See a perimeter's progression

You weigh the same as last month but the mirror says otherwise. The cut might be working in the right shape even if the scale is stuck. You ask the agent for the waist curve to see what the tape has been saying since you started.

What the agent needs to hear

Behind your sentence, the agent works backwards to two pieces: which site (waist, biceps, chest, thigh, hip, neck, calf, forearm) and the window (default: last 3 months). You can name one site, several, or all. Window overrides come in three flavors: named ranges (« since the cut started »), relative (« last 8 weeks »), or explicit dates.

One site, several, or all

Single-site reads are the cleanest, most actionable shape. « Show me waist » draws one curve, picks a scale that fits the range, and gives you something you can read at a glance. Right call when you have a specific question.

Multi-site reads overlay curves on the same card. « Waist and biceps over the cut » is the recomp question made visible — you want the waist coming down while the arms hold. The agent picks scales so sites with different magnitudes don’t end up looking identical.

All-sites reads return a grid, one small chart per tracked site. Good for the longitudinal picture, worse for decisions — too much surface to read at once. Use it for periodic check-ins, not for tuning the next block.

What the perimeter curve tells you vs the weight curve

A perimeter curve in isolation only tells you what the tape said. Paired with the weight curve, it tells you what your body is actually doing. Weight flat + waist down is recomp working — you’re trading fat for muscle at similar rates and the scale hides it. Weight down + waist down is a clean cut. Weight down + waist flat is the red flag — you’re losing scale weight but not the fat you wanted to lose, which means muscle is going with it. Weight up + waist up is a bulk producing fat instead of muscle.

The two signals together dismantle the most common mistake in body composition: trusting the scale alone. The scale moves for a dozen reasons; the tape at the waist moves for one or two.

When the agent gets it wrong

If the site shown isn’t the one you asked for, name it cleanly and the agent re-renders. If the curve looks too flat for what you’ve been feeling, the gaps may be doing the lying — ask « which weeks are missing » and you’ll see whether sparse logs are flattening a real trend.

The subtler failure is protocol drift, not data drift. If the curve picks up a sharp move that doesn’t match what you remember doing, ask yourself whether you changed where you anchored the tape, the time of day, or the tape tension. The agent logged honestly; the measurement conditions shifted. Naming the drift out loud (« I started measuring at the navel mid-cut, before that at the smallest point ») is the fix — the trend reads cleanly from that point forward.

What makes the trend worth trusting

Three things decide whether this perimeter curve tells you something useful: the measurement protocol stayed consistent across the window, the log cadence is regular enough to read (weekly is the sweet spot; monthly is too sparse for short windows), and the site you’re reading matches the question you’re asking (waist for cut and bulk, biceps for arm hypertrophy — reading the wrong site at the wrong moment is a category error). Paired with the weight curve, this read becomes the input for the next plan adjustment.