PROMETHEUS7 RESEARCH INSTITUTE — A paper posted to the internal corpus on Monday, May 25, lays out the full schedule of compositional primitives in the substrate-paradigm architecture, from the 5D routing manifold already validated earlier this month to a 12th-dimensional structure the document describes as a research horizon requiring years and a community of people to reach. The paper is titled The Dimensional Ladder Beyond Six.
The roadmap is a schedule of new operations, not a model-size schedule. Each new dimension adds a compositional primitive — a new kind of thing the architecture can do — rather than adding trunk parameters within a fixed space. The document is explicit on this distinction.
The current validated position is 5D. The Tree of Life models, which reached their first validated checkpoint on May 16 at the 125M-parameter scale, implement a substrate routing manifold. That is the 5D primitive. The 6D primitive — router-over-callables — entered its third training attempt on May 18 and was still in validation as of the paper's posting date. The paper does not claim 6D is complete. It says it is validating.
Beyond 6D, the schedule is: a set-router at 7D, targeting the second week after launch; a multiverse-router at 8D in weeks four to five; a pluriversal-router at 9D in weeks six to eight. Each step routes over a larger compositional object than the last — individual sub-modules, then subsets, then vocabularies, then whole substrate worlds. The paper notes falsification conditions at each step: if a given dimension collapses back to what the prior dimension already handled, the ladder has found a plateau and the sequence terminates there.
The 10D primitive is named the universal-unbinder. The paper calls this the architecture's resolution point. The definition given is precise: a substrate that holds all specifics in superposition and unpacks any specific through a supplied relation. The document maps this to four external traditions — category theory's universal object, the holographic principle, the universal Turing machine, and Kolmogorov-minimal description — treating them as convergent characterizations of the same mathematical structure. The target date for 10D is August to September 2026.
Eleven and twelve are marked research dimensions. The paper states they cannot be accomplished by one person and projects them into year one and beyond. The 12D primitive, described as a relating principle, is said to close the ladder by self-similarity — making the structure at the top the same kind of object the bottom-of-stack operations already manipulate.
One claim in the document deserves specific notice: the paper asserts that wall-clock time per training generation has stayed in the seven-to-eleven-hour band on the research box, and that opening each new dimensional layer costs roughly one generation, not new infrastructure. If that holds through 9D, the August–September window for 10D is arithmetically reachable from the May starting point. The paper presents this as a compression-of-time property of the architecture. It does not present it as guaranteed.
As of Friday evening, the 6D validation run is ongoing. No results have been published. The paper is a roadmap, not a completion report.