The Weird and Wonderful Emerge In AI

Is Google getting arty? Plus, Figma gets a new competitor with MagicPath, Nvidia faces loses due to trade restrictions, and Black Forest Labs launch a new image model

May 30, 2025

Wait, is Google getting arty?

Late last night I sat mesmerized by Flow TV from Google Labs, a project to showcase the capabilities of their video model Veo. Against a backdrop of ethereal synth music I browsed through universes of infinite imagination..

Some of my favorites were the weird and wonderful channels like Movable Type, Window Seat, and so many more. Dare I say it was a more impressive demo showcase than the Sora launch from OpenAI?

Between this, Jony Ive’s joining forces with OpenAI, and Odyssey’s surreal demo of a world model that generates and streams interactive video in real time — is culture coming back to technology with a bang?

I have to say, I love it. Throughout history, creativity has always catalyzed invention. It’s giving me flashbacks to artist-hacker demoscenes of the 1980s and 1990s, the underground computer art subculture that went hand-in-hand with Doom (1993), the game that catalyzed the pursuit of graphics hardware (inspiring a little company called Nvidia).

In the bleakness of talk about how AI is just going to steamroll humanity, this glimmer of chaotic creative energy is a reminder that we are more resilient and adaptive as a species that we often give ourselves credit for.

I remain excited for what’s next.

-tara

We’re hosting the inaugural edition of Strange Sessions — think: Ted Talks meets Tiny desk concerts — on June 25 as part of SF Deep Tech Week. Join us!

Strange Sessions No. 1

Red Skies Rising: A Vision for Terraforming Mars

SF · 25 June 2025 · 07:00 PM · RSVP

Join us for Strange Sessions (think Tiny Desk concerts meets TED), where visionary thinkers and scientists explore groundbreaking ideas shaping our future.

In this edition, x-NASA aerospace engineer Alex Kling discusses how experiments like harvesting atmospheric energy for airborne Mars exploration and deploying nano-particles to trap heat could help make Mars more livable.

While the Martian atmosphere is cold and thin, the right technology could unlock radical possibilities. Discover how these bold ideas could change how we think about life on Earth and beyond.

RSVP

Anthropic launches a voice mode for Claude: This will allow Claude mobile app users to have “complete spoken conversations with Claude,” and will arrive in English over the next few weeks. Users can also switch between text and voice on the fly, and see a transcript and summary following conversations
Opera teases Neon, its first agentic browser: Opera has announced that it will release one of the first agentic browsers, which it calls Neon. Opera Neon will be able to automatically perform tasks for users– even after they've gone offline. (I have to say their marketing video is a hot candidate for the cringe artifact when we look back at this moment in 20 years’ time).
Odyssey’s new AI model streams 3D interactive worlds: Odyssey, a startup founded by self-driving pioneers Oliver Cameron and Jeff Hawke, has developed a world model that lets users interact with a video that is generated on the fly.
Gabe Newell’s new company is getting ready to put chips in brain: Starfish Neuroscience, the secretive startup owned by Valve cofounder Gabe Newell, is creating a "minimally invasive" chip that will target multiple regions of the brain instead of one, unlike BCIs from competitors like Elon Musk's Neuralink. Newell's startup is making its chip as small possible, with such tiny power demands that it doesn't even need a battery. It'll run, instead, entirely via wireless power transmission.
Huawei Supernode 384 disrupts Nvidia’s AI market hold: The CloudMatrix 384 system delivers a staggering 300 petaflops of raw compute power, alongside 48TB of high-bandwidth memory — a notable leap in AI infrastructure capability. Benchmark tests place the Supernode 384 in a strong competitive position: for e.g. Meta’s LLaMA 3 models ran at about 2.5 times faster than on typical cluster setups.
Black Forest Labs released FLUX.1 Kontext - a suite of generative flow matching models that allow you to generate and edit images. Unlike traditional text-to-image models, Kontext understands both text AND images as input, enabling true in-context generation and editing.
Watch out, Figma? Magicpath, a new AI tool for design, launched this week. Users are able to create beautiful components and functional apps in their browser-based canvas, which automatically builds production-ready code in the background
Nvidia expects to lose billions in revenue due to H20 chip licensing requirements: As Nvidia reports earnings for the first quarter of its fiscal year 2026, the company has released numbers on how the Trump administration’s recent chip-export restrictions are affecting business. Nvidia reported that the requirements will result in a $8 billion hit to the company’s revenue in Q2, which is predicted to be around $45 billion – a significant toll.