Wednesday, September 9, 2026 - 5 Minute AI News

0:00OpenAI claims they just solved a 90-year-old math problem, but a prominent mathematician says they stole his work to do it. We have the details on the allegations. Welcome to 5 Minute AI News for Wednesday, September 9th, 2026. We are also looking at Meta's new autonomous agent that wants to handle your bills, plus a notable shift in chip manufacturing standards. Starting with that math breakthrough, OpenAI announced they generated a solution and a formal lean proof for the Navier Stokes existence and smoothness problem. This is a fluid dynamics puzzle that has stumped mathematicians for 90 years. If the proof holds up, it claims a $1 million millennium prize bounty. Solving it is a clear milestone for AI reasoning, but the achievement is completely mired in controversy today. NYU mathematician Tristan Buckmaster alleges OpenAI stole the unusual solution path. He says he uploaded his own AI-assisted drafts to OpenAI's Codex, which the company then used to claim the breakthrough. The allegations go further than just scraping data from the web. Buckmaster claims an OpenAI researcher actually threatened him,

1:03pressuring him to drop a co-author who happens to be employed by rival and thropic. OpenAI's official blog post is celebrating the formal proof, but the broader tech press is heavily focused on these allegations of intellectual property theft and extortion. It highlights the increasingly cut-throat competition at the very edge of research. Speaking of intense competition, Meta just unveiled its most aggressive consumer AI play to date. It is an economist personally AI agent called Muse. Unlike a standard chatbot that waits for your prompt, Muse operates in the background. It is designed to take independent actions like sending emails, booking travel, and even negotiating your bills. It only returns to you when it needs explicit approval. To handle all of this, Muse runs on a dedicated, secure cloud computer for each individual user. Meta says this architecture is necessary to keep user data isolated while the agent performs complex tasks. It requires broad access to personal data, including your health services and financial payments. That is the real test here. Moving from a passive chatbot to an active, goal-oriented agent is a significant technical step,

2:03but it's successful depend entirely on whether consumers actually trust Meta with deep, unmonitored access to their private communications and bank accounts. Shifting over to the hardware that makes these agents possible, there is a rare unified agreement among rival chip makers. ASML has successfully convinced top manufacturers, including TSMC, Samsung, and Intel, to transition to larger photo masks. This is for ASML's newest extreme ultraviolet or EUV lithography machines. These machines cost around $400 million each, and this new foundational manufacturing standard will boost their productivity and throughput by 40%. Getting competitors to agree on a hardware standard is difficult, but this change will directly increase global advanced chip production capacity. It accelerates the hardware pipeline required to train and run the next generation of increasingly compute heavy AI models. Meanwhile, Huawei is reportedly working with the Chinese equipment maker Yuleiung Shang to bypass reliance on the Dutch company's technology entirely. It shows how fractured the global supply chain remains outside of this new agreement. Back on the software side, OpenAI has rolled out chat GPT images 2.5,

3:03and it includes a new capability they are calling Sketch. The feature allows users to draw rough doodles directly within the chat GPT interface. You then use text prompts to guide the AI and transforming your crude sketch into a detailed polished image. It essentially bridges the gap between text to image and image to image generation natively within their flagship interface. You can also integrate your own reference photos to generate highly personalized results right in the chat. By making spatial and visual control more accessible to everyday users, they are competing much more directly with specialized design software. It gives users a way to dictate composition and layout without needing to describe every single spatial relationship in a long text prompt. Finally, looking at the business of AI, the coding assistant startup Cognition has reached a $48 billion valuation. This valuation multiple actually surpasses what their competitor cursor achieved before its acquisition by SpaceX. Its signals continued investor confidence in the coding assistant space. A valuation at that economic scale illustrates how venture capital views the AI software engineering market right now. They clearly see it as expansive enough to support multiple independent giants.

4:03It pushes back against the idea of a winner take-up scenario dominated by a single player or a big tech incumbent. There is still plenty of capital flowing into specialized vertical applications, especially those that directly increase developer productivity. We will certainly be keeping a close eye on those Navier Stokes theft allegations to see if open AI officially responds to the NYU researchers claims. That is all for today. Please remember to like, subscribe, and leave us a review on Spotify and Apple podcasts. It really helps the show. Thanks for listening and have a great day.