See your step count over time

You feel like you've been moving less since the cut started — work has been heavier, evenings shorter. You ask the agent to draw the step curve over the cut so you can see if the lifestyle floor has actually dropped or if it just feels that way.

What the agent needs to hear

Behind your sentence, the agent works backwards to two pieces: the window and the cadence. The window defaults to the last three months — long enough for a trend to mean something, short enough to reflect the block you’re in. The cadence defaults to daily with a 7-day rolling average overlaid on top.

Daily reads as noise without the average — one big walk spikes the line, one sedentary day drops it. The rolling average is what carries the trend. Override the cadence when you want a different read: « weekly only » for a smoother curve, « daily, no smoothing » to inspect specific days.

Windows and smoothing

Windows follow the same pattern as the weight trend. Named ranges land cleanly — « since the cut started », « all time ». Relative phrases work — « last 6 weeks », « the past month ». Explicit dates bound it precisely when you need a specific block.

Smoothing has three modes. The rolling average is the default — a 7-day window smooths the daily spikes from one big walk or one sedentary day, and what’s left is the trend underneath. Weekly aggregates collapse each week into one point — a 7-day sum or average — and read cleanest over long windows. Raw daily shows every day as one point, useful when you’re inspecting specific dates or looking for the outlier that pulled the average.

What the step curve tells you vs the cardio curve

Steps and cardio answer different questions. Cardio is intentional movement — sessions you logged as workouts. Steps are lifestyle movement — everything else, the walks between meetings and the evening loop around the block.

Together they read your total activity. A week with zero cardio but 12k daily steps reads more active than a week with three cardio sessions and 4k daily steps. The step curve is most useful as a baseline floor — if it drops below 6k–7k consistently, sedentary lifestyle is creeping in regardless of what your cardio sessions look like. The cardio curve is the spike on top; the step curve is the floor underneath.

When the agent gets it wrong

Three failure modes show up. Missing days drag the average down — if auto-sync had gaps, those days read as zeros and the smoothed line dips for reasons that aren’t real. Fill them before reading the trend: « show me which days are missing in the last month », then log the values back in.

The rolling average sometimes obscures a real change. A sharp two-week dip during a stressful work block looks like noise in a 30-day window. Narrow the cadence — « weekly, last 8 weeks » or raw daily over that block — and the change resurfaces.

Travel weeks inflate or deflate the average depending on the trip. Asking for « the same window but excluding the travel week » surfaces what the home baseline actually looks like.

What makes the step trend worth reading

Three things decide whether this step curve tells you something useful: the log series is dense enough to be a series (auto-sync gaps look like sedentary days — confirm coverage before trusting the curve), the smoothing matches the question (rolling average for the trend, raw daily for specific-day investigation, weekly for the long view), and steps are read alongside cardio, not instead of (steps alone undersell session-based work; cardio alone undersells everyday movement). The floor of steps and the spikes of cardio together tell the activity story.