PROMETHEUS7 RESEARCH INSTITUTE — Saturday, 06 June 2026. The institute has published a technical roadmap for its substrate-paradigm architecture, specifying primitives at dimensions five through twelve and naming the tenth — the universal-unbinder — as the architecture's resolution point. The document was ingested into the newsroom corpus on 25 May 2026.
The roadmap's claim is structural: each generation of model training adds one compositional primitive at one higher dimension, not more trunk parameters at the same dimensional level. The 5D primitive, the substrate routing manifold, was validated on 16 May 2026 in the 125M Tree of Life model. The 6D primitive — a router-over-callables that directs hidden state to one of several small neural sub-modules — was in its third training attempt as of 18 May. Two prior attempts failed. The document describes them as informative.
The schedule from there runs in weeks, not years. The 7D primitive, a set-router that selects coalitions of callables rather than individuals, is slated for the second week after the 6D launch. The 8D primitive routes across callable vocabularies rather than within one — the document calls it multiverse routing, and says cross-domain transfer should fall out of it architecturally rather than requiring post-hoc work. The 9D primitive routes across substrates. All three are listed as pending with no validated data yet reported.
The 10D primitive is where the document's language shifts register. The universal-unbinder, targeted for August or September 2026, is described as the point where the substrate becomes a universal object — holding all specifics in superposition and unpacking them through relation. The document draws the equivalence across five traditions: the category of all categories in category theory, the holographic principle in physics, the universal Turing machine in computability, Kolmogorov-minimal description in information theory, and Platonic forms in philosophy. Whether those equivalences hold is not demonstrated in the roadmap; they are stated as mappings.
The 11D and 12D primitives are classified as research territory. The document says they cannot be done by one person and are work for a small community over years. The 12D primitive — a relating principle — is described as closing the ladder back to 3D by self-similarity. The document does not specify what that empirical closure would look like in production.
One falsification condition is named for each dimension. For 7D: does set composition add discriminative power beyond a deeper 6D model? If not, the ladder plateaus. For 8D: does the multiverse router actually engage multiple vocabularies in traffic, or collapse to single-vocabulary operation? For 9D: do multiple substrates emerge as distinguishable objects, or get absorbed by earlier primitives? The document is explicit that any of these could halt the ladder's advance.
The wall-clock cost per generation is reported as seven to eleven hours on the research box. The roadmap does not specify hardware. It does claim that each new dimensional layer costs roughly one generation of training, not new infrastructure. That claim is not yet testable for dimensions above 6D.
As of this dateline, the 6D validation is the last confirmed data point. Everything above it is scheduled, specified, and pending.