Digital Ecosystems brings steerable NCA to browser

This feels less like a flashy demo and more like an interactive lab for artificial life. The useful part is not just the visuals, but the ability to intervene while learning dynamics are still unfolding.

• –Live steering turns NCA work from offline batch analysis into immediate hypothesis testing, which is a better fit for path-dependent systems.
• –Checkpointing and branching are the standout features: they make it easy to compare divergent outcomes from the same starting state.
• –Running entirely in the browser with TensorFlow.js and WebGL lowers friction for sharing, but it also makes performance and device limits part of the product.
• –The launch extends Sakana AI’s pattern of biology-inspired research tooling, which will matter more to AI researchers and ALife people than to mainstream builders.