For a decade I built models that turned market chaos into a single call — buy, hold, or sell. The desk didn't reward the prettiest dashboard. It rewarded the person who could look at a wall of conflicting signals and commit to one move before the window closed.

Then I went home and did the opposite. I owned a ring, a watch, a scale that graphed my body fat to one decimal place. I had more data about myself than any athlete in history — and most mornings I had no idea what to actually do with it. Train or rest? Push or hold? The numbers all pointed in slightly different directions, so I did nothing, or I did whatever I felt like, which is the same thing dressed up.

That gap has a name now, at least in my head: the firehose problem. We've never had more information about our health, and we've never been more unsure what to do with it. And I've come to believe that gap — between everything we measure and the choices we actually make — is the most expensive problem in modern wellness.

Measurement is not the same as a decision

The wearables industry sold us a quiet lie: that if you could just see the number, the right behavior would follow. It doesn't. Visibility and action are different skills, and we've spent a decade over-investing in the first while starving the second.

Think about what your ring actually gives you on a bad night: a readiness score, an HRV trend, a sleep-stage breakdown, a temperature deviation. Four signals, and they don't agree. Your HRV is down but your sleep looks fine. Your temperature is up but you feel great. So which one wins? The app won't tell you — it just renders all four beautifully and leaves the hardest part, the decision, entirely to you.

And the cost of the unmade decision is never zero. Every morning you don't decide is a morning decided by default — by momentum, by mood, by whichever app shouted loudest. A dashboard is a decision you've postponed; twelve charts are twelve postponements. On a trading floor, deferral at least shows up on a P&L by Friday. In a body, the invoice arrives years later, itemized as nothing in particular.

On a trading desk, a model that output thirteen charts and no position would be laughed off the floor. Its entire job is to collapse uncertainty into an action you can be held accountable for. Health tech does the reverse. It expands one honest question — what should I do today? — into a museum of graphs, and calls that empowerment.

Where the years quietly go

Put a rough price on it. You'll make something like a thousand train-or-rest calls in a decade, ten thousand bedtimes, thirty thousand meals. Each one is small; the volume is enormous. Nudge even five percent of them from “default” to “deliberate” and you've made a material trade on the largest position you hold. That's the whole pitch — not perfection, just a slightly better hit rate, compounding quietly for years.

Here's the shape of it in one ordinary week. Monday you wake under-recovered but train hard anyway, because the plan said so and nothing said otherwise. Tuesday you're wrecked, skip entirely, and eat like it. Wednesday through Friday you're negotiating with yourself instead of training. One deliberate call on Monday — go easy today so Tuesday exists — and the week is a different week. Multiply by fifty-two, then by a decade. Nothing about that requires heroics. It requires a decision, delivered at the moment it's useful, in a form you can act on before coffee.

The research keeps circling the same point from different angles: long-term outcomes track daily behaviors, daily behaviors track decisions, and decisions fail at the moment of translation — the moment the number was supposed to become a move. We built an industry that measures everything and asks nothing of us. The asking is the product I actually wanted.

The three moves that actually matter

When I finally pointed my old risk framework at my own body, the method got embarrassingly simple. Three steps. I call it Decision for Performance, and it's the same loop whether you're managing a trading book or a training week.

01 — Measure less, on purpose

Not everything that can be tracked is worth acting on. Most days, six signals decide the call: sleep debt, HRV, resting heart rate, training load, glucose variability, and subjective mood. Everything else is noise that makes the noise louder. Muting a metric is not ignorance — it's a decision about where your attention is worth spending.

02 — Model honestly

A signal only matters relative to your baseline, not a population average printed in a glossy app. The same engines that price risk under uncertainty — Bayesian updating, simple regime detection — work beautifully on a person. They don't need to be fancy. They need to be honest about what they don't know, and to widen their confidence bands when the data gets thin. A model that can't say “I'm not sure” will eventually lie to you with a straight face.

03 — Decide, with a number attached

The output is never a dashboard. It's one line: given your last two weeks, do this today — with a confidence score attached, so you know when to push and when to back off. Deload. Easy Zone 2, forty minutes. Confidence 0.82. That's a decision you can be accountable to tomorrow. Thirteen charts are not.

What this asks of the builder — and of you

A one-line call is a serious thing to hand somebody at six in the morning, so the line has to earn its authority. It must show its work. It must admit its error bars. And it must never pretend to be your doctor — my rules are public and non-negotiable: the system never diagnoses, never replaces your clinician, never sells you a supplement. A model that wants the right to tell you what to do earns it with humility about what it doesn't know.

It asks something of you, too: a willingness to act at less-than-certain, to log what happened, and to let the model be wrong out loud so it can get better in public. Trust, like everything else in this method, compounds.

From measuring yourself to deciding for yourself

The shift I want for people is small to describe and enormous to live: stop measuring yourself and start deciding for yourself. The data was always supposed to be a means to that end. Somewhere along the way we made the measuring itself the hobby, and quietly forgot it was supposed to change what we do.

You don't need another chart. You need the next decision. That's the whole thesis, and everything I build — the open-source engines, the weekly notes, even the little flower project that isn't for sale — is just a different way of closing the same gap: one good decision at a time. The firehose isn't going away; you'll own more sensors next year than you do now. The goal was never more water. It's a cup you can actually drink.

The most expensive lesson of my trading career didn't come from a bad model. It came from a good one I didn't act on. Early on, a position of mine went wrong slowly — the signals were there, my own system was flagging them — and I froze. I kept waiting to be sure. By the time I was sure, the window had closed, and the loss was no longer a possibility I was pricing but a number on a page with my name next to it.

What stayed with me wasn't the number. It was the diagnosis. I hadn't lacked information; I'd lacked a working relationship with uncertainty. I had treated not yet certain as a reason to wait, when on a desk — and, I'd learn later, in a body — “not yet certain” is the permanent weather. The lesson that finally sank in wasn't be more careful. It was: learn to move while unsure, because that is the only kind of moving there is.

The myth of the fully informed decision

A trading floor teaches this fast, because it charges tuition daily. Complete information never arrives. You decide with error bars, or you don't decide at all — and not deciding is itself a position, usually the worst one, because nobody sized it on purpose.

Good deciders, I promise you, are not oracles. They are simply people who have made peace with acting at seventy percent sure — because they understand that waiting for ninety is also a bet, and it's paid for with the one asset you can never buy back: time. Hesitation always has a price tag. It's just printed in invisible ink.

The desk even has a word for how confidence should be expressed: sizing. You don't show conviction by being loud; you show it by how much you commit. Very sure? Act bigger. Seventy percent? Act — smaller. Nothing about that is reckless. Recklessness is the other thing: betting everything on the feeling of certainty, which is a feeling, not a fact.

Your Tuesday is a portfolio

Now look at an ordinary week the way an allocator would. Train hard today, or take the rest day? Push through the afternoon slump, or protect the evening? Accept the third commitment this week, or hold the line? None of these arrive with proof attached. Every one is a bet with a payoff curve, made under exactly the conditions the desk trained me for: partial information, real stakes, a closing window.

Here's what makes health bets kinder than market bets, though — the asymmetry runs in your favor. Most daily health decisions have small, recoverable downsides and quietly compounding upsides. The workout you didn't need costs you a little fatigue; the one you did need pays you for months. You don't have to be right every day. You have to be right slightly more often than not, for years. That is a low bar wearing the costume of an impossible one.

The uncertainty toolkit

Three moves, carried straight from the desk into daily life. They are boring, which is why they work.

01 — Prefer reversible mistakes

Skipping one workout is a position you can unwind tomorrow; quitting for a month is not. When you're torn between two options, take the one you can walk back. Reversibility is how you buy information cheaply: act small, watch what happens, adjust. The irreversible version of the same choice should demand ten times the evidence.

02 — Size the bet

Run a two-week experiment, not a lifestyle overhaul. Change one variable at a time, so the result means something. Your attention is capital — the mistake isn't investing it, it's deploying all of it into one trade because a podcast was persuasive on a Tuesday.

03 — Log the outcome

One line, either way: what I chose, what happened. Not for guilt — for priors. Future-you inherits that ledger, and after a year it becomes the one thing no app can sell you: actual evidence about your own body, under your own life. Decisions get easier at exactly the rate the ledger gets longer.

A worked example

Here's what this sounds like inside my head on an ordinary morning, so it stops being abstract. Tuesday, 6:40 a.m. HRV is down nine percent; sleep was fine; legs feel heavy but mood is good; there's a hard session on the plan and a long day behind it. Certainty on offer: none. Signals in hand: four, mildly contradictory — the firehose in miniature.

The old me would have googled her way into paralysis, or trained hard out of stubbornness — both defaults, neither decisions. The current version prices it like a position: the downside of going easy today is trivial and reversible; the downside of digging a hole before a heavy week is not. Confidence in “full send” sits somewhere around fifty. Confidence in “medium session, light top sets” sits above eighty. So that's the trade. Logged in one line, reviewed in two weeks, no drama at the rack.

Notice what never happened: I didn't resolve the contradiction in the data. I didn't wait for the signals to agree — they weren't going to. I sized a reversible bet in the direction the evidence leaned, and moved. That's the entire skill. It takes about ninety seconds, and it has beaten an hour of research every single time I've run the comparison.

What the mountain adds

Snowboarding compresses this whole essay into a single motion. The work happens before the work: scouting the line, reading the snow, knowing the exits. That part looks like caution, and it is — it's also the only part you fully control. Then comes the drop-in, and the drop-in is binary. You commit or you don't. The one thing you cannot do is half-commit; hesitation midway down a line is the only choice that guarantees a fall.

The mountain also teaches a distinction most dashboards blur: risk is not the same as consequence. A line can be low-risk and high-consequence — unlikely to go wrong, expensive if it does — and that combination doesn't mean don't. It means rehearse, until the danger has been converted into procedure. To a certain kind of colleague, my weekends look like the opposite of risk management. They are risk management, practiced at speed. I model risk for a living — and I ride it on weekends.

The barn taught the quietest version of the same lesson: a horse will not accept a rider who is pretending to be calm. The commitment has to be real, because your body broadcasts the truth of it. Decisiveness, it turns out, is a physiological state before it is a mental one.

Stop waiting

Certainty is not coming. Not on the desk, not on the mountain, not at 6:40 on a Tuesday morning with a training decision to make. I mean that as good news — you've been released from an impossible standard. Weigh the signal you have. Size the bet. Commit. Write down what happened.

The window is open now. It won't be later. That isn't pressure. It's permission.

You insure your car. You insure your house. You probably insure your phone. At the bank, I modeled risk for a living — and one day it landed on me that my own health was the one portfolio with no model on it anywhere. The riskiest asset I will ever hold, the one every other asset depends on, and not a single line of code watching it.

That sentence is the reason everything I build exists. It also names the villain of this whole project, and the villain is not “sick care,” not big food, not your ring. It's unmeasured risk — exposure compounding quietly in the one place nobody thought to point a model.

What a model actually does

Forget the math for a moment, because the math was never the point. A model doesn't predict the future — it prices exposure. It takes consequences that compound too slowly to feel and makes them legible before they become emergencies. That's the entire job: a pair of glasses for the invisible.

The kindest analogy I know is the weather forecast. A forecast doesn't tell you what the sky will do; it tells you whether to carry an umbrella. A good health model works the same way. It will never say you will get X — it can't know that, and anything that claims to is selling something. What it can say is: this exposure has been accruing for six weeks, and the price of ignoring it just went up.

The honesty requirements carry over, too. A forecast that never admitted doubt would be worse than no forecast, because you'd stop carrying umbrellas the first time it lied to you. Same rule here: the model states its confidence, widens its bands when your data runs thin, and says “not enough signal” out loud when that is the truth. In this category, epistemic humility isn't a nice-to-have. It's the product.

Health Portfolio Theory

Point portfolio thinking at a life and the mapping is almost embarrassing in how well it holds. Sleep, training, nutrition, and the people you love are asset classes. Each pays a different kind of return on a different schedule. Each can be over- or under-weighted. And your energy — not your money — is the capital being allocated, every day, whether you're paying attention or not.

The classic blow-up on a trading desk is concentration risk: one position quietly becoming the whole book while everyone admires its returns. The classic blow-up in a life is the same trade, and the position is usually work. Perfectly hedged everywhere else, and all-in on a single employer of your hours, your sleep, and your attention. No allocator would sign off on that book. We live it and call it ambition.

The rebalancing rules read exactly like an allocator's. No single position gets to sink the book. Drawdowns will come — an illness, an injury, a brutal quarter — so the question is never how do I avoid every drawdown? It's how deep can one go before it forces me to sell something I can't buy back? Sleep sold in your thirties, cartilage sold in a careless training year, friendships sold to a project — those are the trades that don't reverse.

Run the book for one honest week and the rebalancing writes itself. Sleep got overdrawn Tuesday and Thursday. The training position is over-weighted, because it's the fun one. The friendship allocation hasn't been topped up in a month, and the deficit is starting to show up in mood — which is to say, in everything. None of that requires an intervention. It requires what any allocator does on a Friday afternoon: notice the drift, trim the overweight, feed the neglected position, and let compounding do the rest.

The Body Balance Sheet

The second lens is simpler and, for most people, sharper. One side holds the assets you're building: muscle, an aerobic base, a sleep rhythm, friendships that hold under load. The other side holds the liabilities you're carrying: sleep debt, the nagging injury you've agreed not to think about, chronic stress with a badge that says temporary.

Both sides compound. Neither sends a monthly statement. The interest simply accrues, silently, in whichever direction you happen to be feeding. Muscle, in particular, behaves like a pension: boring to fund, compounding for decades, and painfully expensive to buy back late.

So I run a quarterly audit — twenty minutes, one coffee, four questions. What am I building? What am I carrying? What did I ignore this quarter? And the one that keeps it honest: if a client brought me this book, what would I tell them? The point isn't judgment. It's marking the book to market while corrections are still cheap.

The thousand-pound biosensor

I learned all of this at a barn before I could name any of it. Years before I owned a wearable, a mare taught me that my internal state was public information. Arrive at the mounting block with a racing mind and she dances sideways, ears flat, done with me before I'm in the saddle. Arrive settled and she settles. No dashboard, no login, no Bluetooth — she reads breath, grip, seat, and she is never wrong in the direction that flatters you.

That isn't mysticism; it's applied ethology. Reading a non-verbal partner is a rigorous, learnable skill, and it rewired what I expect from instrumentation. The signals about your state were always broadcasting. The horse was simply the first instrument in my life honest enough to display them without decoration. Every model I've built since has been trying to live up to her standard.

The never list

This category has thoroughly earned your skepticism, so my rules aren't fine print — they're load-bearing infrastructure, published where everyone can see them:

What the math is actually for: handing you — and your doctor — better questions, earlier. A model's success isn't being right; it's making you harder to surprise. Surprise is the real enemy in both trades — the blow-up nobody priced, the diagnosis that “came out of nowhere” after accruing quietly for a decade. You can't model your way out of mortality. You can absolutely model your way out of being ambushed by things that were visible for years.

You would never run a trading book without a risk report. You run your body every single day. The portfolio is already live — the only question is whether anything is watching it. It should be something kind, something honest, something that shows its work.

We named it sleep debt and then decided to treat it like a rounding error. It isn't one. It's a real loan with a real interest schedule, and the body keeps immaculate books — it just doesn't mail the statement until the balance is embarrassing.

My own ledger is blunt about it. After a run of short nights, the bar moves slower and my patience runs out faster — and both happen well before anything in me consciously reports “tired.” That's the first thing to understand about this particular debt: the borrower is always the last to know.

The interest schedule

A single short night is a payable — annoying, recoverable, settled with an early bedtime and an honest look at what caused it. If that were the whole story, sleep debt wouldn't deserve an essay.

It's the chronic shortfall that compounds, and the research keeps sketching the same schedule from different angles: appetite tilts toward whatever's fastest, cravings get louder, mood thins out — and, the genuinely expensive part, decision quality degrades while confidence stays exactly where it was. You decide worse and you don't notice. Ask any risk manager what the most dangerous combination in the world is, and it's that one: rising error rate, steady self-assessment. The numbers vary by person and the literature argues about size, as literatures do — but the direction is one of the most replicated findings in the field, and direction is all a decision needs.

Worse still, this debt finances itself. Tired-you is precisely the negotiator most likely to approve another late night, another “just one more episode,” another morning alarm pushed back instead of a bedtime pulled forward. The loan approves its own extensions. It is compounding in its purest and least flattering form.

Pricing a fortnight

Let's price a real one, because “sleep more” is advice and a number is a decision. Say your body wants seven and a half hours and the last two weeks averaged six and a quarter. That's a shortfall of roughly seventeen hours — more than two full nights of sleep, carried as an open position into everything you do. You wouldn't run any other deficit that size without writing it down somewhere.

Here's how that position expresses itself, in my ledger and probably in yours: the third set feels like the fifth. The 4 p.m. decision — train, skip, snack, scroll — starts going to the house. Food choices get louder and worse. And the whole time, self-assessment holds steady at “I'm fine,” because the instrument doing the assessing is the one that's under-slept.

Repayment is the un-glamorous part. You can't binge-repay seventeen hours in one heroic Saturday lie-in, and trying mostly dents the rhythm that protects you. You amortize: forty-five minutes earlier for a week or so, two anchor nights protected without exception, morning light, training load trimmed while the balance comes down. Boring, effective — and the balance actually clears.

Sleep is position maintenance

The reframe that finally worked for me came, unsurprisingly, from the desk: sleep is not the absence of life. It's position maintenance — the window where the body marks itself to market. Overnight, the training you did gets consolidated, the things you learned get filed, the hormonal accounts get settled. None of that is optional bookkeeping. It is the mechanism.

This is the part the gym taught me to respect. The workout is only the stimulus; the adaptation — the actual getting stronger — happens later, mostly at night. Do the reps and skip the sleep, and the gains are simply left at the window, unbanked. The same is true of skill, of memory, of mood. Effort unslept is effort partially refunded, and nobody tells you the refund went through.

Traders maintain their positions or watch them decay; there is no neutral. Skipping maintenance isn't a pause — it's a slow un-banking of work you already paid for. Once I saw bedtime that way, it stopped feeling like a curfew and started feeling like the close of business: the hour the day's work actually clears.

It also reframed the trade society keeps offering: an hour of evening awake for an hour of sleep, marketed as a free lunch. It isn't free. It's a loan against tomorrow's judgment, tomorrow's lifts, and tomorrow's patience with the people you love — the three most expensive things you own.

How I program rest

So rest sits inside my programming with the same status as load. Not the leftover at the end of the day — a line item. Three mechanics do most of the work:

01 — Bedtime is a margin call

My alarm is set for going to bed, not just for waking up. Most nights it's non-negotiable — not a mood, not a debate. The rare deliberate late night is a position taken knowingly: priced in advance, paid back on schedule. What it is never allowed to become is a habit with a good story.

02 — Rest is a training variable

Rest days and deload weeks are written into the plan before they're needed, with the same seriousness as sets and load. If you only rest when you're forced to, you were never resting — you were breaking, slowly, with paperwork.

03 — Joy counts as recovery

This one I borrowed from dance, the corner of my life where my spreadsheets have no jurisdiction. A proper rest day is not recovery-guilt in comfortable clothes. It's joy, on purpose — music, movement, people — because joy is what restocks the appetite for effort. Call it an off-season for the soul, scheduled like everything else that matters. It sounds soft until you watch it work: the weeks with real joy in them are the weeks the discipline costs less. That isn't a coincidence. It's the funding mechanism.

The fine print

One honest boundary before you go, because this essay is about the debt healthy sleepers take on voluntarily. If your sleep stays broken despite the basics — a steady schedule, light in the morning, load managed — that is not a discipline problem, and it's not a dashboard problem either. It's a clinician conversation, and no app, including anything I build, should ever talk you out of having it.

Everything else in this method — the training call, the mood, the day's decisions — sits downstream of the night before. Fund the position first. The interest you don't pay is the cheapest performance gain you will ever buy.

I train six mornings a week, and the polite assumption people make is that I must feel like it. So let me be honest about the ledger: motivation shows up to maybe two of those six sessions. The other four belong entirely to a system — and the system, not the feeling, is what people are actually looking at when they call someone “disciplined.”

I'd like to retire discipline-as-personality altogether. It flatters the wrong thing. Discipline as a character trait is something you either have or you don't; discipline as architecture is something anyone can pour, one boring decision at a time. This essay is about the pouring.

Weather and climate

Motivation is weather. It's real, it's glorious when it rolls in, and it is completely unreliable. It arrives after good news and vanishes after bad sleep; it loves January and abandons February. Nothing important should be scheduled on it.

Discipline is climate — the infrastructure you build so the thing happens in all conditions. You don't negotiate with rain; you own a jacket. Planning your training around motivation is forecasting your own mood three weeks out, which is astrology with extra steps. A plan that requires enthusiasm is a plan with a single point of failure.

I learned to distrust motivation in the nicest possible way: by watching it perform. On a shoot, everyone is motivated at 8 a.m.; the set runs on call sheets anyway, because professionals don't leave the outcome to the morning's mood. The call sheet is the respect. The mood is the bonus.

Decide once

The architecture is boring on purpose — boring is the feature. The whole trick is to decide once, in a calm moment, so you never have to re-decide daily in a weak one: training at the same hour every day, kit by the door the night before, the session written down before bed, a default breakfast that requires no thought. Each pre-made decision drops the daily cost of showing up a little closer to zero.

Willpower is a budget, and most people spend theirs at the wrong counter — on the deciding whether, when it should be reserved for the doing. Friction design finishes the job: make the good thing the lazy thing. A gym bag sitting by the door beats raw resolve every single time the experiment is run. I've stopped being embarrassed by how mechanical this is. Mechanical is what reliable looks like from the inside.

Environment finishes what scheduling starts. The phone charges outside the bedroom, so the morning's first negotiation never opens. The calendar block says training the way it says any meeting that matters — because a standing appointment with your own health is a meeting that matters. Design the room, and the room votes with you.

It's the same loop I run on my models: small improvements, applied daily, compounding quietly. Model it, train it, repeat. You don't need a new self. You need a lower activation energy.

What modeling taught me about “effortless”

My other career taught me the same lesson from a stranger angle. Every photograph that looks effortless is sitting on top of scaffolding you will never see: hours of lighting, fittings, a full crew, forty frames for the single one that ships. The industry's entire product is the appearance of ease, and the appearance of ease is the most heavily engineered thing on the set.

Effortless is a finished system. Presentation is construction — I've said that on stages, and the gym is where I live it. What reads as drive from the outside is, up close, mostly design: the fixed hour, the pre-packed bag, the program that removes every morning's philosophy seminar about whether today counts. Even the desk agrees. The calm trader in a storm isn't calmer by temperament; she's the one holding a checklist she wrote on a quiet afternoon.

A week inside the system

To make it concrete: six sessions, Monday to Saturday, same hour. Two of them, most weeks, ride actual enthusiasm — those are lovely, and I treat them as weather: enjoyed, never budgeted. The other four run on rails. I'm often not “feeling it” at the first set; by the third, the feeling has usually arrived late, the way it tends to. The session gets logged either way. The log doesn't have a column for enthusiasm.

Two objections always come up, so let's take them head-on. Isn't this rigid? No — the flexibility is scheduled too: a floating swap day for when life happens, a deload week already on the calendar, a standing rule that illness outranks the plan without a committee meeting. Rigid systems break. Mine bends in designated places, which is exactly why it doesn't. What about spontaneity? Spontaneity is what the system buys. Because the mornings run themselves, the evenings are genuinely free — for dance, for the barn, for saying yes to things. Order in one corner of a life funds freedom in all the others. That's the trade, and it's the best one I've ever made.

The kindness clause

Now the part I refuse to leave out, because “discipline” has been marketed as self-punishment for so long that the word arrives pre-bruised. Discipline is not punishment. It is self-respect with a schedule — the practical form of believing your future self deserves to inherit good decisions.

You will miss days. I do. The system's answer to a missed day is administrative, not moral: next session, same hour, no interest charged. Shame is a terrible creditor — it charges compound interest on debts you already paid. Run the system, not the spiral.

And the schedule includes rest, on purpose, in writing. A system that can't breathe isn't discipline; it's just a slower way to quit. The strongest climates have seasons.

One more quiet benefit nobody advertises: a system is portable in a way a mood never is. Travel week, work crisis, February — the same fixed hour and the same packed kit carry across all of them, scaled down when they must be. Ten disciplined minutes on the worst day preserves the thing that matters most, which is the identity of someone who showed up.

Build climate

Stop waiting for weather. The smallest version of everything above begins tonight and takes ninety seconds: put tomorrow's kit by the door. That isn't a metaphor for the method — that is the method, at one-to-one scale. One decision, made early, so the morning doesn't have to be won.

This one is not for sale. It never will be — that's a feature, not a missing business plan. My partner and I built it on weekends, for free, for no reason except that making it felt like meditation. And after essays full of portfolios and margin calls, it might be the most important thing on this site.

It works like this: at the end of the day, instead of writing in a journal, I draw flowers. However they come out — rushed, careful, crooked, blooming — whatever arrives on the page is the entry. Some nights it's a garden. Some nights it's three tired tulips. That's allowed. That's the point.

How it works

The app looks at the drawing and makes a gentle guess at the mood I was probably in when I drew it. What it reads is deliberately simple — how hard I pressed, how much of the page I used, whether the petals open or close, how many blooms showed up tonight — and it guesses gently, in full sentences, with room for doubt. It is allowed to say “I'm not sure,” which already makes it more honest than half the dashboards I've met professionally.

There is no score. No streak to protect. Nothing announcing that I am twelve percent below optimal. Over the weeks, the little bouquets accumulate into something I have never once gotten from a chart: a picture of how I've actually been. Not how I performed — how I was. Here's your month, in flowers.

Sunday evenings we review the week together, tea in hand — my partner's garden next to mine, two small exhibitions of how it actually went. Some weeks the gardens agree. The interesting conversations start when they don't.

Why the quant loves it

You'd think the person who writes risk models would find this frivolous. The opposite. It is the most honest sensor I own, and I can tell you exactly why. Some weeks, every metric I track says fine — sleep banked, training logged, portfolio balanced — and the flowers droop anyway. Weeks like that, the flowers are usually right, or at least right enough to be worth a look. Something was leaking that the instruments weren't licensed to measure.

There's a law in economics that explains the rest: when a measure becomes a target, it stops being a good measure. Every score invites gaming; every streak eventually gets protected at the expense of the truth it was supposed to track. The flowers cannot be gamed, because there is nothing to win. No target, no optimization, no leaderboard of one. That’s precisely why they stay true.

There's also a professional confession buried in here: the hardest signals to model at the bank were never the prices — they were the humans reacting to them. Sentiment, fatigue, fear. Whole trading floors run on unmeasured mood. So “feelings are data” isn't a soft claim from a hard scientist; it's the most quantitative thing I believe. The flowers are just the first interface that let those signals report themselves without being interrogated.

Why we built it small

Every design choice is a value wearing an engineering costume. There are no accounts, because a mood diary belongs to a person, not a growth funnel. The drawings never leave the device, because trust is easier to keep than to win back — and this is the one dataset of mine I would never let near a cloud. The model itself is tiny, on purpose: sophisticated enough to notice pressure and openness and pace, humble enough to be wrong out loud. It doesn't advise, doesn't optimize, doesn't nudge. It looks carefully, guesses gently, and leaves the meaning-making to me — which is where the meaning was always going to be made anyway.

The origin story explains the constraints. We built it after a stretch when every number I tracked was perfect and I was, by any honest account, miserable — over-modeled and under-noticed, a portfolio outperforming its owner. My partner suggested, half-joking, that I stop journaling sentences I didn't believe and draw what the day felt like instead. The first week of wobbly tulips told me more than a quarter's worth of dashboards had. Then, because we are who we are, we taught a very small model to read them. The joke became the most truthful instrument in the house.

Numbers decide; flowers notice

I think of this project as the soul check on everything else I build. The Decision Engine turns numbers into one clear call; the flowers make sure the person receiving the call is still being listened to. One system is for precision. The other is for affection. You need both in the kit, and most of this industry ships only the first.

Because optimization without noticing doesn't stay a method — it curdles into one more firehose, pointed inward this time. People ask me after talks, “doesn't all this measuring make you anxious?” It can. That's real, and anyone selling you a fully quantified life without mentioning it is skipping a risk disclosure. The counterweight can't be bolted on afterward; it has to be built into the method. Mine happens to be drawn in petals.

My most rigorous system and my least rigorous app are the same belief at two volumes: technology should help you notice your life, not grade it.

Draw me a flower

So, an invitation instead of a conclusion. If we ever meet — after a talk, at the gym, at the barn — draw me a flower. Any flower. Thirty seconds, no artistic standards enforced. I'll tell you what I see in it, and I promise to be kind. That's the whole app. And honestly, most of the philosophy: the engines decide, the flowers notice — keep both, in that order, and the data stays in service of the life. Which was the only point of measuring anything at all.