AI
Google launches Gemini 4 Argon for cyber defense as Deepseek ships Huawei chip tools
Google unveils its most powerful model yet, limited to cyber partners, while Deepseek and Huawei push an open-source challenge to Nvidia's CUDA.
This edition was produced with artificial intelligence. Text and voice are generated automatically.
Google launches Gemini 4 Argon, its most powerful model, for defensive cyber work
Alphabet has launched Gemini 4 Argon, an AI model built for coding, research, and writing, with a focus on cybersecurity. Argon is being rolled out only to a select group of the company’s cyber partners through its Fairwind Program. Google says the model was trained specifically for defensive cyber work and can autonomously find, validate, and patch critical software vulnerabilities. Staff have already used Argon for debugging and codebase migrations, and the company cites its ability to analyze long videos and charts. Google claims Argon scored significantly higher than OpenAI’s GPT-6 Astra and Anthropic’s Fable and Opus models across various AI benchmarks, citing the startup Vals to show Argon leads its AI model index.
Google releases Gemini 4 Argon, called its most powerful model yet →
Anthropic AI model proves long-sought percolation theory conjecture
Anthropic released a proof of a long-standing percolation theory conjecture generated by a large language model. Percolation theory, which emerged in 1957, studies how coherent structure emerges from random local connections and hinges on a threshold pc marking the transition between limited and infinite connectivity. Mathematicians had long sought to characterize this transition in dimensions 3 through 10, a problem considered a holy grail of the field. The AI model showed the transition is continuous there, confirming what mathematicians had suspected for decades. Benedikt Jahnel of the Technical University of Braunschweig said the result evoked ambivalent feelings, citing joy that the conjecture was proven alongside disillusionment that the final crucial step came from an AI.
AI solves a ‘holy grail’ problem from probability theory →
OpenAI and Synopsys partner to build GPT-Synopsys for chip design
Synopsys and OpenAI announced a strategic partnership and signed a multi-year agreement to jointly develop GPT-Synopsys, a specialized model optimized to use Synopsys electronic design automation tools for semiconductor design workflows. The companies will collaborate as preferred partners on R&D and joint go-to-market initiatives under a shared revenue framework, with OpenAI licensing Synopsys’ EDA tools to develop the model. GPT-Synopsys will be available to customers worldwide and is intended to make frontier models expert users of EDA tools, delegating design objectives from power, performance and area optimization to timing and verification closure. Customer data is not used to train the model and remains encrypted. Early technology engagements are underway with leading semiconductor customers.
OpenAI and Synopsys Announce GPT-Synopsys: Frontier Intelligence to Revolutionize Chip Design →
Google rolls out Skills in Gemini, replacing Gems globally
Google is rolling out Skills globally in Gemini chat. Skills are detailed prompts for specific tasks that the AI can help write and refine, replacing Gems, Google’s equivalent of OpenAI’s Custom GPTs, which OpenAI is also sunsetting. The format originated with Anthropic and is available as an open standard. Users save frequent instructions as a Skill and invoke them by typing a slash plus the Skill name. Gemini can also generate Skills from previous chats and run them automatically when it detects a matching prompt, and multiple Skills can be chained for bigger tasks. Skills now support text documents, PDFs, and images. Gems shut down in November for personal accounts, March 2027 for enterprise and nonprofit Workspace customers, and June 2027 for education customers.
Deepseek and Huawei release open-source tools for Ascend AI chips
Deepseek has partnered with Huawei to release open-source programming tools for Huawei’s Ascend AI chips, announced via Deepseek’s official WeChat channel. The release includes libraries for computation and for moving data between chips, with Huawei fully supporting the work. The centerpiece is TileLang, an open-source programming language for AI chips originally developed by researchers at Peking University and used by Deepseek for about a year. Deepseek argues that building an independent software ecosystem for AI chips requires a universal language that is easy to program while still extracting full hardware performance, and says TileLang offers a simpler programming model than Nvidia’s CUDA. The two companies also optimized a supernode cluster of 128 Ascend 950 chips.