its low cost changes how they work.
✓
That’s it for today.The rest of your time is yours.
Your daily technology digest, every morning in English. Free. Unsubscribe anytime.
Sources, in the open.
Stories are selected from Hacker News; accessible original articles are summarized and translated with AI. Summaries may contain errors; check the source beneath each story for details. Preparation starts daily at 06:55 Türkiye time (UTC+3), targeting publication before 07:00.
Stories are read aloud by an AI voice.
This edition covers articles shared on Hacker News on the date shown above.
Now on the App Store for iPhone
Your daily digest. Now in your pocket.
Read or listen to a new edition every morning with Kısa for iPhone, save stories for later and keep your streak.
