🧠 Meta’s chief AI scientist steps away

Good morning, AI enthusiast. One of the godfathers of AI just walked out of Meta, and he’s building something he claims could spark the next revolution in machine intelligence.

In today’s AI newsletter:

• Yann LeCun leaves Meta to build independent AI Lab

• Nvidia smashes records with $57B revenue

• OpenAI launches GPT-5.1-Codex-Max

• How to access and use Gemini 3, Google’s most powerful AI model

• AI tools & more

AI NEWS

🧠 Yann LeCun Leaves Meta to Build Independent AI Lab

Yann LeCun, Meta’s long-time Chief AI Scientist and deep learning pioneer, is leaving the company to create an “independent entity” focused on building AI systems that understand the physical world, reason, remember, and plan. Meta will stay a partner and retain access to the lab’s innovations.

• Memo says LeCun wants to build AI with world understanding, persistent memory, and complex planning

• Reports say he clashed internally as Meta shifted attention toward generative AI and new hires

• LeCun has been openly dismissive of LLM-centered AGI efforts, calling AI safety fears “BS”

• His work focuses on “world models”, systems trained on videos, not text (e.g., Meta’s V-JEPA-2)

• Meta plans to partner with his new company, though structure and funding remain unclear

LeCun leaving Meta signals a major split in AI research philosophy, between companies chasing larger LLMs and researchers betting on embodied, physics-aware intelligence. His new lab could become a counterweight to OpenAI, Google DeepMind, and Anthropic in defining what the next era of AI looks like.

AI ECONOMY

🚀 Nvidia Smashes Records With $57B Revenue

Nvidia reported a record-breaking $57B in Q3 revenue, up 62% YoY, as demand for its AI chips continues to skyrocket. Net income hit $32B, beating Wall Street expectations across the board.

• Data center revenue hit an all-time high of $51.2B, up 66% YoY

• Blackwell Ultra GPUs are “off the charts,” with cloud GPU inventory completely sold out

• Nvidia announced AI factory + infrastructure deals totaling 5 million GPUs this quarter

• Compute demand for training + inference is compounding “exponentially,” says CEO Jensen Huang

Nvidia’s earnings show the AI race isn’t slowing, it’s accelerating. With AI factories, sovereign compute deals, and GPU demand stretching years ahead, Nvidia’s dominance looks less like a bubble and more like the backbone of the next computing era.

AI MODELS

💻 OpenAI Launches GPT-5.1-Codex-Max

OpenAI released GPT-5.1-Codex-Max, its new agentic coding model built for massive context, multi-hour tasks, and end-to-end software development. It replaces the previous Codex version across all interfaces.

• Outperforms rivals with 77.9% on SWE-Bench Verified, beating Anthropic and Google

• Uses 30% fewer thinking tokens while running 27–42% faster on real tasks

• New “compaction” system keeps tasks alive across millions of tokens for day-long sessions

• First Codex model optimized for Windows environments and complex command-line workflows

Codex-Max pushes AI coding from quick helpers to true long-haul engineering partners, capable of debugging, refactoring, and executing end-to-end tasks that used to require full human oversight.

HOW TO AI

🎨 How to Access and Use Gemini 3, Google’s Most Powerful AI Model

In this tutorial, you’ll learn how to access and use Google Gemini 3, Google’s most advanced AI model with PhD-level reasoning,

 🧰 Who is This For

• Anyone wanting the newest Google AI model

• Developers exploring the new “vibe coding” abilities

• Students & researchers needing deep reasoning

• Professionals using Gemini for long documents, PDFs, or videos

• Gemini Advanced users wanting full Deep Think access

STEP 1: Access Gemini 3 Based on Your Tier

Go to gemini.google.com or open the Google/Gemini app.

Depending on your account:

Free users:
You’ll get Gemini 3 Flash or a base Gemini 3 model automatically.

Gemini Advanced (Ultra):
You unlock Gemini 3 Pro + Deep Think mode for complex reasoning.

Google Search users:
AI Mode (toggle at the top) now runs on a version of Gemini 3.

This ensures you’re using the newest model regardless of platform.

STEP 2: Start Chatting With the New Reasoning Engine

Gemini 3 understands nuance, tone, and subtext far better than 2.5.

Just open the chat and type naturally, For example: “I’m feeling cozy and slightly gloomy today. Write something that captures that mood using rain as a metaphor.”

The model captures vibe, emotion, and intention with much higher accuracy.

STEP 3: Use Deep Think & Multimodal Features

Deep Think (Advanced users)

At the top of the interface, open the model picker and select:

• Gemini 3 Pro, or

• Thinking Mode (region-dependent)

Try something complex:

“Break down the weaknesses in this business strategy, then propose three counter-moves.”

You’ll see a Thinking.. phase where Gemini processes before answering.

Multimodal uploads

Gemini 3 can handle long videos, huge PDFs, and full books.

Try these:

• Upload a video:
  “Summarize only the last 10 minutes.”

• Upload a 500-page PDF:
  “Find every mention of ‘renewable energy’ and summarize sentiment changes across chapters.”

The model uses its expanded context window to analyze everything at once.

STEP 4: Explore Dynamic Views & Vibe Coding

Dynamic Views

Ask Gemini to generate an interactive UI instead of plain text.

Try: “Show me an interactive Space Race timeline with clickable events.”

Gemini generates a structured, scrollable visual layout.

Vibe Coding (for developers)

Open Google AI Studio or the Gemini coding workspace.

Describe your app in natural language: “Build a retro 1980s-style Pomodoro timer with neon green text on black.”

Gemini 3 generates the logic, UI, and styling in one go, fully runnable.

Microsoft launched Agent 365, a platform for managing, securing, and governing AI agents, with capabilities like agent registry, performance analytics, and more.

 Elon Musk says xAI plans to develop a 500-megawatt data center in Saudi Arabia with state-backed AI startup Humain; the data center will rely on Nvidia chips.

OpenAI announces ChatGPT for Teachers, designed for K-12 educators and school districts, and says it will be free to K-12 educators in the US through June 2027.

Meta releases SAM 3, a model for object detection, segmentation, and tracking in images and videos, and SAM 3D, which can reconstruct objects and humans in 3D.

✨ Gemini 3: Google’s next-gen engine for multimodal reasoning.

⚙️ Antigravity: Google’s new AI-powered dev platform

🤖 Grok 4.1: Faster: smarter, and upgraded for deeper reasoning

📚 NotebookLM: now works with Deep Research, Sheets, Images, and PDFs.

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