Ffellonics and the Complex-Systems Critique of LLM Emergence

A 2025 paper by David Krakauer, John Krakauer, and Melanie Mitchell — Large Language Models and Emergence: A Complex Systems Perspective (arXiv:2506.11135) — offers a measured critique of claims that large language models exhibit genuine emergent capabilities. The authors argue that true emergence is not simply a consequence of scale. It is a specific phenomenon in which large numbers of locally interacting components produce qualitatively new, higher-level properties that can be described by lower-dimensional effective theories. They further propose that intelligence is an emergent property characterised by increasing efficiency — "less is more" — in which systems achieve greater capability through simpler, more compressed internal models.

Ffellonics provides a precise and minimal illustration of exactly this kind of emergence.

Many-Body Interactions Producing Novel Higher-Level Order

The paper's central claim is that genuine emergence arises when many simple components interact locally, generating new collective phenomena that cannot be extrapolated from the parts alone. Ffellonics demonstrates this with unusual clarity.

It begins with identical spheres obeying one local rule: symmetric nearest-neighbour attachment under free-energy minimisation. Through purely local interactions, qualitatively new higher-level structures appear — the tetrahedron at Level 3, the octahedron at Level 4, the icosahedron at Level 5. None of these forms is present in the individual spheres. They are genuine emergent structures that arise only when sufficient numbers of units interact under the governing rule.

By Level 12, the system reaches the 12-fold FCC/HCP lattice — a qualitatively distinct regime of maximal coordination and stable, infinitely extensible order. Each transition is a precise instance of what Philip Anderson called "more is different": quantitative increase in the number of interacting units produces qualitative leaps in symmetry, stability, and structural coherence.

Finite Depth and Lower-Dimensional Effective Theories

Krakauer et al. emphasise that true emergence replaces high-dimensional microscopic descriptions with lower-dimensional effective theories — compact descriptions of the higher-level order that do not require tracking every microscopic degree of freedom. Ffellonics does this exactly.

The high-dimensional configuration space of arbitrarily many spheres is compressed into a 12-level hierarchy. The final ground state at Level 12 is fully described by a single compact rule: the 12-fold coordination lattice. Once that state is reached, the system requires no further hierarchical levels — it extends indefinitely while remaining completely described by the same low-dimensional geometry.

This is the structure the paper is looking for: emergence that terminates in a stable, self-maintaining effective theory rather than requiring endless, unstructured scaling. The hierarchy has finite depth and infinite lateral extension — and the effective theory that describes the ground state is maximally simple.

Efficiency and Optimal Compression

The paper argues that genuine higher-level order is marked by increasing efficiency: more achieved with less. Ffellonics embodies this at every stage.

The entire 12-level structure is generated by one local rule. Each attachment is the lowest-free-energy move available, producing maximal coordination with minimal waste. By Level 12, the system has reached the global thermodynamic minimum — further hierarchical growth is not needed, and the structure maintains its low-energy state while extending indefinitely.

The natural, smooth pace of Ffellonic self-organisation is the observable signature of this efficiency. The system follows the steepest available descent through the free-energy landscape without detours, backtracking, or high-energy barriers. It achieves maximal global order through minimal local computation.

A Contrasting Case to LLMs

The paper is largely a critique of overclaiming about emergence in large language models. While LLMs display impressive scaling behaviours, the authors question whether these constitute true emergence or sophisticated pattern-matching at scale. The distinction matters: scaling that produces quantitatively better outputs within the same qualitative regime is not the same as the emergence of genuinely new higher-level structure.

Ffellonics offers a contrasting case. Its emergence is deterministic, geometrically explicit, thermodynamically grounded, and hierarchically coherent. It produces genuine higher-level order — the Platonic solids, the 12-fold lattice — without requiring massive parameter counts, statistical approximation, or post-hoc interpretation of what has emerged. The higher-level structures are visible, stable, and fully described by the effective theory they instantiate.

In this respect, Ffellonics functions as a reference model for the kind of emergence the paper defends — one that is rigorous, minimal, and produces interpretable higher-level structure with a definite endpoint.

Conclusion

The Krakauer et al. paper identifies two defining features of genuine emergence: qualitatively new higher-level order arising from local many-body interactions, and the replacement of high-dimensional microscopic descriptions by lower-dimensional effective theories. Ffellonics satisfies both criteria precisely and minimally.

Where large language models offer statistical scaling that may or may not qualify as true emergence, Ffellonics offers a deterministic geometric pathway from first contact to maximal relational coordination — finite in hierarchical depth, infinite in stable extension, and fully described at its ground state by a single compact effective theory.

In the broader conversation about what emergence in artificial systems should look like, Ffellonics serves as a useful benchmark: a minimal model that meets the rigorous criteria the paper sets out, while remaining fully interpretable and thermodynamically grounded.