OpenAI Claims to Have Solved Navier-Stokes Math Problem
OpenAI says an internal model has solved the Navier-Stokes problem, one of mathematics’ seven Millennium Prize Problems. However, the historic breakthrough is being overshadowed by concerns over whether researchers can safely use frontier labs’ tools to work on unpublished discoveries.
- OpenAI says an internal model, more powerful than GPT-6 Astra, worked with 10,000 concurrent AI agents to produce a solution to the roughly 90 year old Navier-Stokes problem in just 88 hours.
- The Navier-Stokes problem carries a $1 million Millennium Prize, though the company said it does not plan to claim it.
- NYU math professor Tristan Buckmaster and Anthropic researcher Levent Alpöge had been looking into the Navier-Stokes problem since last year and had used a mix of LLMs, including OpenAI’s Codex.
- As they prepared to publish, Buckmaster reached out to OpenAI and learned that the company had produced a Navier-Stokes proof using a route he and Alpöge had been working on with Codex. That led him to ask whether the model had been trained on, or had access to, their Codex sessions.
- OpenAI says no specific user data was accessed, but acknowledged that it cannot completely rule out de-identified data from product usage contributing to model improvements.
OpenAI researcher Sebastien Bubeck said the two proofs differ significantly and that the company had not seen the researchers’ work before it was made public. However, the controversy raises thorny questions about who owns work completed with AI tools.
OpenAI Launches ChatGPT Images 2.5 With Faster Generation and Better Editing
OpenAI has released ChatGPT Images 2.5, an upgraded image generation model focused on sharper visuals, faster generation, and more reliable editing.
- Images 2.5 improves lighting, textures, and reference-image consistency while following editing instructions more accurately across multiple turns.
- Generation is now up to 50% faster than Images 2.0, reducing the time needed to create and refine images.
- A new Sketch feature lets users draw directly in ChatGPT and use their sketches as references for image generation.
- OpenAI is also adding templates for common image formats and prompt sharing so users can reuse and experiment with image-generation ideas.
- Developers get two API versions: GPT-Image-2.5 Flare for faster, high-volume generation and GPT-Image-2.5 Sunburst for workflows requiring more precise editing and control.
Images 2.5 is rolling out across ChatGPT, ChatGPT Work, and Codex on desktop, mobile, and web, with access across all tiers.
Suno launches new AI music model trained on licensed data
The company has launched its v6 model family, saying it was trained on licensed music from partners including Warner Music Group, BMG and Believe. The move comes as the company faces multiple copyright lawsuits and prepares to retire its older models.
Instacart introduces AI assistant for smarter grocery shopping
Instacart has launched Clementine, an AI assistant that can turn conversations, recipes, grocery lists, photos and screenshots into ready to buy carts. It can also personalize recommendations based on dietary preferences, suggest cheaper alternatives and surface relevant deals.
Anthropic researcher Jacob Coxon resigns, warning the company is racing toward superintelligence that “could kill us all by the end of the decade.”
His comments were echoed by other researchers, like Anthropic safety lead Evan Hubinger, who said he personally estimates there is a greater than 10% chance AI could “kill all humans” within the next decade.
Gemini’s Daily Brief is now free for US users
Google has made Gemini’s Daily Brief available to everyone in the US, removing the previous requirement for a paid Google AI subscription. The feature creates a daily summary using information from connected apps, including emails and meetings.
Apple finally built a foldable, but who is it actually for?
Apple just validated foldables after sitting on the sidelines for years, but charging $2,000 means they are testing how far brand loyalty stretches in a shaky device market.
- It opens like a passport with a seamless hinge mechanism and drops the physical SIM tray completely in favor of eSIM-only setup worldwide.
- Under the hood sits the same A20 Pro chip driving the Pro lineup, paired with an under-display FaceTime camera and full Apple Pencil compatibility.
- The device splits its battery across both sides to feed the massive main screen, hitting a 50% charge in just 15 minutes.
After watching Samsung and Google refine flexible screens over multiple generations, Apple waited until the hardware matured enough to fix screen creasing before launching its own take.
The entry price effectively puts high-end multitasking into luxury luxury territory, forcing competitors to either lower prices or match Apple’s ultra-premium hardware build.
If you build iOS apps, now is the time to start optimizing layouts for split-screen dual-display workflows before October shipping dates.
Anthropic insider quits, warning AI could end humanity by 2030
Hollywood is turning real-time Silicon Valley board disputes into prestige cinema before the ink on the news reports even dries.
- The Hollywood Shift: Neon released the teaser for Artificial, a Luca Guadagnino film starring Andrew Garfield as Sam Altman, focusing on the 5 chaotic days in late 2023 when Altman was fired and reinstated.
- The Conflict of Interest Drop: Amazon MGM originally backed the film but dropped it right around closing a $50B partnership with OpenAI, forcing indie distributor Neon to take the wheel.
- Wild Ensemble Cast: Beyond Garfield, the cast includes Yura Borisov as Ilya Sutskever, Ike Barinholtz as Elon Musk, Monica Barbaro as Mira Murati, and Mark Rylance as Geoffrey Hinton.
Written by SNL alum Simon Rich, the film spans roughly two decades of machine-learning history culminating in the November 2023 boardroom coup. It premieres at the New York Film Festival in October before hitting theaters nationwide on Christmas Day.
Keep an eye out for how tech giants navigate media representation as corporate drama goes pop culture mainstream.
OpenAI drops Images 2.5, cutting image generation latency by 50%
By rolling out in-chat canvas drawing alongside a faster image engine, OpenAI is admitting that text prompting alone is a terrible way to control visual spatial composition.
- Doodle to generate by invoking @Sketch inside the chat window, giving you a raw canvas where a rough layout guide directs the final rendering logic.
- Latency drops 50% under the new Images 2.5 engine compared to the previous release, alongside much tighter subject consistency across multi-turn edit requests.
- Developers get dual API options split between a high-speed default called Flare for fast volume and a precise model named Sunburst built specifically for targeted inpainting edits.
OpenAI launched this update across all ChatGPT tiers and Codex environments as competitive pressure mounts from design tools integrating native generative engines.
The addition of canvas-first inputs shifts prompt engineering away from descriptive prose toward direct spatial layout references.
If you build AI workflows, test the new Sunburst endpoint for multi-step editing pipelines where keeping reference subjects identical across background changes has historically failed.
AI could end humanity by 2030
Three Anthropic researchers warned that advanced AI could cause human extinction within years, with former researcher Jacob Coxon quitting after accusing Anthropic and OpenAI of racing toward superintelligence without adequate safeguards.
Anthropic alignment lead Evan Hubinger said he personally sees a greater than 10% extinction risk within the next decade, while Samuel Marks said senior AI employees often express stronger concerns privately.
The warnings land as frontier models gain stronger cyber capabilities and autonomy, making this debate harder to dismiss as science fiction when the people building the systems keep raising the alarm.
Meta launches Muse personal AI agent
Meta launched Muse, a personal AI agent that can send emails, book travel, sell a car, and handle other digital chores through its app, website, or WhatsApp.
Muse runs inside isolated cloud virtual machines, while Meta’s Sentinel system monitors activity; Stripe Link creates single-use card numbers for purchases, keeping users’ real payment details out of transactions.
Meta is giving Muse plenty of surfaces to live on: iOS, Android, web, WhatsApp, and eventually AI glasses. The catch is that heavy users will need a paid AI plan.
AI just cracked a $1M math problem
OpenAI published a proof from an unreleased internal model that it says solves the Navier-Stokes problem, one of mathematics’ seven Millennium Prize challenges, each carrying a $1 million award.
OpenAI says roughly 10,000 agents worked for 88 hours using a model significantly stronger than GPT-6 Astra, costing millions in compute, while mathematician Tristan Buckmaster disputes how independently it reached the result.
The credit fight may prove almost as consequential as the proof itself: Buckmaster says OpenAI knew about his team’s work, while OpenAI says it never accessed their specific research.
Hackers are stealing Claude tokens
Claude users have reported unexplained token usage after hackers stole active login sessions, letting attackers generate unauthorized tokens and rack up charges even while account owners remained inactive.
Anthropic says infostealer malware can capture Claude session keys, and the company has invalidated affected sessions and refunded some users, though customers still lack detailed tools for tracing unauthorized usage.
One consultant said Anthropic suspended his account during the investigation and only partially refunded him, pushing him to rival AI platforms with clearer usage tracking and account controls.
Apple launches iPhone 18 Pro and Pro Max
Apple unveiled the iPhone 18 Pro and Pro Max with its new 2nm A20 Pro chip, upgraded cooling, longer battery life, and four finishes including deep black and burgundy.
The A20 Pro brings a six-core CPU and seven-core GPU, which Apple says delivers 40% better graphics performance, while a larger vapor chamber helps sustain performance under heavier workloads.
Apple also added a 48-megapixel variable-aperture camera, pro manual controls, and up to 45 hours of video playback, giving the Pro Max another year of very literal “Pro” upgrades.
OpenAI says an AI cracked a Millennium Prize problem in 88 hours
- OpenAI announced that one of its AI systems solved a Millennium Prize problem — one of the seven unsolved mathematical problems designated by the Clay Mathematics Institute, each carrying a $1 million prize — in 88 hours. (cite index=”19-1″>OpenAI says an AI cracked a Millennium Prize math problem in 88 hours, then walked straight into a fight over credit.
- The Millennium Prize Problems are the hardest open problems in mathematics — the Riemann Hypothesis, P vs NP, the Navier-Stokes Existence and Smoothness problem, and four others. A genuine solution to any of them would be one of the most significant mathematical events of the 21st century.
- The “fight over credit” is the context that matters most. The August Astra proofs controversy — where 30 mathematicians said OpenAI’s results pasted arguments from 2016 and 2019 papers without attribution — established a pattern.
- The broader context: 2026 Fields Medalist Jacob Tsimerman announced the launch of the Mathematical AI Safety Institute (MAISI), a research org modeled on Princeton’s IAS that hosts mathematicians for semester-long AI-safety stints.
If the Millennium Prize solution holds, it is the most consequential scientific result produced by an AI system in history — surpassing the protein binder results, the Riemann-Zeta improvement, and every benchmark score ever reported. If it does not hold under mathematical scrutiny, it is the most damaging AI credibility event since the Astra plagiarism controversy.