Your body is a time series: reading labs and wearables over time

My fasting glucose came back at 95 mg/dL last spring, and the lab flagged nothing. It sat five points under the 100 mg/dL line where “normal” ends and “prediabetic” begins. A clean result, and I almost closed the tab. Then I pulled the two draws behind it: 82 mg/dL three years ago, 88 mg/dL last year, 95 mg/dL now. Every one of those numbers passed on its own. The slope failed all of them: a climb of about 4 mg/dL a year, walking in one direction. The level was fine. The trend was the whole story, and no single frame could hold it.

A lab result is a photograph. One frame, one moment, stripped of everything that came before and after. The same is true of every wearable reading, each night’s sleep score and each day’s HRV. We built personal health around reading these photographs one at a time, a year apart for labs and a day apart for the wearables, and then acted surprised when they didn’t tell us what was happening. The information was never in any one frame. It was in the trend, the motion between them.

A number alone can’t have a direction

Tell me your fasting glucose is 95 mg/dL and I know almost nothing. It’s a dot on a chart with no chart. Is it the bottom of a recovery or the top of a slide? The reading can’t say, because a single point has no derivative. Direction is the one thing a photograph structurally cannot contain.

Add time and the dot becomes a line, and the line has a slope. 82, then 88, then 95 isn’t three normal results. It’s a trajectory with a heading and a speed, and you can do arithmetic on it: at roughly 4 mg/dL a year, that line crosses out of normal inside two years. You cannot make that forecast from the 95 alone, no matter how good your model is. The math isn’t there to do.

That’s the smallest thing time gives you. It turns a level into a slope. The bigger thing is what the slope lets you do across everything else you measure.

Time is the join key

Here’s the problem nobody names. Your blood draw, your ring, your CGM, and your watch don’t speak the same language. A blood draw is a single timestamp once a quarter. A continuous glucose monitor reads every five minutes. Sleep is staged in epochs across a night. Training load is a number per session. Four incommensurable streams, none of them on the same scale. You can’t average them. You can’t put them on one axis. There is no honest way to add ApoB to HRV.

Except one. They all happened somewhere in time. Every reading carries a timestamp, and the timestamp is the one column every stream has in common. Line the streams up on a clock and they become comparable, not because the numbers are alike, but because you can ask what else was true at the same moment. That December blood draw stops being a lonely number and becomes the draw that landed three weeks into a bulk, on a week your saturated fat was up 31%, after a month your deep sleep was down. The reading didn’t change. Its neighbors did.

If a body is a film, the time axis is the strip of sprockets running down the side, the perforations that hold four separate reels in sync so the picture and the sound land on the same frame. Pull the sprockets out and you have four piles of footage and no movie. Every cross-source read Depth does is built on those sprockets. Here are three.

Three trends across labs and wearables only the time axis can see

The lead and the lag. A behavior in December shows up in a panel in March, and the only way to connect them is the calendar between. My ApoB crept up 17 points across three panels (78, then 86, then 95), and on its own that’s just a marker drifting. But the climb starts on the panel after I began bulking in December, and it tracks the saturated fat that went up with it. Three winter months separate the cause from its first lab-visible effect. No single panel contains that link; it lives in the offset between two streams on the same timeline. The dose-response between saturated fat and ApoB is well established in the population, but “true in a crowd” only becomes “true for you” when your own timeline shows your rise following your own change. Swap two weekly red-meat meals for fish and the line bends back by the next draw. The change, the lagged effect, then confirmation on the next draw, and it needs the time axis at both ends.

The two streams that explain each other. Take a single bad night of deep sleep. On its own it’s noise. You slept badly, who knows why. Now overlay the CGM on the same clock. On the nights deep sleep craters, glucose is still elevated past midnight, and those are the nights you ate dinner after 9 p.m. Two systems most people would never connect, a metabolic stream and a sleep stream, and the only thing linking them is the axis you can read them against. The mechanism is plausible on its own: glucose your body is still clearing keeps it out of the deep stages it should be in. But the reason to act isn’t the mechanism, it’s that your own aligned data points the same way. Move dinner two hours earlier and watch both lines settle. Late meals hold your glucose up past midnight, and it’s stealing your deep sleep.

The right cause, and the wrong one dropped. Since March, HRV is up 18% and resting heart rate is down 6 bpm. Real gains. But you added both Zone 2 and HIIT in the spring, so which one earned them? A snapshot sees the result, not the history, so it can’t say. The timeline can. Tag every session and walk the curve: the improvement tracks Zone 2 volume almost session for session, while the HIIT weeks barely move the line. Zone 2 is carrying it. That’s a training decision made for you, and it’s only legible because months of two streams sat on one axis long enough to show which one the body was responding to.

The baseline is the quiet payoff

There’s one more thing the time axis builds, and it’s the one that matters most. Population reference ranges are a crowd’s average; your own normal is the only one that matters, and the only way to know it is to have watched you, over time, when things were good. A resting heart rate of 58 bpm is reassuring unless yours lives at 50 and has drifted up for six weeks. The same 58 is an alarm for one person and a non-event for another, and the difference is entirely in the history. Enough timestamped readings and the baseline stops being the textbook’s and starts being yours. After that, every new reading gets graded against the right curve.

This is the line between an app that lists your numbers and an agent that reasons about them. A list reads each frame alone and hands it back to you flat: here’s your glucose, here’s your HRV, good luck. Reasoning means saying this is because of that, and “because” is a claim about time. The cause came first, the effect followed, the offset is consistent, and changing the one moves the other. Your watch sees one half, your ring sees another, your blood is only a snapshot, and the timeline is what lets us read them as one picture instead of four.

So when your next result lands inside the range and the lab waves it through, don’t ask whether it passed. Pull the last three and ask which way the line is running, and what else on your timeline was moving with it. My 95 mg/dL was never the finding. The slope was, and the December that started it, and the dinner hour I can move tonight. We don’t hand you a frame and wish you luck. We watch the film, every reel on the same sprockets, and tell you which way it’s running.