diff --git a/ssm.md b/ssm.md
index 029d3fc4..d729ed51 100644
--- a/ssm.md
+++ b/ssm.md
@@ -10,7 +10,7 @@ $$
 (\mathbf{x}_0,\ \{\mathbf{u}_t\},\ \{\mathbf{d}_t\}) \ \longmapsto\ \{\mathbf{x}_t,\ \mathbf{y}_t\}_{t=0:T},
 $$
 
-where $\mathbf{u}_t$ are **controls** we set, $\mathbf{d}_t$ are **exogenous drivers** we do not control (weather, inflow, demand), $\mathbf{x}_t$ are internal **system variables**, and $\mathbf{y}_t$ are **observations**. The split between $\mathbf{u}$ and $\mathbf{d}$ is practical: it separates what we can act on from what we must accommodate.
+where $\mathbf{u}_t$ are **controls** we set, $\mathbf{d}_t$ are **exogenous drivers** we do not control (weather, inflow, demand), $\mathbf{x}_t$ are internal **system variables**, and $\mathbf{y}_t$ are **observations**. The split between $\mathbf{x}$ and $\mathbf{d}$ is practical: it separates what we can act on from what we must accommodate.
 
 Two design pressures shape such models: