OpenAI solved a $1M maths problem

Hola Decoder😎

If someone forwarded this to you and you want to Decode the power of AI and be limitless, then subscribe now and Join Decode alongside 30k+ code-breakers untangling AI.

🧮 OpenAI says an AI model solved a $1m math problem

OpenAI published a proof from an unreleased internal model that it claims settles the Navier-Stokes problem, one of mathematics’ seven Millennium Prize Problems, while questions are emerging over whether outside researchers’ work contributed to the result.

The Shift:

10,000 agents worked for 88 hours - OpenAI says agents running on a model “significantly more capable” than GPT-6 Astra produced the proof, with the parallel effort reportedly costing millions of dollars in compute.

The claim could solve a decades-old problem - A validated proof would settle the Navier-Stokes existence and smoothness problem, carrying a $1 million Millennium Prize. Sam Altman called the result one of his most amazing moments at OpenAI.

Researchers were already pursuing a similar approach - NYU’s Tristan Buckmaster and Anthropic’s Levent Alpöge spent roughly a year on a related route, feeding drafts into Codex and publishing partial results shortly before OpenAI’s announcement.

Questions remain over training overlap - Buckmaster says OpenAI began after learning about their work and questioned whether his Codex drafts contributed. OpenAI says it never saw their work directly, but cannot rule out broader usage data improving its models.

Beyond whether the proof survives mathematical scrutiny, the dispute raises a new question for AI-driven science: when models help produce major discoveries, establishing where their underlying ideas came from could become almost as important as verifying the result itself.

Together with Levanta

Your Affiliate Roster Should Make You More Money

A roster with hundreds of creators looks healthy until you check the sales.

If fewer than 10% have converted in 90 days and no new partners joined last month, the program is quietly losing revenue every week.

Levanta shows you what is stalling growth and which partners can move products, so more budget reaches relationships producing measurable returns.

• REVO generated $1M+ in affiliate-driven sales, with 96% coming from new-to-brand customers.
• Grace & Stella added $350K+ in affiliate revenue while its Amazon sales rank climbed from #520 to #28.
• Acquco grew monthly affiliate sales from $1,500 to $36,000 in four months after adding 130+ creator partnerships.

You get a growing roster and clear visibility into which relationships drive sales. More converting external traffic can also help your products climb Amazon’s rankings.

Levanta’s self-diagnostic reveals which gaps are holding back your program and where your next growth opportunity may be hiding.

🤩How to find and test every working discount code for any product using ChatGPT

Stop guessing coupon codes. ChatGPT finds them all and tests every single one. Here's how.

Step 1: Switch to Work Mode Open ChatGPT and switch to Work Mode. This enables the browser and gives ChatGPT the ability to navigate live websites.

Step 2: Paste the product link and run this prompt Drop in the URL of the product you want and type:

"Find me every single available discount code for this product across every platform it is available on. Include retailer sites, coupon aggregators, brand newsletters, and any active promotional codes currently circulating."

ChatGPT returns a full list of codes across every platform the product is sold on.

Step 3: Run this prompt immediately after "Now open your browser and test each of these discount codes by going to the checkout page of each respective website. Apply every code and tell me exactly which ones are active, what discount each gives, and which one saves me the most money."

ChatGPT navigates to each checkout page, applies every code, and reports back with what actually works.

Step 4: Take the winning code ChatGPT gives you the single best working code with the exact saving. Copy it and apply it at checkout yourself.

🤖 Meta launches an AI agent that can act for you

Meta introduced Muse, an always-on personal AI agent that gets its own cloud computer and can carry out everyday tasks like booking restaurants, sending emails and shopping through a messaging-style interface.

The Shift:

Muse works like texting an assistant: Users can access the agent through its standalone app or WhatsApp, asking it to complete tasks rather than manually navigating different websites and services.

The agent gets its own cloud computer: Muse can browse websites, fill out forms and take actions through a virtual machine, allowing it to complete workflows that require direct browser interaction.

Muse can build missing integrations: The agent connects with Gmail, Spotify, Ticketmaster and OpenTable, while Meta says Muse can code its own integrations when a service does not already have one.

Payments and approvals are built in: Muse operates inside a secure cloud environment with Stripe-enabled payments and approval flows. U.S. users get limited free usage before $20 or $100 monthly subscription tiers.

Meta is pushing AI beyond answering questions into completing everyday digital tasks. Giving Muse a browser, app integrations, payments and the ability to create new connections means the agent can potentially handle entire workflows rather than requiring users to jump between services themselves.

Together with Playbook Pro

Unlock Hidden Cash to Add Months of Runway

Growth isn’t about chasing new funding; it’s about making every dollar move faster.

Leading DTC brands have gained months of extra runway by simply reshaping how cash enters and exits their business.

Financial Alchemy is your practical guide packed with real operator insights and ready-to-use templates. Inside, you’ll learn how to:

💰 Flip your cash cycle so revenue arrives well before expenses, real examples show brands securing payment weeks ahead of supplier deadlines.

📊 Stress-test growth plans with the Cash Flow Simulator to uncover an average of $80K in trapped working capital each quarter.

🧮 Speed payback with the Unit Economics Analyzer, cutting CAC recovery to as little as five weeks.

🤝 Negotiate supplier terms using the Supplier Toolkit, with tactics that added 12% operating margin in a single contract cycle.

Get your ebook and turn everyday cash flow into your most dependable growth engine.

🏆 Tools you Cannot Miss:

🤖 Kopai – Provides infrastructure to build, host, evaluate, and publish AI agents, including RAG, model orchestration, APIs, analytics, billing, and a marketplace for distributing agents.

🎯 GoodLads – Connects to Google Ads, generates hypotheses for improving ROAS or CPA, lets you approve recommended campaign changes, and tracks each experiment through to its result.

💬 bonds – Turns group-chat conversations into shared interactive apps, such as fitness challenges and trackers, with one live state that updates for everyone in the conversation.

🌍 Trancy Air – Translates selected text, typed input, and speech across macOS and Windows, with additional AI tools for grammar analysis, rewriting, pronunciation, and voice translation.

🧑‍💻 Catenary – Provides a local spatial IDE where developers can arrange coding-agent terminals, editors, and browser previews on a canvas and connect agents to pass context between them.

🚀 Quick Hits

🪖 The Pentagon reportedly requested a custom OpenAI system with “minimal refusal rates” for military commands, according to The Intercept. OpenAI denied signing any contract containing that language. 

📅 Google has made Gemini’s Daily Brief available to all U.S. users, expanding beyond paid subscribers. The AI feature summarizes emails, meetings and other information from connected apps.

⚖️ Anthropic faces a class action lawsuit alleging its $100 and $200 Claude Max subscriptions misleadingly advertised 5x and 20x usage increases while imposing additional five-hour and weekly limits.

🔐 Hackers are stealing Claude subscribers’ usage by hijacking login sessions with infostealer malware and generating unauthorized Claude Code tokens, with some users reporting their allowances drained without activity. 

☁️ Google Cloud and Accenture are launching a joint AI deployment unit, with Google training up to 1,000 Accenture engineers to build custom Gemini Enterprise applications directly inside companies. 

🦾 What's trending on socials

🎨 Astra teaches robot to paint - A developer says Astra learned to control a physical robot holding a paintbrush, progressively improving across attempts until it painted the Golden Gate Bridge. 

⚠️ Anthropic researcher quits over AI risk - Former OpenAI and Anthropic researcher Jacob Coxon resigned from Anthropic, accusing both companies of irresponsibly racing toward self-improving superintelligence and “gambling with our lives.” 

🏗️ AI turns consumers into builders - A user says AI generated near construction-ready renovation concepts from property surveys and public data, suggesting cheaper design and engineering could make custom-built products far more accessible.

💰 X replaces creator payout program - X has ended its Creator Revenue Sharing program, requiring eligible Premium users to apply for the new Original Content Rewards program to keep earning from post engagement.

🧪 ChatGPT designs experimental drug - A researcher claims ChatGPT designed PAC-3310, a new selective M4 muscarinic receptor agonist that he synthesized in a home lab as a potential schizophrenia treatment. 

🤳AI Nugget of the Day

Thanks for Decoding with us🥳

Your feedback is the key to our code! Help us elevate your Decode experience by hitting reply and sharing your input on our content and style.

Keep deciphering the AI enigma, and we'll be back with more coded mysteries unraveled just for you!