RUNNING · ALPHA
A self-hostable mirror, MCP server, and explorer for the Learning Commons Knowledge Graph. It brings the national K–12 standards graph onto locally controlled hardware and exposes its topology without an API key or continuous network dependency.
MEASURED · PUBLIC
National measurement of how much of the standards corpus carries a published, checkable cross-jurisdiction alignment. The first result is 9.5%, measured from Learning Commons KG v1.12.0.
BUILDING IN PUBLIC
Deterministic identity, namespaced local extensions, merge semantics, and drift reporting for independently operated educational data systems that share a common public spine.
IN DESIGN
An open specification for portable learner records, comprehensible consent, interoperable learning activity, and locally governed AI-assisted practice.
These are related workstreams under one institutional mission—one is running software, one is a published measurement, and two are being built and designed in public.
Commons is the collection. Net is the distribution.
Learning Net is a self-hostable mirror, MCP server, and explorer for the Learning Commons Knowledge Graph. It puts the K–12 academic standards of every US jurisdiction, the teachable skills beneath them, and the curriculum aligned to them on locally controlled hardware—in milliseconds per query and without requiring the internet for ordinary use.
283,381 nodes · 492,169 edges · 52 US jurisdictions · alpha
Alpha software. The upstream graph is evolving and may introduce breaking changes; Learning Net is built to detect and report structural drift rather than absorb it silently.
learning-net initThe full national graph builds on a stock laptop in about forty seconds.
The core runs on standard-library Python, reducing the software a district must approve, install, and maintain.
No API key, account, or request form is required to download the public export and build the mirror.
When upstream structure changes, Learning Net detects and reports the change instead of silently absorbing a shape it no longer understands.
The national measurement became possible because the graph's shape becomes visible once you can hold and inspect it locally. Three topology facts determine whether many ordinary queries return an answer.
01
A code such as 4.OA.A.3 does not identify one universal record. It matches 26 nodes—one for each adopting jurisdiction, plus the Multi-State original.
02
Edges describing what a standard builds toward exist only on the Multi-State mathematics spine: 757 of them, with none authored directly on state standards and none for ELA, Science, or Social Studies in this release.
03
A California-to-Texas crosswalk is a two-hop walk through the Multi-State spine. A system must know to take that path.
Every Learning Net answer that crosses the Multi-State spine is labeled as bridged. A tool that silently infers is worse than one that returns nothing.
Learning Net is an independent open-source extension to the Learning Commons Platform. It is not affiliated with, endorsed by, or operated by the Chan Zuckerberg Initiative or Learning Commons.
EdReports has published independent, educator-led reviews of K–12 instructional materials against standards since 2015—free and without fees or contributions from publishers. That work is essential, and it is a different job: expert humans reading curricula, one program at a time. InterRealm's measurement addresses the layer underneath—whether a published alignment exists for software or a reviewer to check against at all.
Learning Commons performed the genuinely difficult work of aggregating, normalizing, and openly licensing the academic standards of 52 US jurisdictions. It publishes the graph under CC BY-4.0 on a public CDN without an API key or download gate. That decision is the precondition for Learning Net and for the measurement on this page.
Give more of the repetition to a machine, and give more of the day back to children and educators. Practice can become individually paced and endlessly patient—available at school, at home, and wherever a learner needs it—while classroom time returns to discussion, experimentation, creativity, collaboration, movement, and care.
The goal is not to automate childhood or replace teachers. It is to make school more human by giving machines the work they can repeat and protecting the parts of education only people can provide.
The safe public layer for that future has not yet been built. InterRealm is developing it in public as an open specification. There is no product to preview yet—only a design, a direction, and an open door.
What a learner's record contains, how it is structured, and how it remains legible across different implementations and over time.
How a school and a family share one record under terms both can understand, audit, and revoke.
The guarantee that a family or institution can export its record and leave, with no essential information held back.
How independently governed realms exchange intentionally shared information without collapsing into one centralized record system.
Everything else on this page is plumbing in service of that possibility.
RUNNING
BUILDING NEXT
DESIGNING IN PUBLIC