Introduction: LEGO That Builds Itself
Imagine a set of magical LEGO bricks. There is no instruction manual and no architect directing the construction. Instead, every brick is alive and only "whispers" to its neighbours: "If you're red, I'll turn red too," or "If you already have three siblings on your left, I'll shift to the right."
With only local communication, can these bricks grow into an elegant Eiffel Tower on their own?
In computer science, this is called a self-organising system, and the colour-changing, moving bricks are known as Neural Cellular Automata (NCA).
But scientists have long been stumped by an awkward problem: without a "god's-eye view" blueprint, these bricks tend to fall into collective confusion. They don't know where the head or the tail is, and what grows out is often just a messy blob of plastic rather than a tower.
In May 2026, researchers from the IT University of Copenhagen and Sakana AI (Milton L. Montero et al.) published a remarkably "alive" paper: Learning Developmental Scaffoldings to Guide Self-Organisation. They borrowed a secret from nature to solve this maddening problem.
What Is a "Developmental Scaffold"?
As Feynman noted, nature doesn't always need complicated formulas; it often uses the simplest physical differences to guide the development of complex life.
When a fertilised egg develops into an embryo, cells don't grow blindly from scratch. Before they begin dividing, the mother has already embedded invisible "signposts" in the egg—for instance, a chemical that is more concentrated on the left and more dilute on the right.
These invisible chemical gradients are the "developmental scaffolding."
Cells never see the whole blueprint. They only need to sense the local chemical concentration to understand: "Ah, I'm in the high-concentration zone—I should become a head."
The Paper's Breakthrough: Giving AI a Draft
The core idea is charming: don't expect AI cells to conjure miracles from a blank slate—learn to give them a draft first.
The researchers designed a two-step system:
1. Build a "draft generator" (SIREN): it doesn't draw details; it only scatters a faint, blurry background pattern across space (the scaffold). 2. Let the "AI cells" enter (NCA): the cells then perform local fine-tuning, growth, and differentiation based on this blurry draft.
Why This Makes AI Stronger
With this "draft first, refine later" approach, the researchers found qualitative improvements:
1. Curing "Symmetry Blindness"
Previously, AI cells facing empty space had no idea where to grow. With a scaffold, they can instantly tell north from south. This addresses the notoriously hard problem of symmetry breaking in biology.2. Impressive Robustness
The researchers deliberately injected noise and disturbances during growth. Scaffold-equipped systems stayed remarkably calm: even when some cells died, the survivors could quickly regenerate the original pattern using the background draft.3. High Compression Efficiency
Most strikingly, because the scaffold carries part of the "macro guidance," the internal rules of the AI cells (parameter count) can be greatly compressed. This proves that splitting the task between the environment (scaffold) and the individual (cell) is far smarter than making the individual bear all the pressure.Why This Paper Matters
The paper touches on a profound proposition: intelligence doesn't have to live in a single, monolithic "brain."
It suggests that future AI might grow like a forest, a coral reef, or our own bodies—emerging spontaneously from simple individuals under subtle environmental induction.
Summary
We don't need an omniscient god's-eye view. As long as we learn to scatter a few "key signposts" across space, AI can grow astonishing complexity on its own in the digital world—just like an embryo.
Next time you marvel at nature's exquisite biological structures, remember: beneath that beautiful skin lies a silent strategy called "self-organising scaffolding." Humanity has finally learned to rewrite that strategy in the world of AI.