Knowing when not to use it: the envelope

A tool that claims to win everywhere is a tool you cannot trust. Stateful Wave Computing ships with the opposite promise: it tells you, up front, when to keep what you already have.

Three regimes where it leads

SWC’s advantage is concentrated, not universal. In simulation it leads in three places: tight budgets, where you cannot afford to spend measurements estimating gradients; drift, where the target moves and one measurement per round must track it; and anytime, real-time loops, where you need a usable configuration every round and a gradient method cannot assemble a step before the target drifts away.

Three regimes where it declines

Just as important are the places it does not help. On static problems with ample budget, a well-run classical optimizer reaches the same answer. With a hand-tuned PID and a good calibration, you already have what you need. On tightly coupled plants with no calibration, there is no clean residual to exploit. In each of these the runtime returns DECLINE — it tells you to keep your current method.

Why honest declines are a feature

For a buyer, a method that knows its own envelope is worth more than one that promises everything. It means you can deploy it exactly where it pays and nowhere it does not, without babysitting it or discovering the hard way that it was the wrong tool. The decline is not a weakness in the pitch; it is the reason the rest of the pitch is credible.

How the envelope check works in practice

Before you commit budget, the client can answer the question for you: a single call returns USE, MARGINAL, or DECLINE for the regime you are in. You find out whether this is your fight before you spend a measurement on it.