Welcome Back! Here's What Happened in AI While You Were on Holiday

Thought the AI world would slow down over Thanksgiving? Not a chance! While you were enjoying turkey and time with family, the AI industry kept moving at full speed.

Here are the highlights from just the past two weeks:

  • Claude Opus 4.5 launched by Anthropic
  • Kling O1 video editing model released
  • Kling 2.6 adds native audio support
  • Retake tool launched by Lightricks
  • Gen-4.5 video model unveiled by Runway
  • FLUX.2 model launched by BFL
  • Gemini 3 Deep Think now live for Ultra users
  • Workspace Studio by Google lets you build AI agents
  • OpenAI Code Red alert issued internally
  • Garlic model teased (yes, that’s the name!)
  • OpenAI leaks show ads in chats being tested
  • Perplexity AI adds memory to conversations
  • DeepSeek rolls out V3.2 and Speciale models
  • Mistral 3 models now available
  • Suno x Warner Music announce major partnership
  • Stories feature launched by Character AI
  • NotebookLM introduces slide-deck creation

Some of these stories—like Gemini Deep Think and the $500K Google hackathon—are so exciting, we’ve broken them down in detail just below. 👇

Build with Gemini 3 Pro in Google AI Studio

Gemini 3 Pro Hackathon Image

Google DeepMind is running a one‑week AI challenge from December 5–12, 2025. It’s called the “Vibe Code with Gemini 3 Pro” hackathon, and it’s open to anyone who wants to build smart apps using Gemini 3 in Google AI Studio.

Winners will share $500,000 in Gemini API credits!

Think of it like a science fair for AI. You give Gemini a prompt, and it helps build an app that actually works—with video, text, code, or even diagrams. If your idea helps in fields like education, health, or business, you could win big.

To enter, you’ll need to:

  • Build your project in the AI Studio "Build" tab
  • Post a demo on Kaggle
  • Share a short write‑up and 2‑minute demo video

Judging focuses on:

  • Real-world impact (40%)
  • Tech depth (30%)
  • Creativity (20%)
  • Clear presentation (10%)

You’ve got one week—go make something awesome!

Source: Kaggle

Read More on Kaggle

Gemini 3 Deep Think launches for Google AI Ultra users

Gemini 3 Deep Think Banner
  • Google has introduced a new Gemini 3 Deep Think mode in the Gemini app, initially available to Google AI Ultra subscribers.
  • The mode is designed for hard reasoning problems, including advanced math, science and logic challenges.

Performance and benchmarks

  • Scores around 41% on the tough Humanity’s Last Exam benchmark—without any external tools.
  • Reaches 45.1% on ARC-AGI-2 using code execution, showing real problem-solving power.

Technology and lineage

  • Uses parallel reasoning—thinking through many possible answers at once to improve accuracy.
  • Builds on Gemini 2.5 Deep Think, which performed at the top in math olympiads and coding contests like ICPC.

How users can try it

  • Open the Gemini app, choose Gemini 3 Pro as the model, and select “Deep Think” in the prompt bar.
  • It’s part of the Pro/Ultra subscription, made for people who need serious reasoning power.

Source: Google Blog

Read More on Google Blog

Is GPT‑5.2 Coming Next Week?

GPT-5.2 Code Red Image

Multiple reports say OpenAI is preparing to launch a GPT‑5.2 upgrade around December 9 as a direct response to Google DeepMind’s new Gemini 3 models. It’s described as a “code red” situation, but there’s no official confirmation yet—everything is based on leaks from tech sites and insiders.

What reports say about GPT‑5.2

  • Several outlets mention that GPT‑5.2 is ready for release and could come as soon as early next week.
  • The update is framed as OpenAI’s first major answer to Gemini 3, which recently topped benchmarks and triggered internal urgency.

Expected focus of the update

  • Leaks suggest GPT‑5.2 aims to close Gemini 3’s lead in reasoning, coding, speed, and science tasks.
  • It may roll out in stages—internal testers first, then partners, followed by public ChatGPT/API deployment.

How solid is this information?

  • Many sources refer back to a “code red” memo reportedly from OpenAI, covered by major tech journalist reports.
  • Until OpenAI issues an official announcement, treat this as well‑sourced but unofficial news — plans could still change.
Original Instagram Post Screenshot

🎓 AI Tutorial: Exploring Angles & Styles of the Same Image

Using a single scene, you can generate multiple cinematic perspectives with AI — from drone views to close‑ups. Here’s how I did it, and how you can try it yourself.

Alternative Angle of Desert School Image

🧠 Original Prompt That Created the Scene

A wide‑angle landscape photograph capturing an open‑air desert school in rural Pakistan. In the immediate foreground is a large, weathered white marble block with the black Urdu calligraphy 'استاد بنے رہنے کے لئے شاگرد بنے رہنا ضروری ہے۔' engraved upon it. Resting on the marble is a vibrant orange 'Rahber' water cooler with a metal cup on top, and a stack of four well‑worn books: 'Why Nations Fail', 'Rich Dad Poor Dad', 'The Alchemist', and 'Sapiens'. In the midground, under the expansive shade of a large acacia tree, a male teacher in traditional shalwar kameez writes on a wooden blackboard, instructing a diverse group of children seated on patterned mats. The background is a vast, arid desert landscape with rolling hills under a clear blue sky. The sunlight is bright and natural, highlighting the textures of the marble, sand, and foliage.
            

📸 More Angle & Style Prompts to Try

  • 🔺 Drone View: Show the exact same scene from a drone’s aerial view directly above the subject.
  • 🔄 360° Rotate: Render the same scene in a full 360° spin around the subject, changing viewpoint smoothly.
  • 🔁 Reverse Shot: Create the scene from the opposite direction — what’s behind the original viewpoint.
  • 🔳 Top‑Down / Overhead: Generate a top‑down cinematic view of the full setup with dramatic lighting.
  • 🎥 Wide Lens Focus: Render the scene using a wide 14 mm lens for maximum environment immersion.
  • 👁️ POV View: Create the scene as seen from the main character’s eyes (first‑person perspective).
  • 🔍 Zoom Detail: Focus tightly on one object in the scene (the water cooler, books, or marble) using a macro lens.
  • 📽️ Cinematic Bars: Reformat the image to 21:9 cinematic ratio with black letterbox bars and moody grading.
  • 🌅 Time of Day Shift: Render the scene at a different time — golden hour, sunset or night — while keeping composition constant.

Feel free to attach your own image and run any of these prompts in your favorite AI tool. Different angles and styles can totally change the mood — and often make the same scene feel fresh and cinematic.

💡 What would you love to see more of?
Just hit reply to this email — I'd genuinely love to hear what AI topics, tutorials, or tips you'd like us to explore next.
Urdu Ai

Urdu Ai

Urdu AI Updates — The fastest-growing AI newsletter in Urdu, bringing the latest AI tools, trends, and insights to Urdu speakers worldwide. Part of the 1 million+ strong Urdu AI community.

Ai for GCC

Ai for GCC

AI for GCC is a FREE newsletter uncovering how artificial intelligence is transforming business, innovation, and policy across the Gulf. Discover real opportunities, regional AI trends, and practic...

Keep Reading