E31 SUBSTRATE
Seven services that turn local compute into a system people can actually use — and that an accreditor can actually review.
E31.AI · Substrate has a dedicated platform site with service specifications and an engineering contact.
Visit e31.aiWHAT IT IS
Substrate is the layer between an appliance and the applications people run on it. Underneath sits compute, storage, and an operating system. Above sits whatever your teams build. Substrate is everything in between: the primitives that make local models useful and the controls that make them accountable.
The division of labor is deliberate. Your application brings the agents, the workflows, and the domain knowledge. Substrate brings memory, retrieval, serving, tuning, tracing, integrity, and governed connectivity — the parts every serious deployment needs and nobody wants to rebuild. Because it is preloaded and sealed onto the appliance, the stack that is reviewed is the stack that runs.
ARCHITECTURE
Each layer is named for what it does. Implementations are selected per deployment and are not part of the interface you build against.
SEVEN SERVICES
Persistent semantic memory so agents stay coherent across sessions.
Agents forget between sessions. Memory gives them a durable record of what has already been established — entities, decisions, and prior conclusions. It persists on the appliance and is never synchronized off the box.
Grounded answers over your own documents, with citations.
An answer is only useful if it can be checked. Retrieval indexes your document holdings on the appliance and grounds each response in specific passages. Every claim carries a citation back to the source material.
Local model serving behind a stable, hardware-independent API.
Models change. The interface should not. Inference serves open-weight models from the appliance behind one stable API, so applications are written once and survive changes to both the model set and the hardware beneath it.
Fine-tune open models on your data, entirely on-premise.
General models do not speak your domain. Adaptation tunes open-weight models on your own material, on the appliance, with nothing leaving the enclosure. The resulting weights belong to you and stay where they were produced.
Full tracing and audit records for every AI operation.
A system that cannot be audited cannot be accredited. Evaluation records each prompt, retrieval, and response as a traceable event. The record is written locally and is designed to support review by your own oversight functions.
Continuous integrity verification and tamper-evident logging.
Trust in an appliance is a claim that has to be re-proven. Watchman continuously verifies the software image against its signatures and monitors the enclosure's tamper indicators. Deviations are written to an append-only record.
Policy-governed connectivity to on-premise systems. Nothing else.
Useful AI has to reach the systems where the work happens. Local Bridge connects the appliance to on-premise systems over explicitly declared, policy-governed routes. No default path to the public internet exists.
MODELS
Every model runs on the appliance. Weights are open, held on local storage, and executed by local compute. There is no per-token cost because there is no metered service, and no data egress because there is nowhere for the data to go. Prompts, documents, and outputs stay inside the enclosure for their entire life.
The model set is selected per deployment against the mission and the tier of hardware, then fixed at provisioning so behavior is reproducible. Refreshes arrive as signed update bundles, verified before they are applied and recorded when they are. Nothing changes on the appliance without a signature and an entry in the log.
UPDATES
An air-gapped appliance still has to be maintained. Substrate handles that with signed release bundles: a complete, versioned set of software and model artifacts, assembled and signed by Element 31, then carried across the gap on physical media. The appliance verifies the signature and the contents before anything is applied, refuses bundles it cannot verify, and records what was installed and when. Updates never require the appliance to reach out, and no update path exists that bypasses the signature.
We will walk the service model, the interfaces, and the update process end to end.