Compiled, not interpreted
Your formulas aren't read one line at a time. They're compiled, turned into something the engine can run at full speed, the same way real software is built.
Serene on the surface. Underneath, the water never stops moving.
Every calm, instant moment in Novi (a number that updates the instant you change a driver, a model that opens without a spinner) rides on an engine working at full tilt out of sight. The peace you feel is engineered.
A swan glides across the water without a ripple. Below the surface, it is paddling hard. Novi is built the same way. The simplicity you experience (the stillness, the answer that's just there) is only the visible half. The other half is an engine running at full effort so you never have to think about it. Simplicity isn't the absence of machinery. It's machinery you never have to see.
The hardest thing to build is something that feels effortless to use.
Most tools keep your model on disk and fetch it a slice at a time. You ask, then you wait. SFMT doesn't. It treats the combined memory of a cluster of machines as one vast, superfast space, and keeps your whole model live inside it. Nothing is loaded from disk in the moment you need it, because it's already in memory. That's the difference between querying a database and working inside a data grid: there is nothing to wait for. Change a driver and the result is already there.
A new generation of in-memory engine: your model as a single, living calculation.
We won't open the whole engine room. The deepest mechanics are ours to keep. But the shape of it is simple to describe: formulas compiled rather than read line by line, two purpose-built machines doing the maths, and data laid out so it's ready the instant you reach for it.
Your formulas aren't read one line at a time. They're compiled, turned into something the engine can run at full speed, the same way real software is built.
A purpose-built virtual machine evaluates the calculation cells, the C-class, the instant their inputs change. No recalculate button, no waiting your turn.
A second virtual machine rolls values up through every hierarchy, the A-class, so totals are always current, never stitched together by hand.
Where it helps, data is held in columns, so the values you read together sit together, and the answer is assembled the moment you ask for it.
Beyond this, the engine's internals are proprietary, and deliberately so. What matters to you isn't how it works underneath, but how it feels on top: instant.
An engine this fast needs more than calculation around it. A set of sidecars travels alongside the engine: services that share its speed and sit right next to your model, instead of in some far-off server.
CLARA and the conversation live next to the model, so a question is answered against the live numbers, not a stale copy.
The moment a number moves or a task lands on you, you know. The engine pushes the change out as it happens.
Approvals, submissions and the steps of a process run as their own service, in step with the model they act on.
Turns the live model into a narrative (board commentary, a report, a Story) drawn straight from the current numbers.
Everything above the waterline is calm by design. Everything below it is the engine, working so that the calm holds. Both are always true at once. That is the whole idea.
Stillness. The part you were meant to notice.
Effort. The part you were never meant to see.
The engine is the part you were never meant to notice. The best way to understand it is to use what it powers. See it inside the Workspace →
The machinery is ours to build and keep. The stillness is yours to use.
The fastest way to understand SFMT is to watch Novi move, then forget the engine is there at all. See it work, and tell us about your team.