I'm an ER doctor. I built an AI coach because I was managing everyone's energy but my own.

By Jason Nau, MD

There's a particular kind of tired that doesn't show up on any of my devices.

My Oura ring can tell me my HRV dropped overnight. My calendar can tell me I worked three of the last four weekends. My to-do app can tell me I have forty-one open items. What none of them could tell me was the only thing I actually needed to know on a Tuesday morning before a shift. Given how depleted I really was, and given what I actually cared about, what should I do today, and what should I let go of?

That question sits in a gap. On one side are wearables, which are extraordinary at measurement and completely silent on meaning. On the other side is productivity software, which is good at managing time and blind to your body. I own instruments on both sides of that gap, and I lived in the canyon between them for years.

So I built a bridge. This is the story of why, what I learned, and what I think it says about the next decade of consumer health software.

The measurement trap

We are living through the golden age of personal measurement and the dark age of personal meaning.

I can tell you my resting heart rate to the beat, my deep sleep to the minute, my readiness score to two decimal places. That's miraculous. A decade ago this data lived only in sleep labs and ICUs. But almost nobody talks about the second-order effect: measurement without interpretation is just a new source of anxiety. I have watched the quantified version of myself check a number, feel briefly worse about it, and do nothing differently. The dashboard becomes a mirror you flinch at.

The wearable companies know this, which is why Oura Advisor and Whoop Coach exist. They're bolted-on chat layers trying to answer "so what?" The trouble is they answer it from inside a fitness frame. They can tell you to sleep more or train less. They cannot tell you that your recovery has been wrecked for three weeks because you've been avoiding a hard conversation with your business partner. They don't know you have a business partner, or that you value loyalty in a way that makes confrontation feel like betrayal, or that the last two times this pattern showed up it ended in a quiet, expensive collapse.

The data that would actually help you isn't biometric. It's biographical. And the companies sitting on the biometric data have no model of your biography. Worse, they have a business model that makes them the wrong people to hold it.

Why I didn't trust anyone else to hold the data

A physician notices something a Silicon Valley founder may be incentivized to turn a blind eye to. The more useful this product becomes, the more dangerous the data behind it gets.

For an app to coach me well, it has to know my values, my failures, the things I avoid, how my body responds to alcohol, the shape of my marriage, the parts of my work I find meaningless. That isn't a wellness profile. It's close to the most complete record of a person you could assemble outside of a confessional, because it pairs the recovery data with the inner life that explains it.

I would not give that to an ad-supported company. I would not give it to a startup whose exit depends on the data being an asset. And I increasingly would not give it to any business whose servers can be subpoenaed, breached, or quietly repurposed when the funding gets tight and the board gets nervous. Health-data regulation is tightening in real time, and the apps holding the most intimate data are the ones with the most to lose.

So I made a constraint that looks like a limitation and turns out to be the entire moat. The data never leaves the device. Your reflections, your health history, your story stay encrypted on your phone, with no server-side copy anywhere. When the AI coach needs to think, it works from an ephemeral, derived slice of context and then forgets. There is no Reserve database of human beings. There is nothing to breach, nothing to sell, nothing to subpoena, nothing for me to change my mind about later.

The part I didn't predict is what privacy does to quality. When you actually trust that no one is keeping a copy, you tell it the truth. You write the entry you would never put in a cloud journal. And the truer the input, the better the coaching. Trust isn't a compliance checkbox here. It's the input that makes everything else work. Privacy and quality turn out to be the same axis.

Energy is the constraint. Values set the direction.

The productivity industry sold us a lie of omission: that the scarce resource is time, and the fix is a better system for slicing it. Anyone who has stared at a perfectly organized task list at 9pm with nothing left in the tank knows that's false. You didn't run out of hours. You ran out of yourself.

Energy is the real constraint, and energy isn't willpower. It's a physiological state you can now measure. That's the synthesis the whole product is built on. Your wearable already knows your energy budget for the day. The only thing missing was a coach that spends it wisely.

So Reserve turns your recovery data into a simple daily energy state, and then the coaching adapts to it. This is the part that feels different the first time it happens. On a depleted day it asks less of you and means it. On a recovered day it aims higher. The interface itself softens when you're crashed. An app that respects your physiology instead of nagging against it is a strange and good experience, and I'm a little surprised I had to build it myself.

But energy management on its own is just sophisticated self-care. The second half is direction. A short conversation when you start surfaces what you actually value. It's a real conversation that listens for what drives you, not a personality quiz. From then on, every coaching note and every suggested task connects back to it, across all of life and not only the slice that fits in a work app. The thesis, in one line: energy, not time, is the constraint, and values, not urgency, set the direction.

The thing that surprised me most: it compounds

Most software is the same on day 100 as it was on day 1. You learn it, it plateaus, you eventually churn. Reserve does something I have never experienced in a consumer app. It gets measurably better the longer I use it, because it knows more.

After three months it had noticed things about me I hadn't consciously connected. That my task completion roughly doubles on mornings I exercise. That my HRV takes a specific, predictable hit for about 48 hours after I drink. That I rate my creative life area lowest right when I'm proudest of my work, because I've quietly stopped making anything that isn't for someone else. None of that came from a model trained on millions of strangers. It came from my data, analyzed on my device, surfaced honestly with the sample size shown so I could trust it.

That's the part I'd bet on, and the part I think the incumbents will struggle to copy. Not the prompt engineering, which is replicable, but the accumulated, private, irreplaceable context. Switching away wouldn't cost me a subscription. It would cost me my own story. That's a different kind of moat than the industry is used to, and you can only build it if you've earned enough trust for people to hand you the truth in the first place.

What I think this means

I'll resist the urge to over-claim. I'm one physician who codes, I built this for myself first, and its earliest users are a few dozen people across medicine, law, tech, and finance who live in the same canyon I did. I'm running a small clinical pilot because I would rather know whether it actually moves burnout scores than assume it does. The honest answer to "does this work for everyone" is that I don't know yet, and I'm trying to find out the rigorous way.

But I'm confident about the shape of the thing. The next decade of consumer health isn't more sensors. We have enough sensors. It's the interpretation layer, the software that connects what your body is doing to what your life is for. The companies with the data are structurally the wrong ones to build it, because the trust problem and the business-model problem are the same problem. The opening belongs to whoever is willing to hold the most intimate data the most lightly: on the device, ephemeral, never for sale.

That's the bet. Energy over time. Values over urgency. Privacy not as a policy but as the product. A coach that knows your body and your biography, and forgets it the moment it's done helping.

I built it because I needed it before a shift one morning and it didn't exist. It exists now. If you live in the same canyon, if you're already pushing hard and want to push smarter instead of just harder, I'd love for you to try it.