A week in Generative AI: Aleph, Study Mode & NotebookLM
News for the week ending 3rd August 2025
No big frontier announcements this week, but lots of great news and updates nonetheless. I think we’ll be having a big August/September so this feels a little like the quiet before the storm!
First up we have a great new video editing model from Runway that allows you to edit videos with natural language. OpenAI also released study mode for ChatGPT and Google released a new video overview feature to NotebookLM.
I’ve added a new ‘Web 4.0’ section to the newsletter to share and keep track of research and reports relating to how generative AI platforms are changing the nature of the internet. hopefully you find this useful and interesting, and I’ll comment on any news that I think is significant.
In Ethics News there are reports of search engines Indexing public ChatGPT conversations and exposing personal/sensitive information. There’s also more commentary around the US prioritising open models, and some research from Microsoft around which jobs it expects AI to replace, and which it thinks are ‘safe’.
Runway Introduces Aleph Video Editing Model
This is pretty cool - a video model that knows how to edit and transform videos as well as generate new content. This is exactly the sort of thing that video (and image) models have needed so that you’re not constantly iterating your prompts and generating lots of versions until you get exactly what you’re after.
I’m sure this isn’t 100% reliable (yet) and there will still be lots of iteration, but being able to more finely direct the generation through natural language prompts is a big step forward. It’ll be interesting to see if this is reliable and accurate enough be used alongside/instead of traditional editing software as well. If it is then it’s a big game changer.
ChatGPT launches study mode to encourage ‘responsible’ academic use
This is a great example of how we can make huge improvements to how generative AI works without needing to improve the underlying model. Even without new models there’s at least 5-10 years of innovation ahead of us.
Study mode is really interesting as it aims to take a major criticism of how generative AI is landing in educational settings and turn it into a strength. The opportunity here is for everyone to have access to their own personal tutor that tailors the learning experience to your knowledge and capability. I think this will be huge in professional settings as well to help train young talent and onboard new team members.
Google’s NotebookLM rolls out Video Overviews
This is a really nice new feature in NotebookLM if like me you prefer to take in information visually rather than just via audio. The new Video Overview feature pulls in images from the documents you’ve shared as well as generating new images to use as part of the visual presentation. This is great for good for explaining data, demonstrating processes, and making abstract concepts easier to understand.
Adam-U has human-like motion with 31 flexible joints
The first part of this video is nothing you won’t have seen before - a humanoid robot picking and sorting items. Classic stuff. But if you go to 1:47 on the video you can really see the 31 flexible joints in action and how it gives the robot very fluid, human like motion.
I think it’s easy to assume that when humanoid robotics platforms start being adopted in lots of different settings that they will be slow and clunky. Whilst the speed issue still persists, I think by the time the AI robotics software is ready for prime time the robotics hardware will be incredibly sophisticated and it looks like we’ll see very human-like movement.
Web 4.0
I thought I’d try something new and add links to lots of different ‘Web 4.0’ related articles as there is lots more research, reports, and interesting stats now available on how generative AI platforms are changing the nature of the internet. I don’t want this newsletter to become a Web 4.0 newsletter, so will only comment on research/reports that I think are significant, useful, and interesting. However, its worth sharing and tracking Web 4.0 related news as it happens:
Google’s AI Mode will ‘fundamentally redefine’ digital ad industry
GenAI apps doubled their revenue, grew to 1.7B downloads in first half of 2025
Zuckerberg: AI increased the time spent on Facebook and Instagram in Q2
AI Ethics News
Your public ChatGPT queries are getting indexed by Google and other search engines
Anthropic studied what gives an AI system its ‘personality’ — and what makes it ‘evil’
Trump’s AI plan is a massive handout to gas and chemical companies
Big tech has spent $155bn on AI this year. It’s about to spend hundreds of billions more
Zuckerberg signals Meta won’t open source all of its ‘superintelligence’ AI models
OpenAI to launch AI data center in Norway, its first in Europe
Mark Zuckerberg promises you can trust him with superintelligent AI
Long Reads
One Useful Thing - The Bitter Lesson versus The Garbage Can
Technology Review - The two people shaping the future of OpenAI’s research
Mark Zuckerberg - Personal Superintelligence
“The future is already here, it’s just not evenly distributed.“
William Gibson