Sunday Rundown #73: Anthropic's Blitz & AI Rhymes

Sunday Bonus #33: Cool Midjourney style references, chapter 3.

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.

Damn, this must’ve been the busiest week of the year!

👩‍💻 AI releases

New stuff you can try right now:

1. Anthropic is shipping:

   1. The already awesome Claude 3.5 Sonnet got a serious upgrade and shows “particularly strong gains in agentic coding and tool use.”

   2. Speaking of which, the new Computer Use feature turns Claude into a crude agent that can “see” your screen and use your computer like a human would:

   3. The new Analysis Tool in Claude can write and run JavaScript code to analyze uploaded CSV files and visualize its findings. (A lot like ChatGPT’s “Code Interpreter.”)

2. Canva is also shipping with lots of “Droptober” drops here in Octob—oh, I get it now:

   1. A text-to-image cool called Dream Lab to quickly make and integrate the resulting images into the rest of Canva’s templates

   2. AI-powered whiteboards that let you use AI to create a text summary of all the visual whiteboard elements.

   3. Improved Magic Write that makes smarter suggestions and gives you more control.

   4. Custom Mockups (coming soon) that creates mockup templates from photos.

   5. …and more, here:

3. ElevenLabs released a “Voice Design” feature that lets you create a new voice by simply describing it. Check out my result for this prompt:

   “Old woman with a low, husky voice. She speaks in an intense whisper”:

   What? Don’t look at me. You heard the lady!

4. Genmo released a base 480p version of its open-source video model called Mochi1. Try it for free. HD version is coming soon. Here’s my quick test for:

   “A steampunk robot playing the violin on stage.”

5. Google overhauled its Music FX DJ (which I tested in March) with more intuitive controls and better sound quality.

6. Haiper AI released version 2.0 of its video model that “sets a new industry standard for quality.” Here’s my steampunk robot violinist again:

7. Ideogram introduced a creative space called Canvas where you can work with multiple images as well as inpainting and outpainting tools:

8. Midjourney unleashed an “External Image Editor” that finally lets you upload and manipulate your own images. It’s awesome (I tried it here and here.)

9. Perplexity’s Pro Search now has a Reasoning Mode that can handle deep, multi-layered queries.

10. Playground released Playground v3, which apparently follows prompts better than top competitors like FLUX, Ideogram, and Midjourney.

11. Replicate now has its own “Playground” (no relation) that lets you easily run and compare different image models.

12. Stability AI is back with a new Stable Diffusion 3.5 model that’s customizable, efficient, and capable of handling different styles. (Try it on Hugging Face.)

---

🔬 AI research

Cool stuff you might get to try one day:

1. Anthropic also teased a new version of its small but affordable and speedy Claude 3.5 Haiku model to come out imminently.

2. Meta previewed SpiritLM, a model that can seamlessly switch between and combine speech and text inputs and outputs. (Request access.)

3. Microsoft announced autonomous agents for Copilot Studio to come out in public preview in November along with 10 prebuilt agents for Dynamics 365.

4. OpenAI showcased a “consistency model” framework (sCM) that can instantly produce images comparable to diffusion models in just two sampling steps.

5. Rhymes AI previewed yet another text-to-video model called Allegro. (View demo clips and join the waitlist.)

6. Runway is starting to roll out a feature called “Act-One” that can turn a recording of a single actor into multiple expressive characters:

---

📖 AI resources

Helpful AI tools and stuff that teaches you about AI:

1. “Developing a computer use model” - a look at Anthropic’s research process behind the new "computer use” feature.

2. “Solving complex problems with OpenAI o1 models” [VIDEO] - a showcase of practical ways to use the o1 reasoning models by OpenAI.

3. “The New Claude 3.5 Sonnet” [VIDEO] - wonderful deep dive into Claude 3.5 Sonnet and “computer use” by AI Explained.

---

🔀 AI random

Other notable AI stories of the week:

1. Google open-sourced SynthID, a watermarking system that embeds imperceptible markers in AI-generated content to make it easy to identify.

2. Hugging Face announced Hugging Face Generative AI Services (HUGS) to help people build AI applications using open models.

3. OpenAI stuff:

   1. Head of Policy Research Miles Brundage is the latest big name to leave the company.

   2. After leaving last month, former CTO Mira Murati is reportedly raising capital for a new AI startup.

4. Timbaland has partnered with Suno as a strategic advisor and showcased its AI features on his new single “Love Again.”

---

🗳️ Sunday poll

🤦‍♂️ AI fail of the week

I don’t know who “Magic Bob” is, but I want one! (Intended version.)

---

💰 Sunday Bonus #33: Four fun Midjourney style references (Vol. 3)

Every now and then, I burn through my Midjourney credits to roll the dice and discover cool new --sref styles.

I shared a few of them before: Part #1 and Part #2.

If you want to learn more about --sref, you can:

• Read this post

• Check out my “Midjourney Masterclass” workshop.

I know some of you are especially interested in these --sref codes, so I’ll share four more today.

To use them, just paste their --sref [number] at the end of your prompt.

Today’s showcase subjects are¹:

• Sunflower

• Owl

• Hat

I give each style below a “nickname” to describe it, but what matters is the --sref number itself.

Have fun!