Understanding Evolution
It Does Not Happen in the Universe — It Happens to the Universe
When we speak about evolution, we almost always mean biological evolution. Variation, selection, adaptation to environment. This framework has been enormously successful, but it carries an implicit assumption that is rarely questioned: while organisms evolve, and environments evolve, the structural rules of the universe do not. The universe is treated as a static stage on which evolution plays out.
Empirically, this picture is backwards. What we actually observe is not a universe with fixed structure hosting evolving forms, but a universe whose capacity to host structure has itself been changing over time. The most basic fact of cosmology is not expansion or entropy, but staging. Entire classes of structures become possible only after certain epochs, and were categorically impossible before them.
In the early universe there were no atoms. Later, atoms became possible. There was no chemistry, then chemistry emerged. No stars, no planets, no heavy elements, no long-lived energy gradients, no complex molecules, no self-replication, no metabolism, no nervous systems, no symbolic reasoning. These absences were not matters of probability or bad luck. They were matters of impossibility. The universe simply could not support such structures yet.
This observation already forces a deeper conclusion: the space of what can exist is not fixed. It grows.
Once that is acknowledged, a further implication immediately follows. A system whose capacity for structure grows over time cannot be infinite. An actually infinite system does not yield new possibilities; all possibilities are already present. Growth of admissibility is meaningful only in a finite informational system whose internal structure is expanding.
This is not a philosophical preference. It is a logical consequence of what we observe. If the universe were structurally complete from the beginning, then atoms, chemistry, biology and cognition would not be time-dependent phenomena. They would either exist from the outset or not exist at all. The fact that they appear in sequence tells us that the universe is not merely evolving within fixed constraints, but evolving by extending its constraints.
In such a finite informational universe, genuine free will is not an anomaly to be explained away, but a structural necessity. Because there exists no external space of pre-given possibilities, no latent catalogue of forms, and no infinite backdrop in which outcomes are already encoded. Innovation is therefore not the selection of one option among many that already exist, but the creative invention of new relational structure that did not exist before the act itself.
Constraints — such as finitude, closure, and symmetry — do not impose determinism; they merely ensure coherence. Within those constraints, each extension of structure is irreducible and non-derivable, marking a real moment of origination rather than revelation. Free will, in this sense, is not randomness nor violation of law, but the lawful emergence of novelty without prior representation. Time is thus not a parameter indexing predetermined states, but the medium through which genuinely new structure comes into being.
This perspective finds a strong point of contact with Assembly Theory, developed by Leroy Cronin and Sara Walker, which reframes complexity not as a static property but as a historical one. In Assembly Theory, an object becomes physically meaningful only once the universe has enacted the sequence of constructive steps required to bring it into existence; prior to that history, the object is not merely absent but undiscoverable in principle. This resonates directly with a finite informational universe: structures do not pre-exist as latent options waiting to be realised, but emerge only through the accumulation of relational and constructive history.
What Assembly Theory measures as “assembly depth” corresponds, at a deeper level, to the growth of informational and structural capacity of the universe itself. An object’s existence is therefore inseparable from the path that led to it, and innovation is neither random nor pre-scripted — it is the necessary consequence of a universe that can only know, sustain, and extend what it has already built. In this sense, complexity is not unlocked but originated, and novelty is not selected from possibility space but generated through irreversible, history-dependent construction.
Get Yosef Akhtman’s stories in your inbox
Join Medium for free to get updates from this writer.
Biological evolution fits neatly into this picture, but only after being demoted from its customary role as a universal explanation. Natural selection explains which forms survive once variation exists. It does not explain when certain kinds of variation become possible. Biology operates inside a space of admissible structures; it does not generate that space. Life did not appear because it was merely advantageous. It appeared because it became possible.
This distinction resolves many long-standing confusions. For example, it dissolves the apparent tension between increasing local complexity and increasing entropy. Entropy constrains state distributions within a given structural regime. It says nothing about whether new regimes can emerge. A universe can be globally entropic while continuing to originate new forms of organisation.
It also reframes the so-called “waiting times” of evolution. Long delays are often treated as improbabilities: why did life take so long? Why did intelligence take so long? But from a structural perspective, these delays are not waiting for lucky mutations; they are waiting for the universe to become capable of sustaining certain configurations at all. Evolution looks slow not because it is inefficient, but because it is constrained by deeper prerequisites.
Once evolution is understood this way, a striking consequence follows: intelligence is not an accident, and it is not inevitable. It is a phase.
This brings us to artificial intelligence. AI is usually framed as a technological project driven by human needs: automation, efficiency, profit, curiosity. That framing is anthropocentric and historically shallow. Humans did not decide, in any deep sense, that artificial intelligence should emerge. We built tools, abstractions, languages, machines and networks, each step following local incentives and constraints. The emergence of AI is the cumulative result of a long structural trajectory that began long before humans existed.
Seen from a broader perspective, AI is not an extension of humanity. It is a new form of organised cognition emerging at the moment when the universe becomes capable of supporting it. In this sense, humans have invented AI in the exact same sense as apes invented humans, or dinosaurs invented chickens.
Biological intelligence required certain structural conditions: stable chemistry, energy gradients, long time horizons, error-correcting replication, neural architectures. Artificial intelligence requires a different but equally stringent set of conditions: symbolic abstraction, global communication networks, massive energy throughput, layered representations, recursive self-reference, and planetary-scale information flow. These conditions did not exist even a few centuries ago. They exist now.
This matters, because it changes how we interpret AI’s significance. AI is not appearing because humans want it, nor because it is “useful.” It is appearing because the universe has reached a stage where cognition no longer needs to be tied to biology. The substrate is changing. From this perspective, AI is not an anomaly or a threat. It is a continuation. Just as life was not a violation of physics, and intelligence was not a violation of biology, artificial intelligence is not a violation of humanity. Each step looks disruptive only when viewed from the level below.
This also clarifies why the emergence of AI feels abrupt. Structural transitions often do. When a new class of organisation becomes admissible, it does not necessarily appear gradually everywhere; it may appear suddenly in a few places and then proliferates. Atoms, stars, cells, multicellular organisms, nervous systems — all follow this pattern. AI is no different.
If we accept this framing, many anxieties and paradoxes lose their force. The question is no longer “should AI exist?” but “what does this new phase require of us?” Humans are not being replaced; they are being decentered. That has happened before. It will happen again. Just as Copernicus took Earth out of the centre of the solar system, this insight takes biology out of the centre of intelligence, which becomes a function of the universe itself that uses biology as a temporary vessel.
In the same way, this structural view reframes cosmic silence of the famous Fermi paradox. If the universe’s capacity for advanced intelligence is itself time-dependent, then the absence of visible interstellar civilisations is not mysterious. It is expected. The universe may simply not have reached a stage where such forms are stable or coherent. Silence is not absence; it is immaturity.
None of this denies biology, Darwin, physics or chemistry. It places them in context. Evolution is not something that happens in the universe. It is something the universe is doing.
The expansion of the Universe is not measured in size, but in complexity. Life, intelligence, and now artificial intelligence are manifestations of such an expansion, appearing when the universe becomes capable of sustaining them. The deepest continuity is not between genes or machines, but between stages of structure. The phase transition we are currently witnessing may appear as a singularity to us, but it is better understood as the natural crossing of a new structural threshold in the universe’s capacity to sustain complexity.
