I started writing this piece because I wanted to know whether I was, finally, allowed to feel tired. For roughly twenty-six months — I tracked them, because I track everything — I woke between 4:30 and 5:00 a.m. with the same particular feeling, a low electrical hum behind my sternum, and could not get back to sleep. I had tried, in order: melatonin, magnesium glycinate, glycine, L-theanine, three different weighted blankets, a sleep coach, a meditation app I came to actively resent, and a six-week elimination diet that left me thinner and angrier than I had been at the start.

None of it worked, and the failure of all of it was the most interesting thing about it. There is a kind of pattern that emerges when a problem refuses to yield to any reasonable intervention, and I had begun to suspect — somewhere around the third sleep coach — that I was solving for the wrong variable.

What finally moved the needle was not a product, not a protocol, and not, as my husband had been suggesting for some time, “just trying to relax.” It was a single sentence from a pulmonologist friend over dinner, delivered without any particular ceremony, that reframed two years of waking up at 4:47 a.m. as something other than a personal failing. I want to walk through what he said, what I did with it, and what the past nine months have looked like — because if even one of the things I tried might help you, I think the version of this you’re reading is worth the time.

What I had been telling myself

The story I had been telling myself was this: I was a light sleeper. I had always been a light sleeper. My mother was a light sleeper. It ran in the family the way blue eyes ran in the family, and the most I could hope for was to minimize the inputs — less caffeine, less screen time, less stress, less wine — and accept whatever remained.

This story had several things going for it. It absolved me of effort that wasn’t working anyway. It explained my mother. And it was the kind of intuitive folk diagnosis — you are a particular sort of person, this is what your sort of person does — that is satisfying in the way astrology is satisfying. The only problem was that it predicted nothing. It did not, for example, predict why I woke up at exactly 4:47 a.m. and not at 3:00 or 6:00. It did not predict why the wake-up was worse on Sundays. It did not predict why six weeks in Lisbon, with all of the same inputs, made it briefly disappear.

When I described all of this to Elias — a pulmonologist I’ve known since residency, who studies sleep apnea and whose patience for unfalsifiable theories is roughly zero — he listened for about three minutes and then put down his fork. “Maya,” he said, “you don’t have an insomnia problem. You have a 3 a.m. cortisol spike, and you’ve been training it for years.”

I have thought about that sentence almost every day since. Most of what follows is what I did with it, beginning with the thing I should have done two years earlier, which was stop treating the symptom and start mapping the upstream.

What the data actually showed

I want to be specific about the data, because vagueness here is a kind of dishonesty. For ninety consecutive nights I wore a chest-strap heart-rate variability monitor, kept a paper log of bedtime and wake events, and ran a galvanic skin-response patch on alternating weeks. None of these devices is exotic; all of them are available to anyone with a credit card and a long-suffering partner.

The pattern was unambiguous. On the nights I woke at 4:47, my heart rate had begun climbing — by an average of nine beats per minute — roughly forty minutes before I was conscious of being awake. My skin conductance, which is a reasonable proxy for sympathetic-nervous-system activity, had begun rising about an hour before that. The wake-up was not the event. The wake-up was the cleanup operation at the end of a much longer cascade that had been running in the dark for the better part of two hours.

This is, in retrospect, embarrassingly obvious. The body does not produce a sudden 4:47 alarm. It produces a long, slow autonomic ramp, and the conscious mind — my conscious mind — arrives at the end of it, blinks twice, and decides this must be insomnia. I had spent two years debugging the wake-up. I should have been debugging the ramp.

The three interventions that moved the curve

Most of what I tried in the months that followed did nothing. I am going to skip past the failures, with one note: I think the field is full of products that work on people whose problem is not what mine was, which makes the testimonials genuine and the recommendations useless. The three interventions that moved my ramp curve, all confirmed with the same chest strap, are below.

One: a hard cutoff on protein after 7 p.m. Not a reduction. A cutoff. A late-evening protein load, even a modest one, appears to drive a small but consistent nocturnal cortisol release in some people, and I am apparently one of them. Removing it cut my average overnight heart-rate ramp by about a third within two weeks.

Two: a 90-minute screen-free window before bed, which I had been claiming to observe for years and which, when I actually measured it, turned out to be closer to twelve minutes. The honest observation made far more difference than the imagined one.

Three: the monitor itself. This is the part that surprised me most. Having a real-time read on my own nervous system — not in the morning, when the data is retrospective and toothless, but at 10:30 p.m., when I could still do something about it — turned out to be the most behaviorally powerful thing I tried. I wrote about the specific device I ended up using and why I kept coming back to it over the others I tested.

Together, these three things cut my 4:47 wake-ups from roughly six per week to roughly one. I am not going to claim they will work for everyone. I will claim that the framework — debug the ramp, not the wake-up — is more useful than any single intervention I’ve come across, and that the equipment to actually see the ramp has become, in the past two years, remarkably accessible.

What I’d tell the version of me at month one

If you are nine months behind me in this, here is the compressed version. You are probably not a light sleeper. You probably have a slow autonomic ramp that has been training itself, every night, for somewhere between six months and ten years, and the most useful thing you can buy is not a supplement. It is a way to see the ramp in real time, while there is still time to interrupt it.

I do not, generally, recommend products in these pieces. I am recommending this one because, having spent two years on this and roughly $4,000 of my own money on things that did not work, I would have paid the price of the monitor on day one if anyone had told me what I am telling you now.