Sormena is a software company in Salt Lake City building the layer underneath teaching software: a prerequisite graph of what a learner must understand before anything else makes sense, and assessments that are graded by deterministic code rather than by a language model.
What we build
A prerequisite-graph curriculum engine. State a goal in plain language. The system resolves it to a concept in a directed graph, walks everything that concept depends on, and returns a curriculum ordered by mastery; 652 of 2,979 concepts carry a deterministic, machine-graded check today (105 of 345 in the featured course), and the rest a definition, most with a worked example. Language models propose structure; deterministic code decides what is correct. No grade of record is ever produced by a model.
A workforce and occupational data layer, in development. A reproducible database designed to unify federal occupational data — occupational profiles, wage and employment projections, the standard occupation, industry and instructional-program taxonomies, and five official crosswalks including the military occupation crosswalk — under stable identifiers, engineered so that unknown renders as unknown. It is not yet built.
What is built today
- 676 of 676 grader specifications pass a deterministic self-consistency check; answer keys are not yet independently reviewed.
- A 242,370-item standards corpus.
- A year-long high-school course authored as a concept graph: 345 concepts, 446 prerequisite edges, 1,018 learning objectives, built and taught by the founder in a classroom.
This is pre-pilot infrastructure. It has no users yet, and that is the honest state of it.
Where it runs
The engine runs locally today on commodity hardware. The next stage is cloud: generating graphs across a large corpus, serving assessments, and running the extraction pipeline at a scale a single machine cannot reach.
Contact
founder@sormena.io · Sormena LLC, a Utah limited liability company · Salt Lake City, Utah