Ad
Skip to content

Matthias Bastian

Matthias is the co-founder and publisher of THE DECODER, exploring how AI is fundamentally changing the relationship between humans and computers.
Read full article about: Clay Mathematics Institute says the Navier-Stokes Millennium Prize Problem has "apparently been settled"

The Clay Mathematics Institute (CMI) has officially weighed in on the Navier-Stokes problem, one of seven Millennium Prize Problems announced in Paris in 2000, each worth $1 million. The problem asks whether the equations governing fluid motion in three-dimensional space always have a smooth, complete solution.

The CMI says "the Navier-Stokes problem has apparently been settled" and hopes "to see waves of new human understanding unleashed as the innovations behind this work are analyzed and interrogated." The solution is now under review. "The process is deliberately unhurried, but we will provide updates," the institute writes. It also notes that the "increasing ability of new technologies to accelerate mathematical research has heightened this sense of anticipation."

The potential solution comes with a bitter dispute. Mathematician Tristan Buckmaster accuses OpenAI of redirecting resources toward the problem after rumors about his research leaked, using his drafts in training data, and rejecting his co-author, Levent Alpöge, who works at Anthropic, from authorship.

Comment Source: CMI
Read full article about: Sam Altman calls for pacing AI development but promises rapid progress will continue

Sam Altman is doubling down on his call to pump the brakes on AI, at least a little. He wants a uniform, nationwide safety framework for advanced AI but says the industry shouldn't wait for lawmakers, even though international coordination needs U.S. government support. To that end, OpenAI now sets explicit safety protocols before training runs that could bring major capability jumps, and Altman hopes other companies "will learn from our approaches and propose their own." He also makes clear that "pacing" doesn't mean "stopping." Progress will "continue to be" rapid, just not as fast as it could.

Pacing will be well worth this cost; no amount of American competitive pressure should justify recklessness, or let capabilities get ahead of alignment and monitoring.

Sam Altman, OpenAI

Over the weekend, Anthropic CEO Dario Amodei, former Deepmind CEO Demis Hassabis, Microsoft CEO Satya Nadella, and Elon Musk all backed better-controlled AI development. According to The Information, OpenAI, Anthropic, and Google have been talking for months about self-regulation through an independent oversight body. Smaller players like Cohere worry this could turn into a cartel.

Read full article about: Elevenlabs makes Music v2.5 available via app and API with free and pro tier options

Elevenlabs has released Music v2.5 for ElevenMusic. The company says generated songs now sound fuller and more natural. In a blind test with 47,885 comparison pairs, listeners preferred v2.5 most of the time, especially for R&B, Soul, Hip-Hop, Rock, and orchestral music. Users retain the rights to their tracks. The free tier includes five lossless downloads per day and the Pro tier 400 per month.

Commercial use is allowed depending on industry and purpose, though the free tier requires attribution. Downloads of tracks based on other artists' songs are blocked, and imitating existing musicians isn't allowed. Music v2.5 is also available through Elevenlabs' API, while v2 remains accessible.

Elevenlabs recently signed a licensing deal with Universal Music Group, but says it only applies to future, separate products, not Music v2.5. The company says it trained its existing Music models on "licensed stems and music" without detailing the data. That still sets it apart from competitor Suno, which was sued for training on copyrighted content without rights holders' consent. Suno also released new models this week.

Read full article about: Altman, Musk, and Hassabis back Amodei's call to add independent oversight

OpenAI CEO Sam Altman and Elon Musk are backing calls to slow down AI development. Altman, Musk, and Ex-Deepmind-CEO Demis Hassabis at least partly endorsed Anthropic CEO Dario Amodei's proposals, all agreeing on the need for independent oversight inside AI labs. Musk has been sounding the alarm on AI risks for years, so his support isn't exactly a surprise.

Altman and Amodei agree: AI labs should be evaluated independently. | Image: via X

Altman also told Fortune that OpenAI won't go public this year, citing safety concerns. He had already shared that decision internally in June. Another reading: OpenAI's financials simply don't support an IPO right now, especially compared to Anthropic's. Both could be true.

Not everyone agrees. Google researcher Peyman Milanfar pushed back, arguing that recursive self-improvement (RSI) is a naive assumption because the feedback loops are inherently unstable. The harder a system optimizes itself, the deeper it falls into blind spots, and reliable evidence only comes from the real world, not benchmarks. That makes Amodei's "speed limit" unnecessary, Milanfar argues: "Systems that reliably improve themselves will be governed, damped, slow, and bounded by margins that look wasteful. Because stability is the speed limit."

Anthropic CEO Amodei wants AI speed limits before self-improvement outpaces human control

Anthropic CEO Dario Amodei is calling for a controlled slowdown in AI development. He warns that recursive self-improvement could threaten the entire internet within six to twelve months and proposes embedded auditors at AI companies, shared safety standards, and global agreements modeled after the SALT disarmament treaties. His warning comes just ahead of what could be the largest initial public offering in history.

Read full article about: GPT-6 Astra appears to show a "step change" in spatial reasoning based on early benchmarks

GPT-6 Astra appears to be a big leap forward for spatial reasoning. A new robotics benchmark called StationeryBench pits OpenAI's GPT-6 Astra against Ai2's MolmoAct2 across five desk-object tasks like uncapping a marker, pouring out paper clips, or passing a ruler between two robot arms. Both models controlled the same dual-arm YAM robots across 200 trials. Astra fully completed 7 out of 100 tasks; MolmoAct2 completed zero. Astra's median progress score hit 46 out of 100, MolmoAct2 managed 12. All results, videos, and code are on GitHub. OpenAI has long-term plans to build its own consumer robots.

Yoav Artzi, an AI researcher at Cornell and Google DeepMind, calls Astra a "step change in spatial reasoning." On the still-unpublished REMAP benchmark, GPT-Astra reaches accuracy close to human level, though Artzi notes that "even ASTRA doesn't get to what humans do in other scenarios." He suspects OpenAI trained the model on large amounts of 3D data such as Blender scenes. That lines up with Astra's particular improvement on 3D tasks.

Read full article about: Nvidia wants to pour up to $10 billion into Anthropic's record-breaking IPO

Nvidia is in talks to invest up to $10 billion in Anthropic's planned IPO, Reuters reports. The company behind Claude wants to raise up to $100 billion and land a valuation of around $2 trillion. That would make it the largest IPO in history. Nvidia would come in as an anchor investor, locking in shares before the stock hits the open market.

The two companies are already closely linked. Anthropic runs on Nvidia GPUs and committed in 2025 to buying $30 billion in Azure compute packed with Nvidia chips. Revenue went from around $9 billion at the end of 2025 to over $65 billion by July 2026. The IPO is expected to wrap up before the US midterms in November.

Nvidia is still playing central bank for the AI industry, investing in customers like Anthropic and OpenAI while backing some $300 billion in guarantees that help data centers secure financing. Most of that money circles right back to Nvidia in chip orders.

GPT-6 Astra needs leaner prompts and fewer guardrails, OpenAI recommends

Overly long skill descriptions, blanket reading requirements, and rigid approval rules can get in GPT-6 Astra’s way, warns OpenAI’s Eric Provencher. More capable models need less hand-holding, so developers should tie instructions to specific tasks and spell out when the job is done.