Responding to a June 14 post by Microsoft CEO Satya Nadella on enterprise AI, a Fast Company contributor argues that a company's lasting advantage lies in its learning loop rather than its foundation model. The author contends that the real challenge for enterprise AI is architectural design that keeps authority separate from the model itself, a concept the writer says their company has developed over three years.
fastcompany.com
· 2026-10-19
Nvidia introduced Nemotron 3.5 Super VL, a 120-billion-parameter open-weight model aimed at enterprise agent deployments, alongside two smaller Bolt models built to run locally on new RTX Spark and DGX Spark hardware. Nvidia says Super VL matches the accuracy of its larger Nemotron 3 Ultra model while completing agentic tasks faster than rival models from Anthropic and OpenAI, and lists companies including Skydance, EY, Ramp, Crowdstrike, Perplexity and Salesforce as users.
techcrunch.com
· 2026-10-19
CERN and French AI company Mistral have signed a statement of intent to form a long-term partnership developing foundation models for scientific research. Their first project is a single pretrained Large Physics Model, trained on billions of simulated proton-proton collisions from the Large Hadron Collider, intended to handle multiple analysis tasks such as particle reconstruction, identification and anomaly detection.
home.cern
· 2026-10-19
Meta says the first phase of its gigawatt-scale Prometheus supercluster in New Albany, Ohio is complete and was used to train its Muse family of foundational models, including the new Muse Wonder models. The company plans to scale the cluster toward roughly one million GPUs and says it uses closed-loop liquid cooling, having achieved over 85% goodput in pre-training and over 90% in reinforcement learning runs for Muse Wonder.
engineering.fb.com
· 2026-10-19
Doubl has unveiled Maxwell, described as a new kind of model that operates between source code and the hardware executing it, aiming to let software adapt during execution rather than run purely as fixed, deterministic instructions. The announcement frames traditional computing as splitting all 'thinking' into the programming stage, with execution remaining mechanical, and positions Maxwell as a way to bring AI-driven adjustment into the execution process itself.
heymaxwell.ai
· 2026-10-19
Anthropic has circulated a confidential draft AI safety bill to lawmakers, obtained by Politico, that would let the government block release of a new AI model only if it threatens at least $1 billion in damage or 50 or more deaths or serious injuries, and only when that risk stems from bioweapons, loss of developer control, or cyberattacks. The bill would route such decisions through independent evaluators reporting to the Commerce Department within 72 hours, with a federal judge in Washington, D.C. ultimately deciding whether to halt release, and no other federal agency allowed to intervene.
techspot.com
· 2026-10-19
DeepSeek has released V4.1 Flash 1017 (Max), a multimodal model accepting text and image input and producing text output, with a 1 million token context window. The model scored 45 on the Artificial Analysis Intelligence Index, well above the comparable median of 19, while running at 230 tokens per second.
artificialanalysis.ai
· 2026-10-19
Nature has released an audio long read discussing concerns that heavy reliance on large language models could weaken people's critical thinking abilities. The episode draws on research from education and cognitive science specialists who study how reasoning skills develop and can be maintained. It outlines practical strategies researchers suggest for staying cognitively sharp while using AI tools.
nature.com
· 2026-10-19
Researchers writing in Nature contend that large language models, despite appearing intelligent, are not conscious and are unlikely to become so under current development approaches. The piece addresses claims from figures such as Richard Dawkins, Geoffrey Hinton, and Anthropic's research on its Claude model suggesting computational signatures of consciousness.
nature.com
· 2026-10-19
A blog post examines recent discussion around Jev, a family of AI models accessed via OpenRouter, focusing on its 'decision model' outputs. It describes three output types—choice, score, and noul—and demonstrates them with a sample API call to Jev 1.13 rating review sentiment on a 'Lukewarm' to 'Thrilled' scale.
pncnmnp.github.io
· 2026-10-18
In a guest essay on Terence Tao's blog, mathematician Francesco Maggi argues that large language models have become genuinely powerful tools for generating mathematical proofs, citing examples like an LLM-assisted construction of a finite-time blow-up solution to the Navier-Stokes equations and a proposed proof related to the Riemann Hypothesis. He outlines two possible futures for science: a 'golden age' where AI empowers broader human expertise, or a 'dark age' where output surges but knowledge becomes concentrated among few people and increasingly incomprehensible to humans.
terrytao.wordpress.com
· 2026-10-18
Speaking on the Sources podcast, OpenAI CEO Sam Altman confirmed the company will build a humanoid robot, arguing that a human-like form makes sense since the world's doors, tools and keyboards are designed for people. He said the bigger challenge is building the 'brain' that controls the robot, not its physical design, and told Fortune a demo could arrive in 2027 with wider deployment years further off.
engadget.com
· 2026-10-17
GoKawiil