Sunday Rundown #108: China Strikes Back & Rainbow Mnemonic
Sunday Bonus #68: Custom GPT that turns images into copy-paste prompts.
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): an exclusive segment for my paid subscribers.
In case you missed it, here’s the last Thursday deep dive:
Let’s get to it.
🗞️ AI news
Here are this week’s AI developments.
👩💻 AI releases
New stuff you can try right now:
Adobe launched Acrobat Studio, an AI-powered workspace for working with PDFs and turning them into creatives like infographics, slides, etc.
Alibaba open-sourced Qwen-Image-Edit that can make precise, high-quality image edits and accurately render English and Chinese text. (Try it for free.)
ByteDance open-sourced Seed-OSS, a 36B reasoning model that can handle 512K tokens of context and lets you adjust its thinking budget.
DeepSeek open-sourced V3.1, a hybrid thinking/non-thinking model that’s faster at responding and better at agentic tasks. (Try it for free.)
ElevenLabs news:
The new Chat Mode in the Conversational Agents platform lets users deploy text-only agents in minutes.
The following are now available via API: Eleven v3 (alpha), the company’s most expressive text-to-speech model, and Eleven Music, which lets you generate music tracks approved for commercial use.
Google news:
AI Mode in Search expanded to 180+ countries and got new agentic capabilities.
You can now create audio versions of your Google Docs documents and have them read out loud with natural voices.
Grammarly launched several AI Agents that help you find citations, grade your paper, predict reader reactions, and more.
LTX Studio added Camera Motion controls, letting you choose from dozens of presets when generating videos with its models.
Meta added AI translations to Facebook and Instagram that automatically dub and lip-sync your Reels into another language.
Microsoft introduced =COPILOT function in Excel that lets you write formulas in plain language and bring AI-powered results into cells.
NVIDIA open-sourced Nemotron Nano 2, a family of small, fast, and efficient reasoning models. (Try them for free.)
OpenAI launched ChatGPT Go, a low-cost plan for users in India with extended access to GPT-5, image generation, file uploads, and custom GPTs.
Runway announced Game Worlds Beta, featuring text-based, AI-generated game sessions with custom stories, characters, and gameplay. (Join the Beta)
Z.ai is launching a free smartphone AI agent that can books hotels, order food, install apps, and generate media.
🔬 AI research
Cool stuff you might get to try one day:
Google news:
Gemini Live is getting many upgrades like real-time visual guidance, better integrations with Google apps, and more lifelike voices.
Google Photos will soon be able to make conversational edits to your images by letting you simply describe the requested changes.
“Nano Banana” is a new image model with state-of-the-art image editing and consistency. Google is widely suspected to be behind it.
NotebookLM will reportedly soon incorporate Deep Research that pulls and analyzes web or Google Drive sources right in your notes.
📖 AI resources
Helpful AI tools and stuff that teaches you about AI:
Anthropic Academy [COURSES]—a collection of free AI fluency courses by Anthropic.
FlowbyGoogle [SOCIAL ACCOUNT]—Google’s X account that shares useful tips for its AI filmmaking platform, Flow.
WordGreed [GAME]—a fun word game that one of my readers,
, made with AI. Soundtrack by Udio. Coding by Gemini.
🤦♂️ AI fail of the week
Ah, yes! It goes “Red, Orange, Gello, Breen, Invigo, and Blank.” Thanks, ChatGPT!
Send me your AI fail for a chance to be featured in an upcoming Sunday Rundown.
💰 Sunday Bonus #68: Turn any image into a copy-ready prompt with my custom GPT
Here’s the problem with image prompts: Make them too vague, and you miss important details. Throw in too many keywords, and you’ll end up with a splatterprompt.
So how do you hit the sweet spot: just enough detail without overkill?
Why, you use my new Image-to-Prompt Converter, of course. How convenient!
This custom GPT takes any uploaded image and gives you a clean, natural language prompt to paste into your favorite image model. You can also incorporate tweaks to any part of the image.
Here’s how it works:
Upload an image.
Choose “Copy” (recreate the image) or “Remix” (tweak some elements).
In “Remix” mode, you’ll first see a breakdown of key elements to modify.
Get a clean and detailed prompt, ready for copy-pasting.
(Optional) Ask the GPT to generate the image natively with GPT-4o image generation.
Check it out here: