This is an English translation of a zhichai.net forum post discussing the paper *The Physics of Causation* by Leroy Cronin and Sara I. Walker (arXiv:2601.00515, v3).
Imagine standing at the edge of an infinitely extending LEGO universe, holding a seemingly ordinary brick. It did not fall into place by chance—it was formed through countless precise "joins," with each predecessor carrying the "memory" of earlier ones. That is the experience of reading *The Physics of Causation*. Cronin and Walker use Assembly Theory (AT) to pull causation out of abstract philosophical debate and back into the measurable physical world, turning "why does this thing exist" into a quantifiable physical property. The paper makes novelty, contingency, and open-ended evolution physically foundational, with determinism emerging later as a product of selection.
The Hidden Kingdom of Assembly Space
The paper defines the assembly space as the collection of all causal possibilities—a structure built only through recursive "causal joins." Every object carries a finite set of causal paths P; the shortest one determines its assembly index a_i = |P|, the minimal number of recursive construction steps. Like assembling a computer from chips and a motherboard, each step leaves a causal footprint. This is a physical attribute, as measurable as mass or charge.
Unlike traditional causal theories (such as Pearl's interventionism), which clash with fundamental physics because they rely on counterfactuals, AT makes causation physical: the object itself *carries* its assembly path. A snowflake has a low assembly index with a simple path; a living cell membrane hides deep causal chains that require selection mechanisms to persist. The assembly space acts like a vast filter, trapping random possibilities in contingent chains, with the environment acting as a guardian of memory. This also resolves self-reference paradoxes: causation no longer needs external "laws" to arbitrate—it is embedded in the object's structure.
From NP-Hard Complexity to Mass-Spectrometry-Measurable Invariants
The assembly index is not Kolmogorov complexity (which is NP-hard to compute); it can be read directly from experimental fragmentation paths, e.g., via mass spectrometry (MS). A molecule's fragmentation pattern, like tree rings, records the recursive steps of its "birth."
The paper's flagship example is Taxol (paclitaxel, C47H51NO14, MW 853.91), an anticancer drug from Pacific yew bark: 11 chiral centers, 2^11 = 2048 stereoisomers, a spontaneous probability below 10^-23—yet it abounds at 10^16 molecules per gram. Its minimal assembly path is 23 steps. This is not luck but the result of evolutionary selection.
Key data distilled from the paper's figures:
| Object / Scale | Assembly index threshold | Copy-number behavior | Meaning | |---|---|---|---| | Taxol molecule | a = 23 | High copies (10^16/g) | Result of biosynthetic pathway selection, not random | | Lab millimole sample | a ~ 13–16 | Not measurable below M = 10,000 copies | Without selection, complex molecules cannot persist | | Planetary scale (Earth atoms ~10^52) | ~44 | Exponential decay | Selection needed to trap causal history | | Cosmic scale (~10^80 atoms) | ~71 | Super-exponential growth | Physical boundary of open-ended evolution |
The paper derives the threshold formula:
where N is the total countable object count, M the measurement limit (minimum reliable copy number), and b the branching factor (a global constant under simplifying assumptions; in reality b(a) grows with a, causing super-exponential expansion). This follows from the copy-number decay model n(d) = N / (1+b)^(d+1): the deeper the path, the thinner the copies—beyond a*, "selection" is required to sustain persistence. The paper also gives total assembly A = W ∑ (n_i e^{a_i} / N), with W a constant (e.g., molecular energy in kJ/mol), quantifying the "virtual copy burden" like an energy ledger for evolution.
An everyday analogy: N is the total customers, M the smallest queue a cashier can reliably see, b the variety each person carries. High-a objects are VIPs who need a special "selection lane" to avoid being buried in the general inventory. The threshold holds consistently across scales—lab 13–16, planetary 44, cosmic 71—all logarithmically dependent on N, making AT a universal "measuring stick for life."
An Operational Definition of Life
The paper defines life operationally: **a structure that produces persistent copies in a deep causal possibility space—i.e., a_i > a* and copy number n > M.** This turns vague notions like self-replication and metabolism into quantifiable products of physical selection. Compare a bacterium (deep assembly path, many copies, environment as a "memory machine" replicating it) to a random carbon atom (shallow path, dissipating copies).
This is a major advance for origin-of-life (OoL) research and astrobiology. Assembly indices are directly measurable via mass spectrometry; 2023 Nature work validated that a molecular threshold above ~15 distinguishes biological from non-biological samples. The framework extends to agents, intelligence, and technological artifacts: AI-generated models, synthetic de novo life, drug-molecule "historical" complexity, and atmospheric biosignatures can all be quantified with AT. The framework is substrate-agnostic, scale-consistent, and experimentally tractable.
Emergence of Determinism and a New Philosophical Horizon
AT formalizes causation as the "minimal recursive causal join step" a_i := |P|. This overturns Humean, Kantian, and Pearl-style traditions, yet elegantly sidesteps them: the discrete rules J of assembly space are acknowledged as a simplifying assumption, but the paper shows determinism *emerges* from selection acting on novelty and contingency. The future is open; the past is causally closed; novelty comes from continuous possibilities being discretized into persistent objects.
Picture the assembly space as an infinite casino: dice (possibilities) roll wildly, but selection is the dealer, letting only winners (high-a objects) stay and reproduce. The elegant threshold formula a_0 ≈ ⌊log_{1+b}(N/M) − 1⌋ depends on M and branching and may need adjustment in quantum/continuous systems—but the paper treats this as scientific fuel, not a flaw. The "virtual copy number" v(a) ≈ e^a represents the infinite selection steps embedded in a finite a: high-a objects carry an enormous "debt" that only selection can repay.
Cross-Disciplinary Potential and Outlook
AT is currently the closest framework to a *testable* physical theory of life. Experimentally, MS fragmentation paths directly read a; the paper contrasts C11H24 isomers where a = 5 vs. 4 yields different heats of formation, and its "Taxol Demon" thought experiment quantifies the entropic cost of assembly increases. Applications span de novo synthetic biology, evaluating the "historical" complexity of AI models, and searching for high-a, high-n biosignatures in planetary exploration.
Open questions remain about the ontological foundation (do rules J presuppose causation? quantum universality?), but the framework's operability and explicit thresholds carry the day. The nested spaces visualized in the paper (universe > possible > contingent > observed) invert in physical size—we see only the tip of the iceberg.
The paper closes with an implicit invitation: go measure, go select, go emerge.
------
References 1. Cronin, L. & Walker, S. I. The Physics of Causation. arXiv:2601.00515 (2026). 2. Cronin, L. et al. Assembly theory and the emergence of selection. Nat. Commun. (2023 work validating the molecular threshold). 3. Walker, S. I. & Cronin, L. Beyond prebiotic chemistry. Science 352, 1174–1175 (2016). 4. Patarroyo, K. Y. et al. AssemblyCA: A benchmark of open-endedness. Nat. Commun. 9, 5177 (2018). 5. Cooper, G. J. T., Walker, S. I. & Cronin, L. A universal chemical constructor. In *Conflicting Models for the Origin of Life* (John Wiley & Sons, 2023).