🧠 AI just discovered new Physics

Good morning, AI enthusiast. A new AI model from China just made a breakthrough that feels straight out of science fiction, an AI that can discover the laws of physics on its own.

In today’s AI newsletter:

• AI learns to derive Physics equations

• Perplexity to power Snapchat search in $400M deal

• Meta goes omnilingual with 1,600+ language ASR

• How to generate a complete product blueprint for your AI projects

• AI tools & more

AI SCIENCE

⚛️ AI Learns to Derive Physics Equations

Researchers from Tsinghua University, Peking University, and others have built an AI called PhyE2E that can automatically derive physics equations directly from raw data, no human intervention needed.

• Trained on real scientific datasets and known physical laws

• Uses transformer models + symbolic reasoning to generate human-readable equations

• Produces compact, unit-consistent formulas scientists can actually test

• When given NASA astrophysical data, it not only matched human findings, it improved them, even refining equations for solar cycles

PhyE2E moves AI beyond pattern recognition into true scientific reasoning. It’s not just fitting curves, it’s learning how the universe works and expressing it in the language of physics. We’re witnessing the dawn of automated discovery, where AI could accelerate scientific progress by thousands of times.

AI PARTNERSHIPS

👻 Perplexity to Power Snapchat Search in $400M Deal

Snap announced a partnership with Perplexity to integrate its AI search engine directly into Snapchat’s My AI chatbot, giving users instant, conversational answers without leaving the app.

• Perplexity will pay Snap $400M in cash and equity for the partnership

• Integration gives Perplexity access to 940M+ Snapchat users worldwide

• The AI search feature will roll out early next year, with revenue recognition starting in 2026

• Snap reported $1.51B in Q3 2025 revenue, up 10% year-over-year

• Snap’s subscription tier, Snapchat+, has now surpassed 17M users

This deal cements Perplexity as a serious contender in AI search, and marks Snap’s biggest step toward turning its chat ecosystem into an AI-powered discovery engine. With this integration, Snapchat could become the first social app where users chat, search, and shop, all in one feed.

AI NEWS

🗣️ Meta Goes Omnilingual With 1,600+ Language ASR

Meta’s new Omnilingual ASR system is the most expansive speech recognition model ever released, capable of turning spoken language into text across nearly every major and minor tongue on Earth, no retraining required.

• Natively supports 1,600+ languages, 50x more than OpenAI’s Whisper

• Zero-shot learning extends coverage to 5,400+ languages

• Released fully open source under Apache 2.0, free for commercial use

• Includes a 7B-parameter audio model and 4.3M hours of multilingual speech data

Meta just made speech tech truly global, empowering local communities, researchers, and enterprises to build voice tools in their own languages without costly retraining or licensing. It’s not just an ASR release; it’s a statement, AI shouldn’t speak only the world’s top 100 languages.

HOW TO AI

💻 How to Generate a Complete Product Blueprint for Your AI Projects

In this tutorial, you’ll learn how to create a detailed AI Product Requirements Document (PRD) that outlines every part of your product, from tech stack and UI flow to user experience and database structure, all in under a minute using a free online tool.

🧰 Who is This For

• Developers who want to give AI coding tools precise project instructions

• Founders building MVPs or prototypes fast

• Designers and product managers documenting new product ideas

• Anyone who wants AI to build exactly what they envision

STEP 1: Access the PRD Generator

Head over to PRD Generator.

On the homepage, you’ll see four main options, Product Roadmap Generator, MVP Plan Generator, Product Launch Plan Generator, and Feature Prioritization Tool.

Make sure you are in Product Requirements Document (PRD) Generator to begin.

STEP 2: Describe Your Idea

You’ll see several text boxes where you can describe your project:

In-depth explanation of the idea: Explain what your product does and how it works.

Example: “A simple to-do list that lets users schedule tomorrow’s tasks, sends reminders, and automatically rolls over incomplete tasks.”

Problem it solves:

Example: “Helps busy professionals manage daily tasks quickly and stay organized.”

Target audience:

Example: “Freelancers, remote workers, and small business owners.”

Features:

Example: “Task lists, reminders, calendar sync, voice commands.”

Once filled, double-check everything to make sure your descriptions clearly match your product’s vision.

STEP 3: Generate Your PRD

After you’ve entered all details, click the Generate button at the bottom.

Within seconds, the tool will create a complete AI Product Requirements Document containing your project overview, objectives, core features, tech stack, and suggested design flow, ready to copy, share, or export.

STEP 4: Export and Customize Your 3D Model

Now take that generated PRD and paste it into any AI coding assistant.

Because the PRD clearly defines every detail, your AI tool will now build what you actually want, with the right architecture, UI flow, and logic from the very first try.

Anthropic expects to break even in 2028, while OpenAI projects ~$74B in operating losses that year before turning a profit in 2030.

Intel CTO and AI officer Sachin Katti says he is leaving the company after four years to join OpenAI; Intel says CEO Lip-Bu Tan will take over its AI efforts.

Google launched the File Search Tool, a managed RAG system that helps ground Gemini using users’ own data.

OpenAI urged the Trump administration to extend the Chips Act tax credit to include AI data centers, servers, and power infrastructure.

🎨 Creative OS: Canva’s new AI-powered visual suite

🧠 Kimi K2 Thinking: Moonshot AI’s new open-source advanced reasoning model

🤖 Grok-4 Fast: xAI’s new lightweight model with a massive 2M token context window

🤖 Excelmatic: Use AI to analyze, visualize, and automate your data

THAT’S IT FOR TODAY

Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!

See you tomorrow :)

- Dr. Alvaro Cintas