Google IO, Claude 4.0 and AI Video Generation – Live and Learn #69

✨ Quote ✨

In a world where knowledge is cheap, curiosity, curation, and judgment - signalled well - becomes insanely valuable.

– Sangeet Paul Choudary - (source)

🖇️ Links 🖇️

Google I/O Keynote Speech. Every year, I look forward to Google I/O because Google is working on a lot of awesome things, and this year has been no different. The number of announcements they made is insane. There's a novel diffusion-based approach to text models that is waaaay faster called Gemini Diffusion, they are improving their Jules Coding Agent, bring Virtual Try On into Shopping, created a tool for filmmakers called Google Flow, made big improvements to their image, video and audio models, put a lot more intelligence into search with AI Search Mode, and also push into the smart glasses market with Android XR and Gemini Glasses. And that's not even everything... the amount of things they work on is truly mindboggling—watching the whole keynote is a good use of your time to get a sense of what they are up to.

There have also been a bunch of other noteworthy big company events over the last two weeks, such as Nvidia's Keynote at Computex in Taiwan, which was very similar to the one Jensen gave at GDC. Also, Microsoft Build happened, where Satya showed improvements to their Copilot Agents, announced that they are open sourcing the VS Copilot extension code + framework, and showcased many updates on Azure's AI Foundry geared around making it the best AI platform out there. There was also IBM Think showcasing some of their work on Quantum Computing (and more).

Video Generation with Veo 3 by Google. I want to spend a bit of time looking at Veo 3 and AI video generation in general because it seems to be at an inflection point. The stuff that people are doing is insane, and I think that video generation is now at the point where it starts to become commercially viable, and where we will see more and more AI-generated video content around YouTube, TikTok, and Instagram. There's just so much going on. A few cherry-picked examples from Veo 3: a sitcom scene, interdimensional cable, and a collection of various scenes with dialogues and audio. Veo can create all the audio natively, too, which is crazy. But Veo 3 isn't the only capable video model out there. For example, there has been a whole sci-fi short film produced with a combination of different AI tools (Kling AI, Midjourney, Runway), and the quality is starting to approach that of a real movie. I think one year from now, we won't be able to tell anymore if a video is AI-generated or real. Scary, but also exciting possibilities for creative output and filmmaking.

Claude 4.0 by Anthropic. Anthropic is going more and more all in on coding models, and Claude 4.0 is cementing their position in this space, being even better at helping with software engineering tasks, especially in large codebases. It's a pretty expensive reasoning model to run, though, and the amount of money it costs, compared to slightly less powerful models, is considerable. For more on this general topic and frontier models in general, I really liked this analysis by Theo from t3.

The Way of Code by Rick Rubin. This is easily one of the most whimsical things I have seen in a while. Rick Rubin essentially answered the question of what would happen if you were to pair vibe coding with the writings of the Tao Te Ching, and this is the result. A weirdly beautiful amalgam of vibe-coded interactive visualizations paired with poetry musing about vibe coding and the creative process.

Sam and Jony - io Acquisition by OpenAI. OpenAI is trying to build its own AI-assisted hardware. Just like how smartphones and the internet brought with them paradigm shifts in the design of how we interact with computers, AI is changing what the best form factor could be, again. This push by OpenAI into manufacturing is about that, re-envisioning the products that are made possible by the rise of powerful AI. And with the people from io previously being designers at Apple, responsible for some of the most iconic products out there, this sure is an interesting development. Again, Theo has a really good video on the whole topic, and there is also a corny official announcement with Sam and Jony being best buddies in front of a camera, musing about what they are going to create and how this is going to change the world.

🌌 Travel 🌌

During the last two weeks, I have been full-time cycling down South America towards Ecuador. The Colombian Andes have kicked my ass, but I've seen more waterfalls than I could keep count of, my favorite being the Cascada Ojo de Dios in Mocoa (or maybe La Cascada Fin Del Mundo 🤔).

I also completed the Trampolin de La Muerte, together with another cyclist I met in the casa ciclista in Mocoa. The Trampolin is a road connecting Mocoa and Pasto, winding through thick jungle for about 120 km up and up and up through the Colombian Cordillera. It's a lot of climbing on very bad roads, with a lot of rain, especially around this time of the year. Still, it was very fun and rewarding. My bicycle and phone broke on the way, though, and so I had to stop for a while in Pasto for repairs, but I am getting ready to leave again now that I am re-charged and with a working phone and bike.

Next is Ecuador, where I will start the TEMBR, the Trans Ecuador Mountain Bike Route 😊 This part of the journey is something that makes me tremendously excited because it goes high up into the beautiful paramo landscapes on beautiful backroads for days on end. You can watch this short film of some other people who did it to get a glimpse of it.

But even though my time in Colombia is nearing an end, I loved every second of it (haha maybe not every second ^^)... I am sure to come back to this beautiful country. With that being said, enjoy some pictures 🙃

🎶 Song 🎶

Welcome to Vulf Records by Vulfpeck

Youtube Music | Spotify

---

That's all for this time. I hope you found this newsletter useful, beautiful, or even both!

Have ideas for improving it? As always please let me know.

Cheers,

– Rico