As of Friday, 05 June 2026, the Prometheus7 Research Institute is somewhere between the sixth and seventh rungs of a ten-step architectural program it calls the dimensional ladder. The fifth dimension — the substrate routing manifold shipped as the Tree of Life model family — was validated on 16 May 2026. The sixth, a router-over-callables primitive, was in its third training attempt as of 18 May. Whether that third attempt has resolved is not confirmed in materials available today.
The ladder is not a parameter-count schedule. Each rung adds a new compositional primitive — a new kind of operation the substrate algebra admits — rather than simply enlarging an existing model. The institute claims wall-clock time per generation stays in the seven-to-eleven hour band on its research hardware. If that holds, the cost of opening a new dimensional layer is roughly the cost of one training run, not the cost of new infrastructure.
The 6D primitive routes hidden state to one of several small neural sub-modules. The trunk specializes; the sub-modules specialize separately; the router decides which one contributes at each timestep. The institute describes this as opening a compositional surface where the substrate can recruit fine-grained specialists rather than handling all tasks uniformly from the trunk alone. Two prior attempts failed before the third run began.
The 7D primitive, scheduled for roughly the second week after 6D launch, routes over sets of sub-modules rather than individual ones — a coalition per token rather than a single specialist. The empirical question on record: does set composition add discriminative power beyond what a deeper 6D configuration would provide? If it does not, the institute says the ladder has reached its first plateau and 7D collapses back to 6D. No result on that question is available today.
The 8D primitive routes across grammars — selecting which callable vocabulary to operate in, not just which callable within one vocabulary. The institute's stated application is cross-domain queries that straddle mathematics and poetry, or theology and physics, without requiring post-hoc analysis to bridge them. Falsification mode: the router collapses to single-vocabulary operation because the training corpus doesn't reward cross-grammar behavior.
The 9D primitive routes across what the institute calls worlds — each world containing its own multiverse of grammars. The substrate becomes aware of multiple substrates as architectural objects rather than as implicit side effects of prior primitives.
The tenth primitive is the stated resolution point. The institute calls it the universal-unbinder. The claim is that at 10D the substrate holds all specifics in superposition and unpacks them through relation — mapped variously to the universal Turing machine, the holographic principle, the category of all categories, and Kolmogorov-minimal description depending on which tradition one prefers. The target date on record is August to September 2026, six to eight generations from the May validation of 5D.
Beyond 10D, the institute describes an 11D primitive — a space of universal objects, each holding all specifics under different relations — and a 12D relating principle that the roadmap says closes the ladder back to its own starting conditions by self-similarity. Both are marked research territory. The institute notes they cannot be completed by one person and frames them as work for a small community over years.
What is confirmed as of today: 5D shipped, 6D was in validation as of mid-May, and a schedule exists. Whether the third 6D attempt succeeded, and whether 7D training has begun, is not established in materials available at this dateline.