we've been making an ai thing with Google DeepMind

join the waitlist at

read the blog post

The project started as "ai prompts on the canvas", with a graph of connected blocks. Updating each block would that trigger the next block to generate content.

But pretty quickly we realized that we could use the AI model (in this case, Gemini Flash) as a kind of natural language code interpreter, referencing stored information and operating on it

Since every block's output could serve as the input for a new generation, regardless of type of content (sound, text, image, etc) the flows could get really wild

We're still finding out what we can use this thing for, or if it even makes sense to anyone outside of the office here... but we're pretty sure that it's incredible. For now, join the waitlist and get early access at

Fixed a few bugs on mobile: - scroll on the account page - font size on landing page - device height on iOS devices

@GoogleDeepMind I made a tldraw computer band name generator!

@GoogleDeepMind cannot wait to build the most horrendous contraption of all time

@GoogleDeepMind So many flows, so little time ⏰