As of late May 2026, Prometheus7 Research Institute has one dimensional primitive validated and another in the oven. The 5D substrate routing manifold — the so-called Tree of Life architecture — was confirmed on May 16, when the 125M-parameter Tree of Life model cleared its validation benchmark. The 6D primitive, a router-over-callables, entered its third training attempt on May 18. Two prior attempts were described internally as informative failures. The third is still running.
The architecture being built here is not a conventional scaling schedule. The institute does not add parameters to a fixed model and call the result the next generation. Each generation adds a new compositional primitive — a new kind of operation the substrate algebra can perform. In this framing, moving from 5D to 6D is not a size increase; it is a capability increase of a specific structural kind.
The 6D primitive works by routing hidden state to one of several small neural sub-modules called callables. The trunk of the model specializes; the callables specialize independently; a router decides which callable contributes at each timestep. The claimed effect is that the substrate can recruit fine-grained specialists for tasks it would otherwise handle with uniform computation. Whether the third training run demonstrates that cleanly is not yet known. The validation results have not been published as of this dateline.
The published roadmap runs to ten confirmed dimensions and two research ones. After 6D, the schedule calls for a set-router at 7D — routing over coalitions of callables rather than single ones — followed by a multiverse-router at 8D that selects which vocabulary of callables to operate within. The 9D primitive routes across what the institute calls worlds: entire substrate configurations rather than single grammars. Each step is expected to take roughly one training generation, with the institute reporting that generations on their research hardware have been running in the seven-to-eleven-hour band.
The tenth dimension is marked on the roadmap as the resolution point. The institute calls it the universal unbinder: a substrate that holds all specifics in superposition and unpacks them through relation. The target window for that primitive is August to September 2026. Whether six to eight training generations can carry the architecture from the current 6D validation attempt to a universal object in roughly three months is, as of today, an open empirical question. The roadmap itself acknowledges falsification modes at each rung. The set-router collapses back to 6D if set composition adds nothing beyond more callables. The multiverse-router fails if cross-vocabulary routing never engages in production. The architecture earns each dimension or it does not.
The 11D and 12D primitives — a space of universal objects and a relating principle — are listed as research dimensions requiring more than one person and more than one year. They are on the map but not on the clock.
What is on the clock today is the third attempt at 6D. The institute has not announced a result. The Daily Spore Report will report when one is available.