Loading...
Loading...
Every memory system remembers what you said. The NOMARK SDK infers how you reason from that history and compiles instructions every model you use should follow, before you type a word. It is the open-source engine underneath every NOMARK engagement and the studio's product portfolio.
Memory has been the default answer to AI personalisation for two years. Larger context windows. Retrieval-augmented generation. Long-term memory products. All of them solve the same problem: making the model remember what you said.
That's the wrong problem.
The gap isn't recall. The gap is translation, between the way you actually reason and the way the model defaults to answering. Memory tells the model you work in TypeScript; it doesn't tell the model to skip preamble, lead with working code, surface assumptions, or compress to conclusion. Those are instructions, not facts. And they don't survive a switch from Claude to ChatGPT to Cursor.
The NOMARK SDK is built around a different primitive: the compiled instruction set. Not what you said, but what the model should do with it.
Parses your conversation ledger across providers: Claude, ChatGPT, Gemini, Cursor, local models.
Extracts preference signals against a 10-dimension cognitive framework.
Resolves contradictions through promotion gates (staged, stable, rejected) so a bad-mood opinion doesn't rewrite your profile.
Compiles the result into instructions the model receives before you type.
Traces every rule back to quoted evidence in your own history. Reject any rule and it's gone.
Apache-2.0. Local-first. No account required for the open-source tier. A Pro tier (BSL 1.1) adds trust contracts, critical-field gates, audit trails, contradiction resolution, and team baselines. Pricing and the full feature breakdown live at nomark.ai.
The SDK is the most direct expression of NOMARK's operating belief: the outcome is the proof. Every other NOMARK product runs on the same primitive: what matters is what the model does, not what it remembers. Three essays explain the choices in more depth.
Prose specifications can't be verified. A constitutional stack needs a runtime, and a runtime needs structured data.
Agent personalisation usually means hard-coded rules. Treating preference as a weighted, evidence-tracked signal layer changes what agents can do.
Most status updates are designed to protect the sender. Ours are designed to inform the reader.
The product home is nomark.ai: install instructions, full documentation, pricing, the lot. If you want to understand the methodology before installing, the three essays above are the starting point.
Using the SDK in a real engagement and want to talk about the closed parts of the stack or a custom deployment? reece@nomark.au.