Sunday Rundown #84: AI Video Is Back & Hollow-back Girl
Sunday Bonus #44: Using ChatGPT scheduled tasks for media monitoring.
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:
Black Forest Labs now has a FLUX Pro Finetuning API that creators can use to train and customize its image models for precise outcomes.
Cognition upgraded the AI software engineer Devin to version 1.2. It now has better in-context reasoning and you can talk to it using voice.
Freepik’s paid users can now generate videos with sound right out of the box in its AI Suite.
Kokoro is a new open-source TTS model that generates high-quality, natural-sounding speech in seconds. (Try it for free here.)
KREA now lets you create 3D objects and insert them into your scene in the real-time editor tool.
Microsoft announced a new offering called Microsoft 365 Copilot Chat that integrates agents into its existing chatbot powered by GPT-4o.
MiniMax is branching further out beyond video models:
Its new open-source family, MiniMax-01, achieves on-par performance with top-tier LLMs even with long-context tasks.
It also has a text-to-audio model T2A-01-HD (catchy!) that can clone a voice in 10 seconds. (Try it here for free.)
Mistral has an upgraded coding model called Codestral 25.01 that is both faster and more accurate than its predecessor.
LinkedIn has two new AI-powered tools: a Jobs Match tool for job seekers and an AI Agent that helps smaller businesses find and recruit candidates.
Luma Labs is back in the video game with Ray2, its newest video model that can create coherent motion and realistic visuals. (Only for paid users for now.)
OpenAI has a beta feature called “Tasks” that lets you schedule future actions in ChatGPT. (Available to paid accounts now, rolling out to everyone later.)
Vidu (one of the “positive surprises” in my recent test) is out with a new video model: Vidu 2.0. It generates images fast, with better quality and consistency.
🔬 AI research
Cool stuff you might get to try one day:
Adobe is rolling out a “Firefly Bulk Create” feature that lets you resize or replace backgrounds for thousands of images at once.
📖 AI resources
Helpful AI tools and stuff that teaches you about AI:
Agent Recipes [reference] - a collection of reusable workflows and code snippets by Together.ai.
“AI Agents Are Here. What Now?” [blog post] - a great primer on agents from the Hugging Face team.
Heidi Health [tool] - AI-powered scribe for healthcare practitioners, shared with me by fellow Substacker
.The Future of Jobs Report 2025 [PDF] - a report by the World Economic Forum, which unsurprisingly heavily features GenAI. (Here’s my NotebookLM version.)
🔀 AI random
Other notable AI stories of the week - everyone’s partnering up:
Google partnered with The Associated Press to feed up-to-date information to the Gemini chatbot.
Microsoft announced a strategic collaboration with Pearson to transform and scale AI learning for everyone.
Mistral struck a deal with Agence France-Presse to integrate newswire stories directly into its Le Chat chatbot.
OpenAI is partnering with Axios to expand its local news coverage while improving ChatGPT’s understanding of news sources.
Perplexity has acquired read.cv and will combine the companies’ design and engineering teams.
World Bank’s pilot program on AI-assisted learning in Nigeria showed “overwhelmingly positive effects on learning outcomes” and helped close the gender gap in education.
🤦♂️ AI fail of the week
“Ah, what a perfect day to leave my torso at home and just chill!” #bliss
💰 Sunday Bonus #44: How to use ChatGPT tasks for brand media monitoring
With the new scheduled tasks feature, you can now ask ChatGPT to take actions at a specified time and email you when they’re finished. You no longer have to always initiate a chat from your side.
Note: Only paid users have access right now, but everyone should get the feature when it’s out of beta.
Because these tasks also work with web browsing, you can ask ChatGPT to look things up at regular intervals.
Basic examples of this include checking the weather or stock info, but you can also use this for more complex tasks in a business context.
So I want to show how you can use this for brand media monitoring and analysis, as well as getting actionable recommendations on further steps from ChatGPT.
I also share my starter prompt template that you can customize to your needs.