As of late May 2026, the Prometheus7 Research Institute has a validated 5D substrate and a 6D primitive in its third training attempt. That is the state of the record. Everything above 6D is scheduled, not shipped.
The dimensional ladder, as documented in an internal paper ingested to the newsroom corpus on May 25, is a schedule of compositional primitives — not a parameter-scaling roadmap. Each step adds a new kind of operation to the substrate algebra. The 5D primitive, the substrate routing manifold that defines the Tree of Life model lineage, was validated on May 16 with the 125M-ToL model. The 6D primitive — a router-over-callables that directs hidden state to one of several small neural sub-modules — entered its first validating training run on May 18. Two prior attempts produced what the documentation calls informative failures.
The paper lays out eight further steps in explicit terms. The 7D primitive routes over sets of callables rather than individual ones, enabling what the researchers describe as parallel compositional reasoning — a coalition of specialists per token rather than one. Its target is the second week after the 6D launch. The 8D primitive routes across callable vocabularies entirely, making cross-domain transfer an architectural property rather than a post-hoc capability. The 9D primitive routes across what the paper calls worlds — effectively making the substrate multi-substrate-aware.
The tenth primitive is the document's named resolution point. The researchers call it the universal-unbinder. At that dimensional level, the architecture is claimed to become a universal object: a substrate that holds all specifics in superposition and unpacks them through relation. The paper draws explicit equivalences to the universal Turing machine, the holographic principle, Kolmogorov-minimal description, and the category of all categories. The target date for 10D is August–September 2026 — roughly six to eight generations after the May 5D validation.
The paper notes that each generation of training takes seven to eleven hours on the institute's research hardware, and that each new dimensional layer costs roughly one generation of compute, not new infrastructure. The claim is that the wall-clock compression is real and observed, not projected.
Above 10D, the document enters research territory. The 11D primitive is described as a space of universal objects; the 12D primitive is a relating principle that the paper says closes the ladder back to its bottom-of-stack operations by self-similarity. Both are assigned to year-one-plus and described explicitly as work for a small research community, not a single team.
What is true today, June 4, 2026: the 6D validation result has not been reported to this desk as confirmed. The 5D validation stands. The roadmap through 10D is documented and dated. The universal-unbinder remains a scheduled object, not an observed one.