Sunday Rundown #102: AI Browsers & Breaksorcism
Sunday Bonus #62: My entire archive as a Notebook you can talk to.
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 perk 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:
Apple expanded Apple Intelligence with live translation, visual search, Genmoji upgrades, an AI workout buddy, and more.
ByteDance released Seedance 1.0, an image model that somehow snagged the #1 leaderboard spot from Google’s impressive Veo 3. (Try the mini version.)
Genspark launched Genspark AI Browser, a lightning‑fast browser powered by the Super Agent under the hood. (Download iOS here. Windows coming soon.)
Google news:
Veo 3 Fast is a cheaper, faster version of Veo 3 with the same 720p native output. (Available in the Gemini app and Flow for paid users.)
Audio Overviews can now show up for certain search results to give you a quick audio summary. (Opt in via Search Labs.)
KREA AI introduced Krea 1, its new image model with superior aesthetic control, artistic knowledge, and output quality. (Request early access.)
Meta launched Meta AI Video Editing, letting users restyle uploaded video clips using 50+ presets.
Microsoft news:
Transcript-based editing in Clipchamp lets you trim videos by deleting specific text sections. (Rolling out to work accounts now.)
Copilot Vision is an AI assistant that can see your screen, understand apps or docs, and advise you where to click. (Available in the US.)
Mistral AI news:
Magistral is the company’s first reasoning model family with competitive performance and transparent, native chain-of-thought output.
Mistral Compute is an AI infrastructure stack for enterprises, including GPUs, orchestration, APIs, etc. compliant with European regulations.
OpenAI news:
Advanced Voice in ChatGPT now sounds more natural and can effortlessly translate a conversation between two languages.
Projects got even more useful and now support deep research, voice chat, improved memory, and more.
o3‑pro is now the company’s best reasoning model, rolling out to all Pro users. In parallel, o3 is now much cheaper without a drop in performance.
Runway introduced a Chat Mode, which lets you generate images, video, etc. using Runway’s models through an intuitive, conversational interface.
Windsurf launched Windsurf Browser, a “fully functional AI-integrated browser,” letting the coding assistant view and interact with browser tab content.
🔬 AI research
Cool stuff you might get to try one day:
Apple unveiled STARFlow, a scalable model that generates high-resolution text-conditioned images.
Meta AI showcased V-JEPA 2, a model that learns about the world by watching videos, which can help robots gain world knowledge with less training.
Midjourney is gearing up to release its long-awaited V1 Video Model and is running a “rating party” where users can rank video outputs.
OpenAI postponed its upcoming open-weights model to late summer, but Sam Altman believes it will be “very very worth the wait.”
📖 AI resources
Helpful AI tools and stuff that teaches you about AI:
“Seizing the agentic AI advantage” [REPORT] - fresh insights from McKinsey & Company.
🤦♂️ AI fail of the week
These men are unnervingly chill about witnessing a demonic possession.
Send me your AI fail for a chance to be featured in an upcoming Sunday Rundown.
💰 Sunday Bonus #62: Talk to the entire Why Try AI archive via NotebookLM
I’m really psyched about this one!
I’ve long wanted to build an interactive version of my brain you could talk to.
Then, last week, NotebookLM made it possible to share Notebooks with others.
So I extracted the full Why Try AI archive and plugged it into a notebook.
You can now ask any AI-related question, and get the “Daniel” take in response:
![Shall I use "photorealistic" to get better photos from AI image models? No, you generally should not use "photorealistic" to get better photos from AI image models like Midjourney. The sources explain that terms like "photorealistic," "ultrarealistic," and "hyperrealistic" are actually used to describe art that is made to look like a photo, rather than an actual photo itself. When these terms are used with Midjourney, even if you include the word "photo" in your prompt, the resulting images tend to appear as if they were painted to look like photos, rather than being genuine photographs. For AI images that genuinely mimic the look of an actual photo, the term "photographic" is suggested. Instead of using "photorealistic" or similar terms, it is recommended to simply use "photo" on its own if you want a photographic image. Midjourney, for example, is capable of generating impressive photorealistic images that can be virtually indistinguishable from real life. Other models like Ideogram 2.0 and FLUX 1.1 [pro] are also noted for their ability to produce photographic images. For a consistent reflection of requested styles, including photographic ones, GPT-4o is mentioned as being highly effective.](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9b877d41-fb51-46a0-b7d5-984e9c8146d1_899x534.png "Shall I use \"photorealistic\" to get better photos from AI image models? No, you generally should not use \"photorealistic\" to get better photos from AI image models like Midjourney. The sources explain that terms like \"photorealistic,\" \"ultrarealistic,\" and \"hyperrealistic\" are actually used to describe art that is made to look like a photo, rather than an actual photo itself. When these terms are used with Midjourney, even if you include the word \"photo\" in your prompt, the resulting images tend to appear as if they were painted to look like photos, rather than being genuine photographs. For AI images that genuinely mimic the look of an actual photo, the term \"photographic\" is suggested. Instead of using \"photorealistic\" or similar terms, it is recommended to simply use \"photo\" on its own if you want a photographic image. Midjourney, for example, is capable of generating impressive photorealistic images that can be virtually indistinguishable from real life. Other models like Ideogram 2.0 and FLUX 1.1 [pro] are also noted for their ability to produce photographic images. For a consistent reflection of requested styles, including photographic ones, GPT-4o is mentioned as being highly effective.")
What’s even cooler is the “Interactive Mode” for audio overviews.
This kicks off a standard NotebookLM podcast but lets you jump in and ask questions using your voice.
The hosts are quite reliable at digging up my takes, while also not making stuff up when you ask about something I haven’t covered.
Here’s a quick demo:
So if you ever wanted to pick my brain, you now have a personal “AI Daniel” to talk to.
Enjoy: