- 1Safety
Researcher develops printable patterns that can evade common surveillance detectors
Bill Swearingen says he used millions of automated tests and a reinforcement-learning model to generate visual patterns that prevent many widely used detection algorithms from flagging people and vehicles.
- 2Tools
T3 Code reduces agents' management burden and restores focus
The T3 Code tool, after months of parallel agent use, resolves problems caused by threads, half-finished worktrees and the “who does what?” issue; workflows can be created, tracked, merged and…
- 3Business
Situational Awareness invests $400M more in chip startup Source Foundry
AI-focused hedge fund Situational Awareness committed an additional $400 million this week to Source Foundry, bringing its total investment in the chip-manufacturing startup to $500 million.
- 4Regulation
Fed official warns AI investments may pose systemic financial risk
Jeff Schmid, president of the Federal Reserve Bank of Kansas City, warned that the scale and financing of current artificial intelligence investments could create macroprudential risks and potentially make the sector 'too big to fail.' He pointed to nearly $2,400 billion of AI commitments by large tech firms, interlinked financing structures and rising market indicators such as widened credit-default-swap spreads for major AI players as reasons for increased regulatory attention.
- 5Industry
Data centers' waste heat raises local temperatures and urban risks
Recent field and satellite analyses indicate that data centers can measurably warm their immediate surroundings, with local air and surface temperature increases reaching around 0.7–2.2 °C in specific cases.
- 6Research
Stanford builds a 37,000‑agent 'Virtual Biotech' and validates a drug design later confirmed by Merck
At VB Transform 2026, James Zou of Stanford described a shift from single, powerful AI agents to systems of tens of thousands of specialized agents coordinated by an AI orchestration layer.
- 7Model launches
DeepMind open-sources WeatherNext cyclone model, lowering compute barrier for storm forecasting
Google DeepMind has released WeatherNext — the cyclone forecasting model published in Nature — making its trained weights publicly available and providing a lightweight Colab-ready version.
- 8Safety
Anthropic says Claude models are highly resistant to prompt injection
Anthropic announced that they have successfully trained their models to resist prompt-injection attacks that previously fooled early versions of Claude; similar results were observed in a benchmark…
- 9Tools
Hermes Agent runnable with Vercel AI Gateway and Vercel Sandbox
Hermes Agent is now usable with the Vercel AI Gateway and Vercel Sandbox services: the AI Gateway key can provide access to all models and give full visibility into spending, while the Sandbox runs…
- 10Business
OpenAI integrates NextSlide team to work on ChatGPT
Presentation-startup NextSlide has joined OpenAI, and its team will now contribute to ChatGPT development.
- 11Regulation
Hungary launches government website for AI information and deepfake complaints
The Hungarian government has launched a new website, mihivatal.gov.hu, to answer public questions about artificial intelligence and to accept complaints related to AI and deepfake content, Science and Technology Minister Zoltán Tanács announced.
- 12Industry
Debate over the "compute crisis": are users or engineers to blame?
An online forecast says many mistakenly blame everyday users for the computing capacity crisis, arguing that mothers' three daily prompts costing $0.01 each are responsible, while engineers spend thousands of dollars per day on frontier models.
- 13Research
AgentRadio's asynchronous messaging nearly doubles multi-agent performance on enterprise code tasks
Researchers from Coral AI Labs and partner universities introduced AgentRadio, an asynchronous message-passing layer that gives concurrently running coding agents passive awareness of each other.
- 14Model launches
OpenAI removes chat limits for free ChatGPT users with GPT-5.6 Luna rollout
OpenAI is updating the language models behind ChatGPT: free and Go users will receive GPT-5.6 Luna, which removes the text-chat message limit and adds a "Think" button that engages deeper reasoning.
- 15Safety
Open-weight AI Models Narrow Gap with Closed Systems in Cyber Capabilities
An analysis by the UK’s AI Safety Institute finds the most capable open-weight models trail the best closed models by roughly four to seven months in cyber-related abilities.
