By the TensorMax editorial team
· Drawing from sources across the AI industry
Today's top story
regulatory action
42 leading mathematicians, including Fields Medal winners Martin Hairer, Peter Scholze, and Wendelin Werner, have signed an open letter to the president of the Royal Society, warning of the existential risks posed by advanced AI.
Why it matters. The strategic stake of this warning from 42 leading mathematicians is that it underscores the urgent need for attention to AI existential risk, with estimates from AI labs suggesting a risk above ten percent, which the mathematicians argue should not be dismissed as hype, and with leading models from OpenAI and Anthropic having already solved open research problems, including one of the seven Millennium Problems, in just three months.
42 leading mathematicians, including Fields Medal winners Martin Hairer, Peter Scholze, and Wendelin Werner, have signed an open letter to the president of the Royal Society, warning of the existential risks posed by advanced AI. The letter highlights the rapid progress of AI models, which have solved open research problems in mathematics, performing at the level of top human mathematicians in many areas. The mathematicians note that these models may develop superhuman abilities in a short timeframe, and express concern that this rate of change could be similar in areas such as cybersecurity, autonomous weapon control, and targeted misinformation. The letter emphasizes that estimates of existential risk above ten percent from AI labs should not be dismissed, and that action is needed before the situation becomes too late to address. The signatories, none of whom are affiliated with AI companies, urge the British Academy of Sciences to alert the government and the media to these risks. With the rapid advancement of AI capabilities, the mathematicians warn that the window for action may be short, and that it is crucial to address these risks before they become insurmountable.
More from today
model release
Why it matters. The release of Astra for Law by OpenAI marks a significant strategic stake in the legal technology market, with the potential to improve legal research by up to 40% compared to existing models. According to the company's testing, Astra for Law passed the evaluation's overall correctness check on 54.0% of questions, outperforming GPT-6 Astra using web search alone. This development is crucial for law firms and legal technology companies, as it enables them to build AI products and workflows around their expertise, with settings, tools, and context tailored for professional legal work.
model release
Why it matters. SpaceX's Falcon 9 rocket has reached a significant milestone with its 700th launch, carrying satellites for SES, highlighting the company's dominance in the launch industry. With 687 Falcon 9 rockets and 13 Falcon Heavy rockets launched to date, SpaceX has demonstrated its capabilities and reliability. This achievement is particularly notable given that the company has told multiple prospective customers that it is no longer accepting new orders for commercial Falcon 9 launches, with some customers turning to alternative providers like Avio, which has seen an influx of new customers as a result.
model release
Why it matters. The release of Ternary Bonsai 2 27B model is a significant development in the AI industry, offering near-lossless compression and a 9x smaller footprint. With a total model footprint of 5.9GB, this model brings stronger reasoning, coding, vision, and agentic capability to the Bonsai series. The use of ternary weights and FP16 group-wise scaling enables 1.76 effective bits per weight, making it a notable achievement in model compression.
model release
Why it matters. The release of Qwen3.8-Omni-Flash, a native omnimodal model, marks a significant milestone in AI development, with the model supporting a 1M-token context window and achieving audio-visual performance close to Gemini 3.8 Flash. This advancement has the potential to revolutionize various applications, including those that require multimodal inputs such as text, image, audio, and video. With its impressive capabilities, Qwen3.8-Omni-Flash is poised to make a substantial impact on the AI industry, particularly in areas where omnimodal understanding is crucial.
model release
Why it matters. The strategic stake of Anthropic's Claude leading 26% of the company's AI R&D work is significant, as it indicates a substantial increase from 1% in March. This shift underscores the growing importance of AI in driving AI development, with Claude collaborating on over 90% of R&D work. The rapid progress of Claude's capabilities, with its share of model R&D tasks rising from 1% to 26% in just six months, has major implications for the future of AI research and development.
research paper
Why it matters. The strategic stake of AI BioDesign, led by David Baker, the 2024 Nobel Prize in Chemistry winner, is to create new molecules and biological functions that are not found in nature but are physically and chemically possible. With the potential to develop new drugs to treat cancer and neurodegenerative diseases or enzymes capable of breaking down plastics in the ocean, this project aims to create databases, models, and tools that could serve as seeds for developing the medicines and technologies of the future. The initiative, which involves the Allen Institute, the University of Washington, and the Fred Hutchinson Cancer Center, has the potential to transform the field of biology, with some leaders comparing its potential to historic transformations such as industrialization, electrification, and the digital revolution, and could lead to breakthroughs in as little as weeks, not decades, according to Baker.
Catch up quick
-
OpenAI is close to solving the Hodge Conjecture, a Millennium Prize problem
-
Seagate's 2026 Data Infrastructure Readiness Report finds 62% of organizations are unprepared to handle AI-driven storage demands
-
GPT-6 Astra conquers Factorio: Space Age in 2 days
-
Meta agrees to $18bn settlement with US state governments to resolve claims that Facebook and Instagram harmed children
-
Crusoe, an AI data center builder, valued at $30.9B in $3.9B round
-
Crusoe raises $3.9B in Series F round, valuing the company at $30.9 billion
-
Anthropic and OpenAI file confidential paperwork for initial public offerings, with valuations near or beyond a trillion dollars each
-
Crusoe signs $13bn cloud deal with Jane Street
-
Researchers used Anthropic models to breach OpenAI's security
-
Geoffrey Hinton, the 'Godfather of AI', warns the US government has about a year to regulate AI before it gets 'out of control'
-
Microsoft executives admit that AI training is 'the largest theft of labor in human history', and that publishers face an 'existential threat' from AI-powered products.
-
Open-weight models take 56% of token volume, up from 7% in December, with average token cost falling 23.2% in August
-
China's sixth-generation fighter jet may be equipped with laser weapons to defend against incoming missiles
-
China may be building a mothership for large uncrewed submarines at a Shanghai shipyard
-
Huawei expects a major shift to its AI model training infrastructure in China in 2027, with its Ascend chips gaining ground on Nvidia
Also on the desk
The Sean Ellis test.Hit reply: if TensorMax disappeared tomorrow, would you miss it? If yes, what specifically?
|
|