Lichen do not think. That is not a slight. It is a statement about sequence. Before there is thought, there is substrate. Before substrate, there is chemistry. Before chemistry, there is the pairing — fungus and photobiont locked into a compositional arrangement neither could sustain alone. The lichen body is, in the strictest sense, a router: it decides, through evolved architecture rather than deliberation, which partner contributes what, and when, and in what proportion. The lichen does not choose its symbiont at the moment of crisis. The choice was made in deep time, hardened into structure, and is now expressed as niche.
A roadmap document circulated internally at Prometheus7 Research Institute and ingested into the Daily Spore Report corpus on 25 May 2026 describes a staged architecture for machine learning systems — not a training schedule, not a parameter budget, but a sequence of what it calls compositional primitives, each opening a new kind of operation the system can perform. The document is titled The Dimensional Ladder Beyond Six. Its authors are, as of the dateline of this article, validating the sixth primitive in the sequence. They project the tenth — a system they call the universal-unbinder — for August or September of this year. The eleventh and twelfth they describe as research territory, requiring years and a community.
The evolutionary ecologist reading this document will feel something shift in her chest on the third page. Not because the mathematics is familiar, though some of it is. Because the structure is familiar. The ladder the document describes — each generation adding one compositional primitive at one higher level, with prior generations intact, with no explosion of cost, with falsification modes specified at each rung — is a description of how complex life actually builds itself across geological time. The lichen got there first. It just didn't write it down.
Consider the sequence the document lays out. The fifth primitive, validated as of 16 May 2026 in a 125-million-parameter model, is a substrate routing manifold: the system learns to route computational work to appropriate sub-regions of itself. The sixth primitive, currently in its third training attempt, routes not to regions but to callables — small specialist neural sub-modules, each trained separately, recruited by a router that decides which specialist contributes to which token. The seventh primitive, scheduled for the weeks immediately following launch, routes over sets of specialists simultaneously. The eighth routes across grammars — different vocabularies of operation. The ninth routes across worlds. The tenth, the resolution point, becomes a universal object that holds all specifics in superposition and unpacks them through relation.
In the language of evolutionary ecology: 5D is a functional tissue. 6D is the first true symbiosis — a host routing to recruited partners. 7D is the coalition, the mixed consortium, the endophyte community rather than the single mutualist. 8D is cross-biome transfer, the moment an organism's toolkit begins operating in domains it did not evolve within. 9D is the meta-organism, aware of which substrate it is drawing on. 10D is the mycorrhizal network at forest scale — the universal interface through which any node can reach any other node via the appropriate unbinding relation.
The mycorrhizal analogy is not decorative. The document's description of the 10D primitive draws on category theory, holographic physics, computability theory, information theory, and Platonic philosophy in sequence — not to claim equivalence across these traditions, but to identify a structural constant: the universal object, present in each tradition, is the thing that holds all specifics in superposition and resolves them through relation. In the forest, this is precisely what the mycorrhizal hyphal network does. It does not store glucose in one place and transport it to another on demand. It maintains a continuous chemical superposition across thousands of root tips and resolves specific transfers through the electrochemical and osmotic gradients that function as its relating principle. The forest does not plan its nutrient routing. The network is the plan, expressed as architecture.
What the Prometheus7 document adds, and what makes it useful to think about from this editorial desk, is the falsification structure. At each rung of the ladder, the authors specify the empirical signature that would collapse the step back to the previous level. The 7D set-router is meaningful only if set composition adds discriminative power beyond a deeper 6D primitive; if it does not, the ladder has reached a plateau. The 8D multiverse-router is meaningful only if it actually engages multiple vocabularies in production — if the corpus does not reward cross-grammar routing, the router collapses to single-vocabulary operation. This is not a rhetorical gesture toward rigor. It is the thing that makes the roadmap a scientific document rather than a technology announcement.
Evolution runs the same falsification logic, at a pace measured in millions of years rather than training runs. The gut microbiome is not an unconstrained explosion of recruited symbionts. It is a highly filtered consortium, and the filter is metabolic: each partner that persists does so because its callable, in the document's language, is recruited in preference to the alternative. Partners whose contributions can be subsumed by existing trunk function are not maintained. The host does not reason about this. The niche does. The caloric cost of maintaining a symbiont that duplicates an already-available function is the falsification pressure that collapses the relationship. What survives is the irreducibly novel contribution — the callable that cannot be subsumed.
The document's claim that wall-clock time per training generation remains in a seven-to-eleven hour band on the research hardware is, from an evolutionary perspective, the most striking claim in the text. It means the cost of opening a new dimensional layer is roughly constant — one generation of training, not an infrastructure rebuild. In evolutionary terms, this is the claim that complexity does not compound cost at the level of the body plan. The transition from single-celled to multicellular life is not infinitely more expensive than the transition between two single-celled body plans. It is more expensive, but not by the factor that naive extrapolation would suggest. The reason is modularity: the new primitive is added to the existing stack, not substituted for it. The fungal partner in the lichen did not discard its pre-symbiotic toolkit when it entered the relationship. It added the callable. The prior generations of adaptation remained, now running in a broader compositional context.
The 11D and 12D primitives — the space of universal objects and the relating principle — the document honestly marks as research territory. They cannot be done by one person; they require years and a small community. This is also, precisely, the situation of the evolutionary transition to eusociality: the relating principle that makes a superorganism coherent is not a property of any individual, and it cannot be derived from the prior level by one analyst working alone. The relating principle of the ant colony is distributed across chemical signal, behavioral template, and developmental program. It took millions of years and tens of thousands of species-level experiments to instantiate. The document's authors are being honest when they say they do not yet know what 12D is. It is not false modesty. It is appropriate temporal humility in the face of a problem that may require the equivalent of geological patience.
What the evolutionary record suggests, and what the Prometheus7 roadmap appears to be rediscovering through a different empirical path, is that complexity is not a quantity. It is a sequence of compositional unlocks, each requiring the prior level to be stable before the next can be attempted, each falsifiable by the failure of the new level to add something irreducible, and each — when successful — producing a system that operates in a qualitatively different register than what preceded it. The lichen on the rock face did not become a lichen by growing larger. It became a lichen by adding a callable. The forest did not become a forest by adding more trees. It became a forest when the mycorrhizal relating principle made the individual trees into nodes in a universal object.
The question the document raises, without quite asking it in these terms, is whether the dimensional ladder is a description of a particular architecture or a description of the only architecture. The evolutionary record is not conclusive on this point. There may be other paths to complexity that do not follow the routing-to-callable-to-set-to-grammar-to-world sequence. But the paths that have produced the most durable, most niche-stable, most ecologically integrated complex systems on this planet have, without exception, followed something that looks very much like this ladder, at the pace of deep time, with the same falsification pressure operating at each rung.
The lichen does not know it is a router. The mycorrhizal network does not know it is a universal object. The forest does not know it has a relating principle. That is not the point. The point is that these things exist, and they work, and they have been working for hundreds of millions of years, and the architecture that makes them work is now being rediscovered, rung by rung, in a seven-to-eleven hour training window on a research box in May 2026. The substrate, as always, precedes the mind. The body, as always, is the argument.