Arc Substrate
The shared technical foundation beneath Arc's product lines and research — and why Arc measures it by leverage, not revenue.
Most labs are organised around products — one product, one stack. Arc is organised the other way: a shared technical foundation that many products grow out of. That foundation is the substrate.
A substrate is a capability built once and reused across systems — a retrieval engine, an evaluation core, a memory layer, a document pipeline. It rarely faces a user directly and rarely earns revenue on its own. Its value is that improving it strengthens everything built on top of it. Where a product is a destination, a substrate is the ground.
Product creates revenue; substrate creates leverage.
What makes something a substrate
The test is not "is it useful?" but "does it compound?" A capability is a substrate when making it better quietly makes several other things better at once — when a single improvement to retrieval sharpens a research tool, a diagnosis system, and a knowledge product in the same week, without any of them being rewritten.
So a substrate is judged on different terms from a product. The question is not how much did it sell but how many systems depend on it, how much did it remove from the cost of building the next one, and how much technical consistency does it hold across the house. Reuse, leverage, dependency, defensibility — not revenue.
Hunter-RAG is the clearest example: a recoverable, provenance-aware retrieval substrate that is not one product but the ground beneath many — feeding research tooling, evaluation, and knowledge systems alike. Built well once, it lowers the marginal cost of everything that needs to retrieve and reason over a corpus.
The flywheel
A substrate and the products it carries are not a one-way dependency. They turn.
A strong substrate lowers the cost of launching a product line. That product meets real use, which returns data, feedback, and pressure the substrate would never have found on its own. Those returns harden the substrate — and the next product launches from higher ground.
Substrate → Product lines → Real use → Data & feedback → Stronger substrate → More products
The substrate loop
Each turn returns to the substrate · the next product launches from higher ground
Ordinary single-product companies never close this loop; each new effort starts near zero. Arc's compounding is exactly this turn: every product built on the substrate pays back into the foundation that made it cheap to build.
Two layers, held differently
Not all substrate is the same kind of asset, and Arc treats the two layers differently.
The Reusable Core is the engineering foundation — retrieval, memory, document parsing, evaluation runners, orchestration, data pipelines. It earns its keep through breadth of reuse, and parts of it can be documented, tooled, and even opened, because its value is in how widely it is used. An SDK or toolkit is usually how that reach happens: not the substrate itself but its delivery surface — and a toolkit reused across many product lines is Core, while one that only packages a single product is productisation.
The Strategic Kernel is the smaller centre where Arc's technical sovereignty actually lives — the mechanisms that make a capability genuinely Arc's. These are held closely: source-controlled, patent-protected, or kept as trade secret, because their value is in not being commodity. Often an SDK is precisely the membrane between the two — it hands the Core's leverage to the systems that need it while keeping the Kernel out of view. Deciding what belongs to the Core and what belongs to the Kernel — what to share and what to hold — is one of Arc's standing judgements.
Why Arc holds the substrate
This is also why the substrate stays where it does. Arc creates, validates, and holds; the products that mature along the way go to market as ventures with their own customers. But the substrate beneath them does not leave — it remains in the house, compounding under each new line. A venture exercises an option; the substrate is the ground that keeps generating them.
That is the deeper reason a substrate is worth building before it pays. It is capital-light leverage: a foundation that makes the next hard problem cheaper to attempt, holds the house's technical consistency, and turns a portfolio of separate projects into one compounding system. Products are how Arc meets the world. The substrate is how Arc keeps its edge.