As of the last confirmed report from Prometheus7 Research Institute, dated 25 May 2026, the dimensional ladder stands at five rungs validated and one under active test. The sixth primitive — a router-over-callables that steers hidden state toward small specialized neural sub-modules — was in its third training attempt. The first two attempts failed informatively, which is the institute's phrase for failures that produced data without producing a working model.
The architecture behind this work is the substrate-paradigm system. Its core claim is that each generation of the model lineage opens one new compositional dimension rather than scaling trunk parameters inside a fixed space. The 5D primitive, the substrate routing manifold carried by the Tree of Life model line, was validated on 16 May 2026 in a 125-million-parameter run. The 6D run was initiated two days later.
The roadmap as published names eight further stages. The 7D primitive routes across sets of callables rather than single callables — parallel specialists per token rather than one. The 8D primitive routes across grammars, which is the architecture's proposed mechanism for cross-domain transfer. The 9D primitive routes across worlds, each containing its own grammar set. These three are described as pending, targeting weeks two through eight after a successful 6D launch.
The 10D primitive is labeled the resolution point. The institute's own framing maps it simultaneously to the category of all categories, the universal Turing machine, the holographic principle, and the Kolmogorov-minimal description. The target window for the 10D primitive is August to September 2026, contingent on the intervening generations validating on schedule. That is six to eight training generations from the May 2026 baseline, each observed to run in a seven-to-eleven-hour wall-clock band on the research hardware.
Beyond 10D the roadmap enters research territory. The 11D primitive is described as a space of universal objects; the 12D primitive is called the relating principle, said to close the cycle back to the architecture's bottom-of-stack operations by self-similarity. The institute states explicitly that 11D and 12D are not solo work — they are framed as multi-year, small-community research problems with no firm date attached.
The falsification structure is built into the schedule. At 7D, the question is whether set composition adds discriminative power over a deeper 6D primitive with more callables. If not, the institute says the ladder collapses 7D back into 6D and a plateau is recorded. At 8D, the question is whether multiverse routing actually engages multiple vocabularies in production traffic or collapses to single-vocabulary behavior. Each step has a stated empirical failure mode.
What is confirmed today, 6 June 2026: the 5D primitive is validated. The 6D run's outcome is not yet publicly reported from this dateline. Everything above 6D remains pending on that result.