OpenAI says AI solved one of maths’ biggest problems
OpenAI is claiming its unreleased AI model just solved Navier-Stokes, the 3rd most famous unsolved math problem in the world that’s been open for ~90 years. It’s one of the 7 Millennium Prize Problems, with a $1 million bounty.
OpenAI threw ~10,000 AI agents at the problem at once, letting them work in groups, share ideas, run code and search a cached version of the internet. They found the solution after ~88 hours, then GPT-6 Astra spent another 17 hours formalizing and verifying it in Lean.
There’s some drama though. A mathematician who spent a year working on a related Navier-Stokes result (also with AI) says OpenAI got access to his work, raced to finish the bigger problem and then pressured him to drop his co-author, who works at Anthropic.
And a pretty important caveat: the mathematician’s related result has been verified, but OpenAI hasn’t released its proof of the actual Millennium Prize Problem yet. So for now, we mostly have to take their word for it.
AI researchers are starting to quit over ‘out-of-control’ AI fears
Jacob Coxon resigned from Anthropic after three years doing pretraining research at OpenAI and Anthropic, saying neither company is acting responsibly and they’re “racing straight to self-improving superintelligence and gambling with our lives.”
His argument is basically that the people building these systems know how serious this could get. He says researchers privately talk about AI potentially killing everyone by the end of the decade, but the race keeps going anyway.
And other researchers are backing him up. Anthropic’s Alignment Science Lead Evan Hubinger said “we really do earnestly believe AI could kill all humans”, putting his own odds at 10%+ in the next decade. Another researcher who left Google DeepMind in June said Jacob is right and that thinking about how to stop this was “literally my day job.”
Coxon is now calling for labs to slow down and coordinate, potentially even temporarily stopping improvements to model capabilities.
Pretty wild that “should we actually keep building this?” is becoming a real debate among the people building it.
OpenAI agents crack a 90-year-old math problem in 88 hours
A 90-year-old math problem just got cracked. Not by humans. By AI.
The Navier-Stokes equations describe how fluids move, think water swirling down a drain. The big open question: can those equations ever completely break down? Mathematicians had no answer for nearly a century.
OpenAI deployed 10,000 AI agents on an unreleased model and solved it in 88 hours, proving the equations can break down under extreme conditions.
The model found a vortex that tightens and spins ever faster, while the fluid’s energy stays bounded throughout. Think of fluid stretching like spaghetti until the math itself explodes.
Here is what makes this technically credible:
- The result was formally checked in Lean, giving mathematicians confidence it is correct.
- Agents could read from a cached internet version, run code, and communicate within subgroups.
- The model used is significantly more capable than GPT-6 Astra.
Multi-agent systems tackling unsolved science is now real.
OpenAI ships ChatGPT Images 2.5 with faster generation, sketch input, and two new API models
OpenAI just dropped ChatGPT Images 2.5, and it is a solid upgrade to how you generate and edit images, both in the app and through the API.
The big headline: Flare delivers higher-quality images than GPT-Image-2 at 50% lower latency. That matters a lot if you are building anything image-heavy.
Here is what changed and what it unlocks for you:
- Better reference photo handling: subjects look more recognizable, lighting feels more natural, and distinctive features carry through edits.
- Precise local edits: change a jacket while keeping the subject in place, update a background while preserving the product, or revise text without reworking the whole composition.
- Sketch lets you draw directly in ChatGPT. Just type @ Sketch to show it exactly what you have in mind.
On the API side, you get two models. Use Flare for fast, high-quality everyday generation, and Sunburst when editing precision matters most. Both are priced at 2x the GPT-Image-2 rate.
Available now across desktop, mobile, and web.
Mistral raises €3B Series D, the biggest equity round in European tech history
Mistral just pulled off something big. The French AI company raised €3 billion (~$3.5B) in a Series D, the largest equity round a European tech company has ever closed, at a €21 billion valuation. That valuation nearly doubled from their previous round just a year ago.
So what does Mistral actually do, and why should you care? They build AI models you can download and run yourself, no middleman, no usage fees, no one else seeing your data. That is the core pitch.
- The money goes toward bigger model research, more compute power, and expanding globally
- They are building 1 gigawatt of compute capacity in Europe by 2030, basically their own AI data center empire
- They now let customers pick which region processes their AI queries, so data stays local
- Over 125 enterprise customers already, including Airbus and HSBC
The big idea: you get frontier AI performance without being locked into one provider’s cloud. You run it, you own it.
OpenAI just claimed a blockbuster math breakthrough. Some academics are crying foul
Navier what? OpenAI claimed yesterday that an internal model, described as significantly more capable than GPT-6 Astra, has ‘solved’ a version of the Navier-Stokes equations, one of the seven Millennium Prize Problems, and widely considered to be one of the hardest problems in mathematics.
Why this matters: The Navier-Stokes equations aren’t just an abstract curiosity. They describe how fluids move, which means they quietly underpin huge parts of modern life. Engineers use them to design jet engines and model airflow. Doctors rely on them to understand blood flowing through arteries.
The controversy surrounding the claims
- The announcement has landed amid a serious misconduct claim that has put OpenAI on the defensive.
- Two academics — NYU’s Tristan Buckmaster and Anthropic’s Levent Alpöge — claim they spent a year working on a similar problem, using an unusually rare approach.
- Buckmaster claims they’d uploaded their drafts into Codex, and that OpenAI, which had access to those logs, then produced a proof using the same unusual approach.
- Bubeck has called the allegations “false and inflammatory,” and maintains the proof was reached independently. Sam Altman has also backed Bubeck.
The dispute is feeding a broader anxiety: that major AI labs exploit access to users’ private work and absorb their ideas, and this case is quickly becoming Exhibit A for academics who share that concern.
Meta unveils “world’s first personal AI agent”
The Instagram maker just dropped Muse, an AI agent app that connects to your email, calendars, shopping, payments, and smart home devices to carry out tasks on your behalf. Muse has its own computer that works in the background, so it can do things like book movie tickets or manage your inbox. However, some users have privacy concerns. See Muse in action here or try it out here.
OpenAI unveils ChatGPT Images 2.5 and a clever new way to prompt it
The upgrade promises sharper detail, more natural lighting, and up to 50% faster generation than Images 2.0 (see the difference between the two models here). What caught users’ attention, however, is a new feature called ‘Sketch,’ which allows you to doodle directly inside ChatGPT, then have the model turn that rough drawing into a polished image. OpenAI is betting that users will find visual prompting far more hands-on than typing. You can see ‘Sketch’ in action here.
Anthropic safety researcher resigns, sparks concerns over AI risks
Evan Hubinger, who leads alignment science at Anthropic, has unleashed a firestorm on social media, saying that he believes there’s a greater than 10% chance AI could “kill all humans” within the next decade. The comments came after researcher Jacob Coxon publicly quit the company over similar fears. The resignation post (29 million views), amplified by Hubinger’s backing, has reignited the debate about the pace of AI model progress.
The Money: AI coding tools command record multiples as VCs bet on multi-winner market
Two headline rounds this week signal investors see AI-assisted development as a category with room for multiple champions, not a single platform winner. Cognition’s $48B valuation at 53x revenue multiple, higher than Cursor commanded before its SpaceX exit, suggests the market is pricing in sustained enterprise demand rather than consolidation.
Deals to know:
- Cognition (Series C, $2B at $48B valuation) — AI coding assistant Devin grew ARR from $492M to $900M in four months. Enterprise clients include Mercedes-Benz, NASA, Goldman Sachs, Citi. Investors: Andreessen Horowitz, Accel, Founders Fund, General Catalyst, Avenir
- Mistral AI (Series D, €3B at €21B valuation) — Europe’s largest tech equity raise ever. Building 1 GW compute capacity in Europe by 2030; sovereign AI positioning attracts government and enterprise clients across 20 countries. Investors: Samsung Electronics (lead), EQT Scaleup Europe Fund, PSG Equity, a16z, Nvidia, BlackRock
Signal: Capital is flowing to companies solving developer velocity and data sovereignty, the two constraints enterprises cite most when deploying AI internally.
Chrome shifts to 2-week updates as AI accelerates security threats
Google tightens patch cycles to close the gap between known vulnerabilities and deployed fixes as AI-powered attacks evolve faster.
OpenAI claims its AI solved 90-year-old math problem
Company deployed 10,000 parallel agents over 88 hours to crack the Navier-Stokes Millennium Prize Problem, costing millions in compute. Won’t claim the $1M prize.
NSA accuses six Chinese firms of extracting billions of tokens from Claude, GPT, Gemini
DeepSeek, Alibaba, Moonshot AI and three others allegedly ran industrial-scale distillation campaigns using fraudulent accounts and proxy services since late 2024.
Anthropic faces expanded lawsuit over Claude Max usage caps
Developers paying $200/month for “20x more usage” still hit undisclosed weekly limits, class-action complaint alleges, raising questions about how clearly AI subscriptions must define compute restrictions.
Meta debuts Muse AI agent
Meta has launched Muse, a personal AI assistant integrated across WhatsApp, Instagram, and Facebook that can book travel, send emails, make payments, and access third-party apps on behalf of users. CEO Mark Zuckerberg positioned it as Meta’s answer to competing AI agents from OpenAI and Anthropic.
- Muse operates directly within Meta’s messaging platforms, allowing users to complete transactions and manage tasks without leaving WhatsApp or Instagram, with planned expansion to Meta’s smart glasses.
- The assistant handles payments and accesses personal data to execute tasks autonomously, raising immediate questions about security and data handling across Meta’s 3.5 billion monthly users.
- Meta is offering a free tier with paid plans at $20 and $100 per month for higher usage levels, signaling a direct challenge to OpenAI’s ChatGPT and Anthropic’s Claude in the agentic AI market.
For enterprise teams, Muse signals that AI agents capable of autonomous transactions are now reaching mainstream consumer platforms. This changes the competitive landscape for B2B tools built on similar capabilities. Security and compliance leaders should also note that employees using personal devices may now have AI agents with payment and app access running in their messaging apps, creating new vectors for data exposure.
The Genome Search Box
DeepMind released AlphaGenome Atlas, an AI precomputed map of all 9 billion possible single-letter mutations in the human genome. Each mutation gets 27,000 predictions: gene expression, RNA splicing, chromatin folding, 3D structure. The dataset fills a petabyte. Anyone can query it in a browser. Years ago AlphaFold mapped every protein’s shape. Now Atlas maps every mutation’s footprint.
The Human Genome Project took 13 years and $2.7 billion just to read the letters. DeepMind’s AI model ran almost every possible mutation, turned the results into a map, and put it behind a browser tab. A researcher who once needed a GPU cluster, a pipeline, and a bioinformatics engineer now needs a question. The 98% of the genome that was invisible is now searchable.
DeepMind did not solve the genome. It solved the query, and the bottleneck just moved from price to purpose.
Math Problems Are Now an Ad
OpenAI announced that an internal AI system solved the Navier-Stokes existence and smoothness problem, a Millennium Prize problem open since the 19th century. Roughly 10,000 agents consumed 88 hours and 300 billion tokens, then completed a Lean formal proof. The model behind it exceeds GPT-6 Astra. This follows ARC-AGI-3 at 99.9%, ExploitBench at 100%, and an Erdős conjecture solved earlier this year. The press release writes itself. So does the next one.
The pattern is now a product cycle. Benchmark falls, paper ships, headline lands, next benchmark is already in training. The Millennium Prize problems are seven. OpenAI has started treating them like a backlog. Mathematics took a century to build these walls. OpenAI’s agents are clearing them in workweeks, and the announcements arrive with the cadence of a changelog. The question is no longer whether AI can solve a famous problem. It is whether anyone still reads the announcement.
OpenAI solved a century-old problem, again, and the math community is running out of surprises.
Europe Bows to Mistral
Mistral raised 3 billion euros, the largest equity round in European tech history, at a valuation of 21 billion euros. Samsung led. ASML led the last one. The three-year-old French company serves 125 enterprises and is approaching a billion dollars in annual revenue. Mistral’s models have never made much noise. Europe paid the premium anyway.
The buyers are not investing in a model. They are investing in a border. European banks, defense contractors, and governments do not want their data on American servers, and Mistral is the only serious alternative. The company does not need to outscore OpenAI. It needs to be European enough. On that test, it has no competition. Samsung and ASML are not buying equity. They are buying insurance, and the premium just hit 3 billion euros.
Europe cannot build its own OpenAI. So it paid 21 billion euros for the next best thing: an AI that is not American.
DeepSeek Is Being Undercut
DeepSeek cut V4-Flash prices by up to 60%, with off-peak rates returning to pre-August-hike levels. The cut targets Flash only. Pro stays unchanged. At the same time, DeepSeek opened a beta for V4.1 Flash, a cheaper model with native multimodality. DeepSeek made its name as the cheapest frontier model. The cheap ones have multiplied.
DeepSeek launched this price war. But now it just stopped winning it. GLM 5.3 Flash has pulled even with Flash on price, and undercuts Pro by three to five times, for nearly the same performance. So DeepSeek’s only move is to cut again, and hope its high cache-hit ratio makes the discount feel real. The company that broke into the global market by undercutting everyone is now being undercut, and the lever it pulled to get here is almost the only one left.
DeepSeek built its empire on being cheaper. The cheaper ones have arrived, and all that is left is to cut again.
Mistral Closes Europe’s Biggest Round
Mistral AI closed a €3 billion Series D led by Samsung, pushing its valuation past €21 billion — Europe’s largest tech funding round ever. The open-source champion is positioning itself as the continent’s answer to US frontier labs, with fresh capital to scale compute and enterprise distribution.
ChatGPT Images 2.5 Lands
OpenAI released ChatGPT Images 2.5 with significantly improved detail sharpness and more precise editing capabilities. The update, the second since the April image model debut, continues OpenAI’s rapid iteration on multimodal generation as competition from Midjourney and Google Imagen intensifies.
OpenAI Claims Navier-Stokes Breakthrough
OpenAI announced its AI solved the 90-year-old Navier-Stokes existence and smoothness problem, one of seven Millennium Prize challenges. The claim was immediately challenged by NYU mathematician Tristan Buckmaster, who alleges OpenAI used his research logs from a year of Codex usage to preempt his own near-solution. The controversy raises serious questions about AI companies accessing user data for competitive scientific advantage.
Qualcomm Lands $4B AI Chip Deal with Amazon
Qualcomm secured a multi-generation AI chip and optical networking contract with Amazon worth $4 billion. Amazon simultaneously received warrants to acquire $4 billion of Qualcomm stock, creating a deep strategic lock-in. The deal sent AI infrastructure stocks broadly higher and positions Qualcomm as a serious challenger in the custom cloud AI silicon market alongside Broadcom and Marvell.
Astra Rolls Out to All Paid Tiers
OpenAI fully deployed Astra to Plus, Pro, Business, and Enterprise users across Codex and ChatGPT Work. The rollout marks the end of staged access and the beginning of universal AI-assisted coding for OpenAI’s entire paid user base, significantly raising the bar for competitors like Cursor and GitHub Copilot.
OpenAI’s Next Model Already Crushes GPT-6 Astra
OpenAI revealed an internal model that is ‘significantly more capable than GPT-6 Astra.’ Training began on August 28 and the model surpassed Astra across the board in just seven days. The disclosure signals an accelerating internal capability curve and puts pressure on competitors racing to close the gap.
Meta Launches Muse, a Privacy-First Personal AI Agent
Meta released Muse, a personal AI agent designed to compete with OpenAI Operator. Muse emphasizes on-device privacy and ships as a standalone iPhone app. Built on Meta’s Llama models, it handles sensitive personal data locally, positioning privacy as a differentiator in the increasingly crowded AI agent market.
Mistral Closes €3B Funding at $24B Valuation
Paris-based Mistral AI closed a €3 billion funding round led by Samsung, reaching a $24 billion valuation in Europe’s largest-ever tech funding. The open-source champion’s massive raise signals that investors still see a viable alternative path to closed-source dominance, despite Mistral lagging behind OpenAI and Anthropic on benchmarks.
Washington Buys Into Quantum
The U.S. government acquired minority equity stakes in three quantum computing companies: D-Wave, Rigetti, and Quantinuum. The move signals a new federal strategy of direct capital participation in strategic quantum infrastructure, mirroring approaches seen in defense tech and semiconductor policy.
IonQ Launches First Volume-Production Quantum Computer
IonQ launched Superion 256, a quantum computer purpose-built for volume production. The milestone represents a critical step in moving quantum computing from laboratory demonstrations to commercially deployable systems, with implications for drug discovery, materials science, and optimization workloads.
Bluecore Energy Raises $50M Seed in Two Months
Bluecore Energy, a nuclear technology startup, raised a $50 million seed round just two months after incorporation. The rapid raise reflects surging investor appetite for next-generation nuclear technology, driven by AI data center energy demands and growing policy support for advanced reactors.
Robot Simulation Startup Antioch Raises $32M
Antioch Inc., a simulation software company for robotics testing, raised $32 million in new funding. Its platform enables robot developers to validate and iterate in virtual environments rather than physical labs, cutting development time and cost as demand for robotics testing infrastructure grows alongside the embodied AI boom.
Intel Plans 10% CPU Price Hike
Intel is reportedly planning a 10% CPU price increase ahead of a major product launch in March 2027, with AMD expected to follow between June and July. The move suggests the semiconductor industry is entering a new pricing cycle, with implications for data center operators and enterprise buyers already navigating AI infrastructure costs.