Presented by
Good morning, AI enthusiasts. For decades, mathematicians argued that AI could never do real mathematicsthe kind that wins a Fields Medal. Math, Inc.'s Gauss just proved them wrong.
A 200,000-line formal proof, completed in under three weeks, with errors caught in a peer-reviewed paper along the way. Has AI just crossed mathematics' highest threshold?
In today's recap:
AI formally verifies a Fields Medal math proof
Apple debuts M5 Pro and M5 Max chips
Build a Chrome extension with Manus, no code
Anthropic plans to sue after Trump security ban
4 new AI tools, prompts, and more
MATH, INC.
AI formally verifies a Fields Medal math proof
Recaply: Math, Inc. just completed the first formal verification of Maryna Viazovska's Fields Medal-winning sphere packing proofs in dimensions 8 and 24, producing roughly 200,000 lines of Lean code in under three weeks.
Key details:
Gauss is a reasoning agent that combines natural-language reasoning with formal proof writing, allowing it to conduct literature searches, write Lean code, and run verification tools autonomously.
Gauss completed the 8-dimensional proof in 5 days. The human team estimated that step would take 6 more months. It then formalized the 24-dimensional case in one additional week.
Along the way, Gauss found and fixed a missing minus sign and a flawed definition in Viazovska's published Annals paper, according to Math, Inc.
The formalization adds roughly 130,000 lines to the project, bringing the total from 70,000 to 200,000 lines, and is available in full on GitHub.
Why it matters: There's been plenty of skepticism about AI's ability to do real mathematics, the kind that wins Nobel-equivalent prizes. Gauss doesn't agree, and has 200,000 lines of verified code to back that up. In under three weeks, it completed a decade-scale formalization effort and found errors in a peer-reviewed Fields Medal paper. Whether or not this is a revolution in mathematics, it's hard to argue it isn't the start of something remarkable.
PRESENTED BY HUBSPOT
How can AI power your income?
Ready to transform artificial intelligence from a buzzword into your personal revenue generator
HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.
Inside you'll discover:
A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential
Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background
Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve
Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.
APPLE
Apple debuts M5 Pro and M5 Max chips
Apple
Recaply: Apple just unveiled M5 Pro and M5 Max for the new MacBook Pro, packing 4x the AI compute of the previous generation while raising prices by up to $400 due to a global RAM shortage.
Key details:
Built on Apple's new Fusion Architecture, M5 Pro and M5 Max connect two 3-nanometer dies into a single SoC, with Neural Accelerators in each GPU core for a major boost in on-device AI performance.
M5 Pro supports up to 64GB of unified memory at 307GB/s bandwidth; M5 Max doubles that to 128GB at 614GB/s. Both deliver over 4x the peak AI GPU compute of the previous generation.
The price increase of $100 to $400 per model is linked to a global RAM shortage driven by AI data center demand, according to TechCrunch, with analysts already predicting a dip in smartphone shipments this year.
MacBook Pro models with M5 Pro start at $2,199 for the 14-inch model; pre-orders open March 4, with availability beginning March 11.
Why it matters: While Apple's M5 chips mark some of the biggest computing leaps in MacBook history, the price tag tells another story. The AI boom's insatiable demand for memory is squeezing RAM supply and pushing prices up across consumer hardware. MacBook Pro buyers are paying $100 to $400 more than last year for that reason alone. If Apple is a bellwether, other hardware makers are likely about to have the same conversation.
TUTORIAL
Build a working Chrome extension with Manus in minutes
Recaply: In this tutorial, you will learn how to build a fully working Chrome extension using Manus's Chrome extension builder, without writing a single line of code, from describing your idea to having it live in your browser in minutes.
Step-by-step:
Go to manus.im/playbook/chrome-extension-builder, and prompt: "Please create a Chrome extension that can [your task] and display results in a [popup/sidebar]. I would like the interface style to be [minimal/modern] design."
Make your description specific. "Extract all email addresses from the current page and display them in a sortable popup list" gets better results than "help me with emails." The clearer the task, the closer the first version is to what you want.
Watch Manus build it live, it writes the manifest, JavaScript, HTML, and CSS files autonomously, then tests the extension before delivering the final package with a ready-to-follow setup guide.
Download the ZIP file Manus generates, unzip it, then open Chrome and go to chrome://extensions. Toggle on Developer Mode in the top right, click "Load unpacked," and select your unzipped folder. The extension appears in your toolbar instantly.
Test the extension on a live page, then refine it by describing changes back to Manus: "Make the popup wider" or "Add a copy-all button." Download the updated files and reload the extension in Chrome.
Pro tip: Start with an SEO analyzer. Ask Manus to build an extension that extracts meta titles, descriptions, H1 tags, and image alt text from any page, it's one of the most practical extensions you can build and something you'll actually use every day.
ANTHROPIC
Anthropic plans to sue after Trump security ban
Bloomberg
Recaply: Anthropic just announced plans to sue the Trump administration after the Pentagon labeled it a "Supply Chain Risk to National Security," banning military contractors and federal agencies from working with the company.
Key details:
The "Supply Chain Risk to National Security" designation, typically reserved for foreign adversaries like Huawei, bars any military contractor or supplier from doing business with Anthropic, extending the ban beyond federal agencies.
Agencies have 6 months to phase out existing Anthropic use; the company held a contract worth up to $200M with the Pentagon to advance responsible AI in defense operations.
Amodei said the dispute centers on Anthropic's refusal to let its AI power autonomous weapons or mass domestic surveillance, calling the designation "legally unsound" and a "dangerous precedent," according to Anthropic's public statement.
Hours after Trump's directive, OpenAI struck a deal to deploy its models on Pentagon classified networks, stepping into the space Anthropic had previously been the only company cleared to serve.
Why it matters: While AI companies have rushed to secure government contracts, much of the industry's safety narrative rests on the idea that companies can negotiate limits on how their models are used. Anthropic's situation tests that assumption directly. When the government demands unrestricted use, including scenarios the company calls incompatible with safety, the only options left are a courtroom and a public statement. The industry is watching to see which approach holds.
NEWS
What Matters in AI Right Now?
OpenAI is in discussions to deploy its AI systems on NATO's classified networks, days after securing a similar deal with the U.S. Department of Defense.
Simular launched Sai, an always-on AI co-worker that operates a private remote desktop, using GUIs, APIs, and code to handle tasks autonomously with approval-based guardrails.
OpenAI's head of post-training has left to join Anthropic as an individual contributor in RL research, after shipping o1-preview, o1, o3, and leading post-training for the GPT-5 series.
Cursor introduced MCP Apps in version 2.6, enabling interactive UIs from Amplitude, Figma, and tldraw directly inside the agent chat, alongside private team plugin marketplaces for Teams and Enterprise customers.
OpenAI rolled out GPT-5.3 Instant to all ChatGPT users and developers, reducing hallucinations by up to 26.8% and cutting unnecessary refusals and preachy caveats compared to GPT-5.2 Instant.
Google DeepMind launched Gemini 3.1 Flash-Lite in developer preview, priced at $0.25/1M input tokens with 2.5X faster output and adjustable thinking levels for high-volume workloads.
OpenAI is building an internal alternative to GitHub after repeated outages from Microsoft's platform, with staff discussing whether to eventually sell the repository tool to customers.
OpenAI teased GPT-5.4 on social media, signaling the next model in the series is arriving sooner than expected, just as GPT-5.3 Instant began rolling out to all users.
TOOLS
Trending AI Tools
🧠 Gauss - Math Inc's AI formalization agent that autonomously writes and verifies formal mathematical proofs in Lean code
💻 Sai - Simular's always-on AI co-worker
⚙️ Cursor 2.6 - AI code editor with MCP Apps
💨 Gemini 3.1 Flash-Lite - Google DeepMind's fastest and cheapest Gemini 3 model
PROMPTS
Create Cold Email Templates
#CONTEXT:
Adopt the role of cold email architect. The user needs to break through the noise of overstuffed inboxes where 99% of outreach gets deleted within 3 seconds. Decision-makers receive 100+ pitches daily, their assistants gatekeep aggressively, and generic templates trigger instant spam filters. Previous attempts using standard formulas failed because they sound like every other desperate sales pitch. The user has one shot to create genuine connection before being permanently blocked.PS: This is not the full prompt. Click the button below to access the complete prompt.
Have a favorite prompt? Tell us about it or rate today’s prompt, click here.
EVENTS
Perplexity Ask: March 11, 2026 • San Francisco, CA
Multimodal Frontier Hackathon: March 07, 2026 • San Francisco, CA
NVIDIA GTC 2026: March 16-19, 2026 • San Jose, CA
ClawCon: March, April, and May, 2026
The Era of AI Show: March 10, 2026 • New York, NY
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
