So, Google's DeepMind dropped this Project Genie thing, and honestly, it sounds like a pretty neat way to mess around with AI. Remember Genie 3, that AI that could conjure up images and react to your movements in a simulated world? Well, they're letting people outside of Google actually play with it now.

There's a catch, naturally. You gotta be in the US, over 18, and shelling out $250 a month for Google's AI Ultra plan. Which, let's be honest, puts it out of reach for most of us. But hey, if you're already deep into the AI game, this could be a fun toy to experiment with.

Project Genie has a few different modes to tinker with. One that caught my eye is "World Sketching." Basically, Google's Nano Banana Pro model whips up a base image, and then Genie 3 uses that to build your world. You get to describe your character, choose the camera angle (first-person, third-person, the works), and decide how you want to wander around. It even gives you a little sketch preview before fully committing, so you can make tweaks.

I think that is a pretty cool, because it lets you see how the AI interprets your instructions. Plus, you can even riff off prompts that other users have cooked up. It reminds me of those AI art generators, but with a focus on creating explorable spaces.

Don't get it twisted though, Genie 3 isn't a game engine. It might look like a game, and it can simulate some physics-based interactions, but there aren't any real game mechanics. Your world-gen time is limited to a minute, and the output is capped at 24 fps and 720p. It's not going to replace your PS5 anytime soon.

However, it’s a fascinating glimpse into what DeepMind's been cooking up. And if you're already subscribed to AI Ultra, then this sounds like a pretty cool way to check it out.