Sunday Rundown #66: Mystic Magic & No-Iris Nancy

Sunday Bonus #26: Even more fun Midjourney style references.

Heads up: I’ll be traveling at the end of next week, so the next Sunday Rundown will be on Sunday, August 15.

I expect to share the recording of Friday’s Midjourney workshop with paid subscribers in the next few days.

Happy Sunday, friends!

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): a goodie for my paid subscribers.

Let’s get to it.

🗞️ AI news

Here are this week’s AI developments.

👩‍💻 AI releases

New stuff you can try right now:

1. Alibaba released a new multimodal vision model called Qwen2-VL, capable of understanding 20 min+ videos and scoring above GPT-4o on most LLM benchmarks.

2. Anthropic made Artifacts broadly available to all Claude users. (I previously showed how to use Artifacts to turn any material into an interactive course.)

3. Google is releasing a bunch of updates across its platforms:

   1. Subscribers can now upload documents, perform data analysis, and use Gems (customizable AI chatbots) inside Google Workspace.

   2. An AI-powered “Take notes for me” feature will keep a written record of your Google Meet calls.

   3. Imagen 3 is moving out of the experimental AI Test Kitchen and into Gemini Apps.

4. Google also released three new experimental models inside the AI Studio:

   1. Gemini 1.5 Flash-8B: Smaller variant of the original Flash model.

   2. Gemini 1.5 Pro (Improved): Especially good at coding and complex prompts

   3. Gemini 1.5 Flash (Improved): A better version of the original.

5. The Mystic image model is now available in the Freepik Pikasso image generator. (Requires a Premium subscription.)

6. Playground overhauled its website to focus on preset design templates, powered by the latest Playground v3 model. (Try it out.)

7. Runway now lets users extend Gen-3 Alpha videos up to 40 seconds.

---

🔬 AI research

Cool stuff you might get to try one day:

1. Google developed a GameNGen neural model that lets users play a game of DOOM by simulating new frames without an underlying game engine:

---

📖 AI resources

Helpful stuff that teaches you about AI:

1. Achieve More with GenAI [Course] - Paid MasterClass by four AI experts, including Substack’s own

   Ethan Mollick
   .

2. Anthropic’s System Prompts - a transparent list of system prompts Anthnropic uses for all of its Claude models.

3. Data and AI Trends 2024 [Downloadable] - a report by Google Cloud based on input from “hundreds of data professionals.”

4. ”How we built Artifacts with Claude” [VIDEO] - a quick and fun behind-the-scenes look at the making of the “Artifacts” feature:

🤦‍♂️ AI fail of the week

“Hey! My eyes are up h— Oh, oopsie!” (Final version.)

Anything to share?

Sadly, Substack doesn’t allow free subscribers to comment on posts with paid sections, but I am always open to your feedback. You can message me here:

---

💰 Sunday Bonus #26: Four fun Midjourney style references (Vol. 2)

Occasionally, I like to burn through my leftover “fast hours” to discover cool --sref styles.

I shared four of them some weeks ago. (Here’s more about --sref.)

Today, I’m back with four new ones.

Simply copy-paste them at the end of your prompt using their --sref [number] and have fun!