I Failed My Own Protocol 23 Times. My Diary Says I Was Reading the Wrong Number.

Dr. Barry Dublin, MD · Day 49 of 180

Plates going around. Somebody's fork hovering near my elbow. Two separate people asking me, in the kind way families ask, whether I was sure I didn't want any.

I said no twice. And here is the part I still can't quite get over: it cost me nothing.

I wasn't gritting my teeth. I wasn't doing the arithmetic of how many more minutes until we could reasonably leave. I wasn't performing a man who wasn't hungry. I was just a person at a table, in a conversation, not eating — the way you'd sit at a table and not eat the flowers.

And it hasn't only been last night. For two weeks now I've been going out to restaurants with my family, sitting through entire meals watching other people eat, and feeling essentially nothing about it.

You need to understand why that's strange.

Restaurants are what break me. Not "the holidays" as some vague category — specifically, sitting in a restaurant with people I love while food arrives at the table. That exact room has ended more of my attempts than anything else in my life. If you'd asked me a year ago to name the place where my discipline goes to die, I wouldn't have had to think about it.

I have twenty-three failed attempts on record that say last night should have been hard.

It wasn't. And I don't think that's because I finally became a stronger person at sixty. I think it's because I spent this summer reading four years of my own handwriting and discovered I'd been watching the wrong number the entire time.

Twenty-three times

Let me be exact about the failure, because the number is the whole point.

Since August 2022 I have started therapeutic ketosis — a precise, measured metabolic state, not a diet in the way most people use the word — twenty-three separate times.

My best run before this one lasted forty-five days. Most were far shorter. Several didn't survive the week.

And every single one of them ended the same way. Not a graceful tapering off. Not "life got busy." I was tempted, and I ate.

Twenty-three starts. Twenty-three breaks.

I'd like to sit in that a moment instead of hurrying past it, because the humiliation is part of the data.

I'm a physician. Thirty years in metabolic medicine. I have sat across a desk from a patient and explained this — calmly, confidently, with the particular authority of a man who knows exactly what he's talking about — and then driven home and failed at it myself. Again. For the twenty-second time.

There's a specific flavor of shame in being unable to do the thing you're qualified to teach. It isn't dramatic. It's quiet, and it accumulates.

For most of those four years I explained it to myself the way nearly everyone does. I'd been weak. I'd gotten complacent. I needed to want it more.

That explanation has one enormous advantage: everyone accepts it. Your doctor accepts it. Your family accepts it, sympathetically, while privately agreeing. You accept it, because it feels like maturity — like taking responsibility rather than making excuses.

It has one enormous disadvantage. It doesn't fit the evidence.

The one unusual thing I did

Here's the only thing that separates my twenty-three failures from anyone else's: I wrote them down.

Every attempt. Every reading. A drop of blood on a test strip, a number, a date, a time, and a line about what happened that day. Four years of it.

Four hundred and ninety blood ketone readings. Three hundred days. Twenty-three restarts.

I didn't do it to build a dataset. I did it because I'm a doctor and measuring is a reflex, the way some people photograph everything. But at some point the notebook stopped being a habit and became evidence.

That's the real difference, and it's worth dwelling on.

When you fail at something like this, you almost never have a record. You have a feeling. The feeling is I blew it again, and there's nothing to check it against — no file, no numbers, nothing that could possibly contradict it. So the story wins by default. And the story is always, always the same: something is wrong with you.

I had four years of data sitting in a drawer.

And I still believed the story. I just believed it while sitting directly on top of the file that disproves it, which is its own kind of joke.

The theory I was sure about

For years I had a working theory, and I was confident enough to teach it.

The theory: the number protects you.

Beta-hydroxybutyrate — BHB, the ketone I measure — is what your brain runs on when it isn't running on sugar. My clinical experience said that somewhere above 4.0 mmol/L, food loses its grip. Not that you stop enjoying food. That it stops pulling. The cake is still cake. It simply isn't addressed to you.

So the conclusion wrote itself. I break when my ketones are low. Low number plus nearby food equals disaster. High number equals safe.

Clean. Intuitive. Almost certainly wrong.

I found out this summer, when I finally sat down and read the diary properly, and ran into two entries that took the whole thing apart.

May 19, 2025. 9:47 in the morning. BHB 4.8.

That's a good number. That is comfortably inside the range where, according to my own theory, food should have had no power over me whatsoever.

I broke that day.

March 26, 2026. 12:59 in the afternoon. BHB 4.8.

Also good. Also inside the zone. The diary entry for that day, in my own handwriting, reads:

broke 8pm.

One slice of pizza. One pita.

Twice I was standing precisely where my theory said I was untouchable. Twice I went down anyway.

I could have quietly left those two entries out of this article. Nobody would ever have known. I'm putting them at the center of it, because a theory that can't survive its own counterexamples isn't a theory, it's a slogan — and because if I'm going to ask you to believe what comes next, you should first watch me dismantle something I'd been saying for years.

What was actually happening

So I asked a different question.

Not what was my number when I broke. Instead: what had my number been doing all week?

That changed everything.

November 2024. I broke on the fifth. But the two days before, my ketones had already fallen off a cliff — 0.2, then 0.5, then 1.2. And across that entire attempt, the longest I ever strung together above 4.0 was two days.

When the break finally came, it wasn't pizza. It was olives.

I want you to feel how small that is. A handful of olives ended a run. Not a birthday cake, not a weak moment at a wedding — olives, because by then there was nothing at all holding the line. When the fuel is gone, it doesn't take temptation. It takes proximity.

May 2025. I broke at 4.8, which looked like a straight contradiction — until I read the week behind it. I had been above 4.0 for exactly one day. The six days before that: 1.7, 3.0, 2.7, 3.8, 3.3.

I hadn't been living in the zone. I'd wandered into it the previous afternoon and happened to still be standing there when the food showed up.

March 2026. Broke at 4.8 again — and this is the one that taught me the most, because this time I really had been holding. Five consecutive days above 4.0: 4.7, 5.4, 5.0, 5.6, and then 4.8 on the day it came apart.

Five days of doing everything right. Still not enough.

And then the case that convinced me.

October 2025. For eight straight days — the 28th of October through the 4th of November — my ketones stayed above 4.0 every single day. 4.5, 4.2, 7.0, 5.9, 4.6, 6.7, 7.1, 5.0.

Eight days. In four years of trying, the only time I ever held that zone for more than a week.

That attempt contains no break at all.

None. It kept going another three weeks. And when it eventually ended, it didn't end in a restaurant — it ended because my numbers slid out of the zone first. November 5th: 3.4. November 7th: 1.4. And then it was finished.

The protection lasted exactly as long as the level did. Not a day longer.

So here's what my own diary actually says, stated as plainly as I can manage:

It isn't the reading. It's the streak.

One good number on one good day protects almost nobody. What appears to produce real resistance is being above 4.0 and having been there for roughly a week. Every break I documented happened at fewer than seven consecutive days. The single time I got past seven, food stopped winning.

Why a week?

If you want the intuition rather than the biochemistry, think about antibiotics.

You've got an infection. You start the course. By day three you feel dramatically better — fever gone, appetite back, basically yourself again.

And every physician reading this already knows what happens next, because we spend our entire careers losing this argument: the patient stops taking the pills. They feel fine. The signal says finished.

But it's a ten-day course for a reason. Feeling better on day three tells you the drug is working. It tells you nothing about whether the job is done. Stop there and the infection comes back, often angrier, and the patient concludes that antibiotics don't work for them.

That is exactly the mistake I was making. My "feeling better on day three" was a number on a meter. I'd see 4.8 and think I'm protected — when all 4.8 ever meant was the process is running.

Or try this one, which is the version I keep coming back to.

You come home in January to a freezing house and turn on the heat. An hour later the thermostat reads seventy degrees.

Are you warm?

Sort of. The air is seventy. But the walls are cold. The floor is cold. The furniture, the window frames, the entire mass of the building is still cold — and the instant that furnace shuts off, you'll feel it within minutes, because nothing in that house is holding heat. It's all happening in the air.

Run the furnace for a week and something completely different has happened. The structure itself is warm. Now you can open the front door in January and the house barely notices.

The reading on the meter is the thermostat.

The streak is the walls.

I spent four years checking the thermostat and wondering why I kept getting cold.

The number that lies to you

There's a second thing buried in that data, and honestly it's the one that makes me angry, because it explains why none of us ever see the collapse coming.

I don't only track ketones. I track blood glucose too, and I add them into a single figure I call total energy — a rough measure of how much fuel my brain has available, from whichever source. It's a genuinely useful number. Below a certain point I can't work; above it I'm sharp.

Here's what four years of readings show.

When my ketones are at or above 4.0, my average blood glucose is 75 mg/dL. When my ketones drop below 1.0, my average glucose is 102.

Read those twice, because the two halves are moving in opposite directions. As the ketones drain away, glucose climbs to replace them — and almost precisely enough to fill the hole. So the total barely budges. The tank appears to stay full the entire time.

Imagine a fuel tank with two chambers and a gauge that only measures total liquid. The gasoline burns off. Water seeps in behind it. The needle doesn't move. It reads full all the way to the side of the road.

In my diary: of the 244 readings where my total energy looked strong, 105 of them had ketones below 4.0.

Forty-three percent of my good days were not good days. They were vulnerable days wearing a good day's clothes.

The cleanest example is from that March night. One in the morning, just after the pizza. Ketones 0.6 — essentially none. Glucose 137 — high. Total energy 8.21, which is a perfectly respectable figure. If you'd handed me only that total, I'd have said the man was fine.

The man had just broken a seventeen-day run.

The measurement that reassures you is the one that gets you. Feeling fine is not a measurement. A good deal of the time, feeling fine is the symptom.

Where I am now

I restarted on June 17th. Today is day 49.

On July 31st I tied my four-year record of forty-five days, and on August 1st I passed it. That was the milestone everybody could see, and I made a video about it, and it mattered to me more than I'd expected it to.

But under my own theory, the day count was never the number that mattered.

This is:

For twenty-one consecutive days — every day from July 15th through this morning — my ketones have stayed above 4.0.

My previous best, across four years and twenty-three attempts, was eight days.

Inside that stretch sit twelve consecutive days above 5.0 — a level I'd essentially never held for more than a day at a time in my life. My average daily peak across the twenty-one days is 5.39. Of every individual reading in that window, 74% are above 4.0, against a four-year average of 30%.

This is not a longer version of my old attempts. My forty-five-day record — the one I was proud of — averaged a BHB of 3.25, with barely a fifth of readings above 4.0 and a best streak of four days. I white-knuckled that one for six weeks and then broke.

Length is not the same thing as safety. You can survive a long time in a state that is quietly costing you everything you have.

For the first time in four years, I'm not visiting the zone. I'm living in it.

Which brings us back to the restaurant. I don't believe I sat through those meals because I finally grew a spine at sixty. I think I sat through them because I'm on day twenty-one of something I had never before sustained past day eight, and the thing my own diary predicted is happening on schedule.

One honest footnote. Yesterday I hiked for five hours in the Adirondacks — which is its own small miracle, given that seven weeks ago my joints hurt badly enough that I'd stopped going to the supermarket. (I've written before about why inflammation, rather than structural damage, is so often the thing that actually hurts.) My reading that evening was 4.2 — the lowest of the whole stretch.

That's not a bad day. That's an artifact. My legs had just spent five hours burning the very ketones I was trying to measure. The meter tells you what's in the blood at that instant, not what your body is producing — and if you don't know the difference, you'll misread your own results constantly, and draw exactly the wrong conclusions about what works.

The woman who stopped eating

Before I tell you the part I'm not going to publish, I want to tell you about a house call.

I'd seen this patient once before, many months earlier. Elderly. Bedbound. When I arrived, her grown children met me at the door and explained, in the resigned way families do once they've made peace with something, that their mother didn't really speak anymore.

I went in to see her.

And her eyes were sharp.

I want to be careful about how much weight I put on that, but I've spent thirty years walking into rooms like that one, and there is a look that goes with genuine cognitive decline. She didn't have it. She was following me across the room. She was tracking me — intent, focused, entirely present, watching this doctor with an attention that had nothing vague about it.

She was also so thin I didn't recognize her.

I went back out to the family and asked what had been happening. They told me she'd simply stopped eating. They didn't know why. She'd lost around sixty pounds. Her blood pressure had fallen so far they'd had to reduce her medication, and her diabetes medication had been stopped altogether because there was no longer anything to treat.

And then one of the children said, almost in passing:

"The only thing she'll take are the protein shakes you recommended."

I recommend protein shakes to nearly all my elderly patients. They almost never get enough protein, and a shake is a fast, easy way to deliver it. For most of them it's the single most useful thing I can suggest.

But she wasn't drinking them in addition to her meals. She'd replaced eating with them altogether.

I always carry a meter. I asked if I could check her ketones.

4.0.

There it was. Her appetite hadn't vanished for some mysterious reason, and it wasn't the beginning of the end. She was in ketosis — deep, unintentional, sustained ketosis — and one of the most reliable things ketosis does is switch off hunger completely. She wasn't refusing food. She simply had no signal telling her to want any.

And that sharpness in her eyes, the thing that had stopped me in the doorway — that was the other half of the same chemistry. That's what her brain looked like running on ketones.

My instruction to the family was the opposite of everything I do for a living. Give her some carbohydrate. Bring her out of it, gently, and her appetite will come back.

I think about that house call constantly, and here's why.

That woman had achieved, entirely by accident, the metabolic state I have failed to reach deliberately twenty-three times. She got the appetite suppression. She got the mental clarity. She also lost sixty pounds she could not afford to lose, on a body that had begun eating itself, while everyone around her — including, for a while, her doctors — assumed they were watching an old woman decline.

The same chemistry. Completely different meaning. For me, at sixty, deliberate and monitored, it's a therapy. For her, unintentional and unwatched, it was a slow emergency dressed up as peacefulness.

That is the entire reason I don't hand people a set of rules.

The part I'm not going to publish — and why

So. I've told you what I found. I'm not going to tell you exactly how I do it, and I'd rather explain that honestly than pretend it's modesty.

There is a trap in this work, and it's the one that woman fell into by accident.

Protein lowers ketones. That's straightforward physiology — your body can convert protein into glucose, so eating more of it pulls your ketone number down. Which means that the moment you start chasing a number, an obvious move presents itself.

Eat less protein. The number goes up. It feels like winning.

That is the one thing you must never do.

Protein isn't a variable you get to trade away. It's the macronutrient you're most obliged to defend. Cut it while your body is already running a deficit and your body will simply find protein somewhere else — which means it starts taking yours. Muscle first. Then hair. Then strength, which goes quietly, in increments too small to notice month to month until the day you can't get out of a chair.

And it does not politely stop at the muscles you can see. The heart is muscle too. Medicine learned this the hard way in the late 1970s, when crash diets built on poor-quality protein were sold commercially and people died of cardiac arrhythmias while following the instructions exactly.

The people who fall into this are almost never the greedy ones. They're the ones who aren't hungry. That includes a great many people on GLP-1 medications, and anyone far enough into ketosis that appetite simply switches off. They eat a salad. Or they eat nothing. And they feel disciplined doing it, right up until they're losing the tissue they most needed to keep. (The lean-mass problem is one I've written about in more detail here — a substantial share of the weight lost on those drugs isn't fat at all.)

So the real problem was never "get your ketones high." Anyone can get their ketones high by eating almost nothing, and a good number of them will hurt themselves doing it.

The real problem is holding ketones high while protein stays exactly where it belongs. Those two requirements pull against each other constantly. The resolution is different for every person, and it moves as you go.

That narrow, technical, genuinely difficult thing is what thirty years of practice actually bought me. It's also why I do this alongside people, watching the numbers as they move, instead of handing out rules and wishing them luck.

There's a second reason underneath all of it. Every small infraction chips away at the number, and the effect compounds. A little of this, a bit of that, and suddenly you're not at 5.2, you're at 4.1 — and now you need to find room somewhere else, and the whole structure becomes harder to hold.

It's a titration problem, not a willpower problem. Titration requires measurement, and measurement requires somebody reading the results who has seen a few thousand of them.

Which brings me to the thing I'd most like you to take from this, whatever you decide to do:

This is about getting healthier. It is not about vanity. A smaller number on a scale, or a larger one on a meter, is not the same thing as a healthier person. You can move both in the right direction while actively damaging yourself.

I have watched it happen. She had sharp eyes and sixty pounds gone and everyone thought she was fading.

What this doesn't prove

I want to be careful here, because the internet does not need another confident man generalizing from himself.

This is one person. Me. It's what researchers call an n-of-1 — a study with a single subject, no control group, no randomization, and an investigator who is also the participant and has an obvious commercial interest in how it turns out. That isn't nothing. It also isn't proof of anything general.

My diary isn't perfectly daily; I recorded when there was something to record. So "consecutive days" means consecutive days that have readings.

And only three of my twenty-three attempts have the word broke actually written in them. Not because the other twenty ended differently — they didn't — but because on those, by the time it happened, I was too discouraged to keep writing. The silence in a failure diary is data too. It's just the kind you can't analyze.

Most importantly: I'm on day 49 of 180. I have 131 days left, and I could still break.

I know where the hard parts are. Rosh Hashanah in mid-September. Then Sukkot — eight consecutive days constructed almost entirely around eating — landing right around day ninety. And Chanukah in December, falling across days 171 through 179, the final week of the entire experiment, which is either poetic or vindictive depending on which day you ask me. (Yom Kippur, for the record, is a fast day. That one I'm not worried about.)

But the holidays aren't really what frighten me.

It's an ordinary Tuesday. A restaurant, no occasion at all, end of a long week, people I love around the table.

That's the room where I've lost this twenty-three times.

If I break, I'll post that number the same day, on the same channel, with the same face. A record that only contains the good days isn't a record. It's an advertisement.

Day 49

Every morning I put a drop of blood on a strip and write down what it says. Every day I post that number publicly, and I'll keep doing it until the 13th of December.

Twenty-one days above 4.0. Previous best: eight.

I spent four years convinced I was a man with a discipline problem. My own handwriting says I was a man reading the wrong number — standing in the hallway checking the thermostat, over and over, in a house whose walls had never once been given a week to warm up.

I don't know how this ends yet.

That's rather the point of doing it where you can all see.

Follow the daily numbers: youtube.com/@neuraliftmd Free essay — "The Chains We Choose": neuralift.co The founding cohort is five people. If you'd rather do this with a physician watching your numbers than guess at it alone, book a discovery call.

Educational only, and not medical advice. This is a documented self-experiment in one person — me. The patient described here is presented without identifying details. Therapeutic ketosis can interact with prescription medication and is not appropriate for everyone. Talk to your own physician before making any changes.