Sunday Rundown #119: Everything Everywhere & Cooperative Arm Wrestling

Sunday Bonus #79: Swipe file with 100+ Gemini 3 use cases.

Happy Sunday, friends!

🎉Today is November 23, which happens to be my birthday! To celebrate, I’ve unlocked “The Skeptical Writer’s Guide to AI” from my archive. Enjoy!🎉

Welcome back to the weekly look at generative AI that covers the following:

• Sunday Rundown (free): this week’s AI news + a fun AI fail.

• Sunday Bonus (paid): an exclusive segment for my paid subscribers.

---

In case you missed it, here’s this week’s Thursday deep dive:

Gemini 3: Google’s Silent Knockout Punch

Note: If you’re regularly missing out on my emails, remember to check your “Promotions” tab and mark whytryai@substack.com as a “Safe Sender.”

---

Let’s get to it.

🗞️ AI news

Here are this week’s AI developments.

👩‍💻 AI releases

New stuff you can try right now:

1. AI2 open-sourced Olmo 3, a model family that makes every training step traceable, which helps anyone looking to build or improve their own models.

2. Amazon launched Video Recaps on Prime Video: AI-generated highlight reels that summarize past seasons with narration, dialogue, and music.

3. ElevenLabs introduced Image & Video (Beta) to let you create images, audio, and video within a single platform. (Try it here.)

4. Genspark launched AI Workspace, an agentic space that completes complex tasks—from slide decks to entire apps—from simple prompts.

5. Google news:

   1. Antigravity is an agentic development platform where you can orchestrate agents to plan, code, and verify tasks in your editor, terminal, and browser.

   2. Gemini 3 is the world’s best large language model, with top-tier reasoning, multimodal, and agentic coding skills. (It also proves you don’t need hype.)

   3. Gemini app now lets you verify if an uploaded image was made by Google AI by parsing the invisible SynthID’s watermark

   4. Google Search added Canvas in AI Mode that lets you plan trips with real-time flight data, hotels, maps, and booking links by simply prompting.

   5. Google Vids expanded Gemini features like AI voiceovers and transcript trimming to all Gmail users.

   6. Nano Banana Pro is a SOTA image model that excels at rendering text, editing visuals, and grounding results in real-world info. (Try it here.)

   7. Scholar Labs helps break down complex research questions and find relevant academic papers. (Try it here.)

6. Manus released Browser Operator, a browser extension that runs Manus AI directly in your logged-in browser without additional authentication.

7. Meta news:

   • Segment Anything Model 3 (SAM 3) can segment and track objects in images or videos based on text or visual prompts. (Try in Playground.)

   • SAM 3D turns image into accurate 3D reconstructions of objects and people, complete with shape, texture, and pose. (Try in Playground.)

8. Microsoft news:

   • Agent 365 is a “control plane” that helps IT manage and secure AI agents across a company’s suite of Microsoft products.

   • Copilot Actions in Windows 11 let an AI agent complete desktop tasks like organizing files or converting PDFs in an isolated environment.

   • Excel, PowerPoint, and Word agents in Copilot can generate full spreadsheets, slides, and docs, grounded in enterprise data.

9. Midjourney news:

   • Style Creator lets you discover new aesthetics without prompts, relying primarily on visual grids.

   • Profiles give users a customizable space to show off their creations.

10. OpenAI news:

    • ChatGPT for K-12 Teachers is a free AI workspace that helps K–12 educators plan lessons, create study materials, and more. (US-only for now.)

    • GPT-5.1-Codex-Max is a frontier agentic coding model that can refactor entire projects and debug for hours while spending fewer tokens.

    • Group chats in ChatGPT let you bring family, friends, and coworkers into shared conversations to work, plan, and make shared decisions.

11. Perplexity news:

    • The AI shopping agent is now free for US customers, with one-click buying and the ability to recall past chats.

    • Asset creation tools let Pro and Max users create slides, sheets, and docs directly inside search.

    • Comet for Android brings the AI-native browser to your smartphone, including tab summarization, voice chat, and a built-in ad blocker.

12. Quora rolled out group chats for Poe, letting up to 200 people work together with any available AI model in a single conversation.

13. Replit launched Design Mode (powered by Gemini 3) that turns your ideas into interactive mockups or websites in minutes.

14. TikTok now uses invisible watermarks to label AI-made videos and lets users control how much AI content shows up in their feeds.

15. TwelveLabs brought video intelligence features into Frame.io, letting creatives find specific video shots, auto-tag scenes, and flag compliance issues.

16. xAI released Grok 4.1, with better emotional intelligence, improved creative writing skills, and fewer hallucinations.

---

🔬 AI research

Cool stuff you might get to try one day:

1. CraftStory announced Model 2.0, a video model that can create up to 5‑minute long studio-quality videos from one image and a script. (Sign up for early access.)

2. Google news:

   1. Gemini 3 Deep Think will boost Gemini’s reasoning and multimodal skills to tackle harder problems with code execution and deeper logic.

   2. Generative UI will let Gemini craft interactive tools and experiences from a prompt directly inside Search and the Gemini app.

3. Meta unveiled WorldGen, a system that turns text prompts into fully interactive 3D worlds compatible with game engines like Unity and Unreal.

---

📖 AI resources

Helpful AI tools and stuff that teaches you about AI:

1. “AnyLanguageModel” [TOOL]—a Swift tool from Hugging Face that lets developers easily run any local or cloud LLM with a single code tweak.

2. “Google for Education” [REFERENCE]—a collection of Google’s AI-focused courses and certifications for students, educators, and administrators.

---

🤦‍♂️ AI fail of the week

Beautiful. Such unity. You did it, team!

Send me your AI fail for a chance to be featured in an upcoming Sunday Rundown.

---

💰 Sunday Bonus #79: 100+ use cases & prompts for Gemini 3 Pro (swipe file)

Today’s bonus is a bit of a meta exercise.

I got Gemini 3 Pro to pull together and create an interactive swipe file of use cases for itself. It joins the ranks of my other popular swipe files:

• 90+ use cases for GPT-4o image generation (swipe file)

• 45+ practical use cases for o3 (swipe file)

• 40+ use cases for Claude 4 (swipe file)

• 72 use cases for Nano Banana (swipe file)

There are 111 use cases in 9 filterable categories:

Three sample use cases in the “Image Input” category.