Same Number, Two Different Brains

Why one ketone reading can’t tell you how your brain is doing — and what a child with epilepsy, a rat, a tired hitter and my own overtrained heart taught me about the long game

Three days, one number

Picture your car's fuel gauge reading three-quarters full on three different mornings. On one of those mornings the car sputters and stalls. On the other two it runs beautifully. You'd quickly figure out that the gauge isn't telling you the whole story. Something else is making the difference: the engine, the oil, how hard you drove it the day before.

That's what happened to me this month. I'm going to walk you through it slowly, one number at a time, because I lived it and even I get cross-eyed if I go too fast.

Number one: the fuel gauge

Every morning I prick my finger and measure my blood ketones. Ketones are a fuel your liver makes out of fat when you eat very few carbohydrates. Your brain can run on them in place of sugar. The meter gives me a number. Most people on a ketogenic diet land somewhere between 0.5 and 3. I aim for 4 or higher, which is high. (I'll explain why later.)

On three days in the past nine — September 29, October 6 and October 7 — my first reading of the morning was exactly the same: 4.5.

Same fuel gauge, three times.

Number two: how sharp my brain is

Every morning I also take a simple 3-minute test on my laptop. A counter pops up on the screen at random moments, and I hit a key the instant I see it. The computer measures how long I take, in milliseconds, which are thousandths of a second. Lower is better: it means a quicker, more alert brain. This is the same test sleep scientists use to tell when someone hasn't slept enough.

Here's how my three 4.5 days went:

• September 29: 299 milliseconds, the slowest of my whole run.
• October 6: 250 milliseconds.
• October 7: 249 milliseconds.

That's a gap of about 50 milliseconds. It sounds tiny — it's less time than a blink. But think of hitting the brakes on the highway. At 60 miles an hour, 50 milliseconds is the difference between stopping in time and rolling about four and a half feet further. On this test, a slowdown that size is the kind sleep researchers link to losing sleep.

Same fuel gauge. A very different engine.

Number three: the season, not the weather

So what was different? Not the reading. The week behind it.

To see that, I use a third number: my 7-day average, the average of my last seven morning readings. If one reading is today's weather, the 7-day average is the season. One warm day in January doesn't make it summer.

• On September 29, my 7-day average was only about 2.2. I'd been stuck in the 2s for three weeks. That 4.5 was one warm day in the middle of winter.
• On October 6 and 7, my 7-day average was about 4.1 to 4.3. I'd had a full week of 4s. That 4.5 was an ordinary day in the middle of summer.

The fuel wasn't the only thing that was different. The week before September 29 was also short on sleep, including one night of under five hours. And that evening — after my slow test — I did the hardest workout I've attempted in years. More on that later, because it turned into its own lesson.

Number four: how well I recharged overnight

My watch gives me one more number every morning: a recovery score from −10 to +10. Think of it like your phone's battery after a night on the charger. +10 means fully charged. −10 means the charger barely worked. The watch figures it out mostly from my heart rate and the rhythm of my heartbeats while I sleep.

Here's the surprise. Before September 29 and before October 6, my recovery score was −10, the worst it can be. So one bad night can't explain the difference either. Both mornings started with an empty battery. What separated them was the season behind them.

Here are all four numbers side by side:

And it wasn't just those three days. I started the reaction test on September 28, at the tail end of my September slump. My first three tests averaged about 287 milliseconds. Over the next seven, as my 7-day average climbed back toward 4, they averaged about 246, roughly 40 milliseconds faster. Inside that better week, the morning reading didn't predict the score. My lowest reading of the week, 3.7, came with my fastest time, 238. My highest, 5.1, came with nearly the same time, 240. My brain's speed followed the season, not the day.

Saturday: my best number on one of my worst days

On Saturday, October 3, my meter read 5.1, one of my highest readings of the run. For months a number like that would have had me grinning and filming a video. But it was my sixth day of fasting, and my body wasn't celebrating.

My blood pressure (BP) sitting down was 82/62. Blood pressure is written as two numbers. The top number is the pressure in your arteries when your heart squeezes. The bottom number is the pressure while your heart rests between beats. Many healthy adults sit somewhere around 110/70 to 120/80. My 82/62 was the lowest of my whole run. Think of a garden hose with the tap turned way down: the water still comes out, but without much push. When I stood up, my BP rose to 109/80. That told me my body was doing its job and compensating, but it was working hard to do it.

My blood sugar was 111 on an empty stomach, which is high for me. When the body is under strain, such as a long fast stacked on top of a hard workout, it releases stress hormones that push sugar into the blood even when you haven't eaten. It's the body pulling its own fire alarm.

I felt jittery, and my brain felt like a wet sponge. (Oddly, my symbol-matching test that afternoon was my best ever. Bodies are strange. I'll come back to that.)

So the highest reading on my meter came on one of my worst days. The fuel gauge said "full." The engine said otherwise.

Yesterday: a foggy morning, a sharp afternoon

Yesterday morning at 11, I took my tests with a head cold and pre-interview nerves, and most of my scores were the weakest of the week. At 2 PM I recorded my first podcast interview. When we finished, the host told me it was one of the best interviews he has done. (More on that when it airs.)

If one morning's tests couldn't predict that same afternoon, one morning's ketone reading can't either.

The big picture

Over the whole run so far — 109 days with readings — my average has been 4.07, and I've been at 4 or higher on 61 of those days. My best week, in late July, averaged 5.6. From mid-July into August I stayed above 4 for 21 days in a row. Then most of September slipped into the 2s, a cold snap in the middle of my summer. By September 28 my 7-day average had fallen to about 2. By yesterday it was back up to 4.3. (One honest note: since Day 92 I post my first reading of the morning, not my best of the day, so my recent weeks are graded on a tougher curve.)

Last week I wrote that a single good week is weather, and a pattern that keeps coming back is climate. This week I want to push that idea further. The more I read, the more I suspect the brain doesn't care much about the weather. It cares about the season.

What this letter is about

If you measure anything about your health — blood sugar, blood pressure, weight, ketones — this letter is about why one reading can fool you, and what a better question looks like. The short version: don't just ask "what's my number today?" Ask "what kind of week has it been?"

I'll show you the science behind that idea, the science that argues against me, and the time this month I pushed my body too hard and it pushed back.

Quick glossary

Ketones (BHB): a fuel the liver makes from fat when carbohydrates are very low. The main one is beta-hydroxybutyrate, or BHB, measured in millimoles per liter (mmol/L). It's the number on my meter.

First reading: since Day 92, the number I post is my first reading of the day. No retakes.

7-day average: the average of my last seven morning readings. It smooths out single good or bad days so the trend shows through. The weather versus the season.

Nutritional ketosis: usually defined as a blood BHB of about 0.5 mmol/L or higher.

Reaction-time test: a 3-minute test where I hit a key the instant a counter appears on screen. It measures alertness in milliseconds (thousandths of a second); lower is better. Sleep scientists call it the PVT (psychomotor vigilance test), and it's their standard test for detecting lost sleep.

Blood pressure (BP): written as two numbers, like 120/80. The top is the pressure when the heart squeezes; the bottom is the pressure while it rests between beats.

Recovery score: my watch's overnight "battery recharge" score, from −10 (worst) to +10 (best). The watch calls it ANS charge, for autonomic nervous system: the automatic wiring that runs your heart rate, blood pressure and stress responses.

HRV (heart-rate variability): the small changes in time between heartbeats. Higher usually means a rested, recovered nervous system.

Mitochondria: the tiny power plants inside cells that turn fuel into usable energy.

Cortisol: the body's main stress hormone. When the body senses a threat, including a very hard workout, cortisol pushes blood sugar up.

Rival explanation: something else that changed at the same time and could explain the result just as well. Scientists call it a confounder.

"How high is high enough?"

That was one of the first questions the podcast host asked me. It's a good one.

The common answer is that nutritional ketosis starts around 0.5, and around 1.0 many people call it solid. I aim for 4.0 or higher, which is a lot.

Why so high? Because the oldest and best evidence that ketones act on the brain comes from epilepsy, where doctors have used the diet to treat children since the 1920s. In those children, the level mattered.

In one study, seizure control tracked the children's blood BHB, and the clearest benefit showed up above about 4 mmol/L [1]. And the international expert guidelines warn that eating carbohydrate can quickly reverse a child's ketosis and bring the seizures back [3]. The level was doing work. (For what it's worth, my own ketones held at 4.5 through this week's cold.)

Here is the part that matters most for this letter, though. Even in epilepsy, doctors don't judge the diet by one day, or by one week.

Why time may matter more than the day

Let me slow down here, because this is the core of what I want you to take away.

1. The epilepsy clinics give it months. When the diet works, it often works fast, usually within the first one to two weeks [2]. But the international expert guidelines tell families to give it at least three months before deciding it failed [3]. Time on the diet is part of the treatment.

2. The brain takes weeks to change fuel. In a classic 1967 study, researchers placed catheters in the blood vessels going in and out of the brain in three patients with obesity who were fasting under medical supervision for five to six weeks. By then, ketones, not glucose, had become the brain's main fuel [4]. The brain can run on ketones, but the switch happens over weeks. It doesn't flip on with one high reading.

3. In rats, the brain built more power plants. Researchers kept young rats in ketosis for about three weeks. Then they looked at the hippocampus, the brain's memory center, under an electron microscope. They found a 46 percent increase in the density of mitochondria, and 39 of 42 genes for mitochondrial proteins that changed had been turned up [5,6]. Biologists call this mitochondrial biogenesis, which means building new mitochondria. It's a construction project, and construction takes time.

4. Some of the benefit may outlast the diet. At Johns Hopkins, 66 children who had become seizure-free on the diet stopped it after about two years, and almost all of them were off seizure medicines too. Eighty percent stayed seizure-free afterward. Only 13 of the 66 (20 percent) had their seizures come back, typically a couple of years later, and more than half of those became seizure-free again [7]. One caution: these were the luckiest children, the roughly one in ten who became seizure-free on the diet in the first place. Still, something seems to have changed that didn't depend on ketones still being in the blood that morning.

Put those four together and you get the idea that pulled me toward averages. If ketones help the brain, the help probably doesn't come from today's number. It comes from weeks of steady fuel: long enough for the brain to retool, and maybe long enough to build something that lasts.

That idea has a champion. Dr. Christopher Palmer, a Harvard psychiatrist, argues in his book Brain Energy that many mental illnesses are, at root, energy problems in the brain, with the mitochondria not keeping up [8]. It's a theory, not a proven fact, but people are testing it. In a small Stanford pilot study published in 2024, 21 adults with schizophrenia or bipolar disorder who also had metabolic problems followed a ketogenic diet for four months. Among those who started with significant symptoms, 79 percent improved, and the people who kept their ketosis consistent did better than those who kept it only partly [9].

This summer the first randomized trial arrived. Researchers at UCSF randomly assigned 58 adults with schizophrenia or bipolar disorder to a ketogenic diet or their usual diet for one month, and 25 chose to continue the diet for four months [10]. The authors wrote that one month appeared to be too short to capture the mental-health benefits. The gains in mood, symptoms and thinking showed up in the four-month group. Time, again.

My sleep shows the same time course

My own nightly sleep ratings — the "how did you sleep" button I press every morning on the watch — tell the same story across all four of my ketosis attempts [17]:

The first two weeks are the worst stretch I've recorded, worse even than being off. After that, my good nights nearly triple compared with being off, and my bad nights fall from 37 percent to 22. My nights are also less broken in ketosis: my watch's continuity score averages 2.6 out of 5 in ketosis versus 2.1 off, roughly one fewer stretch of lying awake a night [17]. That matches what I've felt for years: rough at first, and by a month, another world. What it does not show is that a higher ketone level, once I'm in, makes the night better. Across 141 nights in ketosis, my 7-day average and my deep-sleep minutes had essentially no relationship. Being in mattered. How high didn't.

☕ Coffee break

If you got lost, here's the story so far.

Three times my meter read 4.5. The day that sat on top of a weak week — weak in ketones and short on sleep — came with my slowest reaction time. The days that sat on top of a strong week came with much faster ones.

The science I trust suggests ketones act on the brain over weeks, not hours. Children with epilepsy are given months. The brain switches fuel over weeks. In rats, the brain built more mitochondria in about three weeks. My own sleep ratings turn around after two weeks, not two days.

So I'm watching the 7-day line, not the daily number.

Now comes the part where I argue with myself.

Where the evidence pushes back

The level may matter after all. In the UCSF trial, within that first month, the people with higher ketone levels had lower depression scores, even after accounting for weight loss [10]. So it may not be either-or. Time on the diet and the level reached both seem to count. What my data argues against is judging the level by a single morning.

The rat finding doesn't always repeat. In a 2021 study, middle-aged mice spent one month on a ketogenic diet. Their leg muscles grew more mitochondria, but their brains, including the hippocampus, did not [11]. Maybe age matters. Maybe a month is too short. Maybe it's the species, or the diet. I don't know. But it's a real result, and it's why I won't tell you ketones build new brain mitochondria in people. That has not been shown.

Two of my tests are fooled by practice. On Monday I titled a video "A week at 4 and my brain woke up." I still feel that. But when I lined up every memory test since August, the scores had climbed steadily the whole time. In late August, with my ketones around 4.3 at test time, I averaged about 6.4 numbers backward. In the September slump, with ketones around 2.1, I averaged 7.4, which is better. Now, back around 4.2, I'm at about 8. My scores kept rising right through my lowest-ketone weeks. That's what practice looks like, because I've taken this test dozens of times. My symbol-coding scores have climbed the same way, from 55 at my first session to the 60s and 70s since, and my best one landed on a day my body was running on empty. A test that rewards practice can't tell me much about ketones yet. So I can't credit ketones, or the average, for those scores.

Why I trust the reaction-time test more. It's the one that doesn't reward practice. When sleep researchers at Penn gave people this brief test over and over, the typical reaction time showed no systematic improvement with repetition [12]. That's a big reason sleep scientists rely on it. So when my typical time dropped about 40 milliseconds, practice is the least likely explanation. Three caveats. First, mine is a home version on a laptop, not a lab device. Second, my slow sessions were my first three, so first-time nerves can't be fully ruled out. Third, my lapses — the moments I zone out completely — didn't improve. Only my typical speed did.

My sleep only partly outlasted the diet. My sleep scores fell back every time I quit. My deep-sleep and REM minutes, in my most recent break, didn't: they matched my ketosis attempts. That could be carryover from years of going in and out of ketosis — or it could be fitness, weight and time. I can't tell yet.

The psychiatry studies are early. The Stanford pilot had 21 people and no comparison group. The UCSF trial was randomized for only one month, and its four-month results came from the 25 people who chose to continue. Encouraging, not proof.

And there's a bigger rival explanation: me. My slow reaction times at the end of September came during a stretch of short nights, right around the hardest workout I've done in years. That brings me to the confession.

Training like I'm 20 at almost 61

This is the part of the month I'm least proud of, and the part I learned the most from.

In September I went back to hard training after years away. First, on September 4, an hour-long run — my first run in about seven years. Then, on September 29, my first Sprint 8: eight 30-second all-out bursts with a minute and a half of recovery between them. My heart rate was hitting about 160 in the bursts — close to the top of what's predicted for my age. The plan was to do it every other day. I did it once.

My body sent the bill.

On the night of September 24, my recovery score was +8.8, with a sleeping heart rate of 42 and an HRV of 88, one of my best nights. Over the eight nights since that workout, I haven't had a single positive night, and three of them hit the floor at −10. My HRV this past week has sat between the 50s and the low 70s. Then came Saturday's sitting blood pressure (BP) of 82/62, the low-pressure garden hose, on the sixth day of a fast.

After that one session I put the plan on hold. More than a week later my recovery still isn't back, partly because I picked up a cold Monday night. The workout took minutes. Paying for it is taking more than a week. The sprints stay on hold until I've recovered, and then I'll build up slowly.

So is my faster reaction time the ketones, or is it simply that I stopped beating myself up and started sleeping again? I can't separate the two. The improvement showed up on October 1, two days after that one sprint session, which is what you'd expect if recovery were the reason. One clue points the other way: two of my fast days, October 4 and October 6, came right after −10 nights. Once the average was up, a bad recovery night alone didn't slow me down. My best guess is that it was both — the stage and the recovery — which is the whole point of this letter.

Here's what I think happened with the training, labeled clearly. I'm not conditioned for that kind of effort yet. To an untrained body, all-out sprints read as an emergency, and the body answers emergencies with stress hormones like cortisol, which push blood sugar up. In one well-known study, cortisol barely moved at easy effort but rose about 40 percent at moderate-hard effort and about 83 percent at hard effort [13]. A trained body reads the same workout as a smaller threat. That's standard exercise physiology, and I think it's likely. But I haven't measured my own cortisol, so for me it's a reasonable guess, not a result.

There's good evidence for a better way. In a small Finnish study, healthy men either trained on a fixed schedule or let their morning HRV decide the day: hard when HRV was up, easy when it was down. The HRV-guided group improved their fitness more [14]. They did the same kind of work, just with smarter timing.

So here's my new rule, and I'm writing it down so you can hold me to it. Base first: walking, strength work, steady effort. Sprints come back only on mornings when my recovery score is positive and my sleeping heart rate is back near my baseline. The watch decides, not my ego.

Slumps and hot streaks

I grew up loving the late-1990s Yankees. I've always wondered why great hitters go into slumps, weeks where nothing falls in, and then hit streaks where they can't miss.

I think brain energy may be part of it. To be clear, that's my hunch, not a finding. Nobody has ever measured a hitter's ketones during a slump.

There is real data, though, that a tired brain plays worse baseball. A Vanderbilt sleep neurologist looked at all 30 teams in the 2012 season. Hitters' plate discipline, meaning how well they laid off pitches outside the strike zone, got worse as the season wore on. For 24 of the 30 teams, it was worse in September than in April. He pointed to the grind of the season [15]. Plate discipline is a brain skill: see the ball, decide, and hold back, all in a fraction of a second. It's not far from my reaction-time test.

A second study measured sleepiness in 80 major-league players before the 2010 season, then checked three seasons later who was still in the league. The sleepier the player, the less likely he was still there. Of players who scored 5 on the sleepiness scale, 72 percent were still in the league; at a score of 15, just 14 percent were [16]. (Both studies were presented at a sleep-medicine conference, not yet full journal papers.)

So fatigue and poor recovery measurably change how the brain performs, over a season and over a career. I trust that. Whether brain fuel is part of the story is the open question I'd love to see someone study.

How sure am I?

Established: my numbers as I've reported them [17]. In children with epilepsy, seizure control tracked blood ketone levels [1]. Expert guidelines say to give the diet months before judging it [3]. During weeks of fasting, the human brain switches to ketones as its main fuel [4]. Among children who became seizure-free on the diet and then stopped it, 80 percent stayed seizure-free [7]. The brief reaction-time test doesn't improve with practice [12]. Fatigue and sleepiness affect baseball performance and career length [15,16].

Self-observation (one person, one stretch): my typical reaction time got about 40 milliseconds faster as my 7-day average rose from about 2 to about 4, and it didn't follow the daily number. It also overlapped with my stopping hard training and sleeping more, so I can't fully separate them. My own sleep ratings are worst in the first two weeks of ketosis and best after a month, across four attempts.

Likely: a single ketone reading says little about how a day will go. For me, recovery — sleep, eating, fluids, training load — says more. Hard training without a base drove my recovery scores down.

Plausible, not proven: that ketones help the brain over weeks rather than days; that stress hormones explain my Saturday; that brain energy plays a part in slumps.

Mixed: whether the level itself matters. My sleep data says no, once I'm in; the UCSF trial found higher ketones went with lower depression scores [10].

Animal-only and mixed: ketones building new brain mitochondria happened in young rats after three weeks [5,6] but not in middle-aged mice after a month [11].

Not supported by my data: that ketones are improving my memory or coding scores. Practice explains them at least as well.

What I'm changing

Four things.

I watch the 7-day line, not the daily number. I still post the first reading every day, because it keeps me honest. But I judge the run by the average.

I train by the morning score. Base first, and sprints only on green mornings.

I'm testing my brain more carefully. This Sunday I run my first full weekly battery, ten tests instead of the daily few, so I can watch the slow trend without getting fooled by one foggy morning or by practice alone.

I'm logging what the watch can't see. Starting this week I fill in a 30-second morning sleep log: how long it took to fall asleep, bathroom trips, whether I fell straight back to sleep, whether I woke disoriented, whether I remembered dreams, and how restored I felt. Those are the changes I notice most in ketosis, and a wrist watch can't measure any of them. If you'd like the one-page template, reply to any issue of The NeuraLift Letter and I'll send it.

The real test of the "lasting" question is still ahead. When this run ends at Day 180, I'll keep measuring. If something was built, it should still be there.

The sentence I trust now

The level sets the stage. Recovery does the performing.

A 4.5 on top of a weak week — low ketones, short nights — gave me my slowest reaction time. A 4.5 on top of a strong week, with a cold coming on, gave me a much faster one. A 5.1 on the sixth day of a fast, while I was under-drinking and still paying for over-training, gave me a sponge for a brain. And a foggy morning didn't stop a good afternoon, because one morning was never going to be the whole story.

If you're chasing a number, any number, try this. Before you look at today's reading, ask how you slept, whether you ate, and how hard you pushed yesterday. Then look at the last seven days. That's the number that knows something.

— Barry

I post my blood ketone number every day — including the bad ones — in a 90-second video: youtube.com/@neuraliftmd. Day one is there. So is every failure since.

If you want to do something like this with a physician watching, I’m taking on a small founding cohort. It starts with a conversation: calendly.com/hello-neuralift/30min

Educational only, not medical advice. Ketogenic diets for epilepsy or psychiatric conditions should be done only with a medical team. Don’t change any medication on your own.

Where all this comes from

1. Gilbert DL, Pyzik PL, Freeman JM. The ketogenic diet: seizure control correlates better with serum β-hydroxybutyrate than with urine ketones. J Child Neurol. 2000;15(12):787–790. — Children on the classic ketogenic diet. Seizure control correlated with blood BHB (p = 0.03), with the clearest benefit above about 4 mmol/L; urine ketones were a poorer guide.
2. Kossoff EH, Laux LC, Blackford R, et al. When do seizures usually improve with the ketogenic diet? Epilepsia. 2008;49(2):329–333. — When the diet works, improvement usually shows within the first one to two weeks.
3. Kossoff EH, Zupec-Kania BA, Auvin S, et al. Optimal clinical management of children receiving dietary therapies for epilepsy: updated recommendations of the International Ketogenic Diet Study Group. Epilepsia Open. 2018;3(2):175–192. — International consensus guidelines: give dietary therapy at least three months before judging it, and remember that carbohydrate, including hidden carbohydrate in medicines, can quickly reverse ketosis and bring seizures back.
4. Owen OE, Morgan AP, Kemp HG, et al. Brain metabolism during fasting. J Clin Invest. 1967;46(10):1589–1595. — Three patients with obesity fasting five to six weeks under supervision. Catheters in the brain’s blood vessels showed β-hydroxybutyrate and acetoacetate had replaced glucose as the brain’s main fuel.
5. Bough KJ, Wetherington J, Hassel B, et al. Mitochondrial biogenesis in the anticonvulsant mechanism of the ketogenic diet. Ann Neurol. 2006;60(2):223–235. — Young rats on a calorie-restricted ketogenic diet for about three weeks; genes for energy metabolism and mitochondrial proteins turned up in the hippocampus. Animal data.
6. Bough K. Energy metabolism as part of the anticonvulsant mechanism of the ketogenic diet. Epilepsia. 2008;49(Suppl 8):91–93. — The author’s own review of his rat work: a 46% higher density of mitochondrial profiles under the electron microscope, and 39 of 42 changed mitochondrial-protein genes turned up. Animal data.
7. Martinez CC, Pyzik PL, Kossoff EH. Discontinuing the ketogenic diet in seizure-free children: recurrence and risk factors. Epilepsia. 2007;48(1):187–190. — Of 557 children started on the diet at Johns Hopkins, 66 stopped it after a median 2.1 years because they were seizure-free (92% were also off medication). Thirteen (20%) relapsed, at a median 2.4 years; seven of them became seizure-free again.
8. Palmer CM. Brain Energy. BenBella Books; 2022. — A book-length argument that mental illnesses are metabolic disorders of the brain. A hypothesis, not a trial.
9. Sethi S, Wakeham D, Ketter T, et al. Ketogenic diet intervention on metabolic and psychiatric health in bipolar and schizophrenia: a pilot trial. Psychiatry Res. 2024;335:115866. — Single-arm pilot: 21 adults with schizophrenia or bipolar disorder and metabolic abnormalities, four months. Among those with elevated symptoms at the start, 79% improved at least one point on the Clinical Global Impression scale; closer adherence went with more benefit. No control group.
10. Abram SV, Kyner JM, Vu A, et al. Metabolic improvements with a ketogenic diet correlate with symptom improvement in psychosis: a randomized controlled trial. Schizophr Bull. 2026;52(4):sbag082. — The first randomized trial: 58 adults with schizophrenia-spectrum or bipolar I disorder, a one-month comparison against usual diet, then a four-month single-arm extension (25 people). Higher ketones tracked lower depression scores independent of weight loss; symptoms and thinking improved by four months. The authors judged one month too short.
11. Zhou Z, Vidales J, González-Reyes JA, et al. A 1-month ketogenic diet increased mitochondrial mass in red gastrocnemius muscle, but not in the brain or liver of middle-aged mice. Nutrients. 2021;13(8):2533. — Middle-aged mice, one month. Mitochondria increased in leg muscle but not in the prefrontal cortex, hippocampus or liver.
12. Basner M, Hermosillo E, Nasrini J, et al. Repeated administration effects on psychomotor vigilance test performance. Sleep. 2018;41(1):zsx187. — Repeated sessions of the brief 3-minute reaction-time test. Mean and median reaction time and lapses showed no systematic change with practice; a few secondary measures changed slightly.
13. Hill EE, Zack E, Battaglini C, et al. Exercise and circulating cortisol levels: the intensity threshold effect. J Endocrinol Invest. 2008;31(7):587–591. — Cortisol rose with exercise intensity: little at low effort, about 40% at moderate-hard effort and about 83% at hard effort.
14. Kiviniemi AM, Hautala AJ, Kinnunen H, Tulppo MP. Endurance training guided individually by daily heart rate variability measurements. Eur J Appl Physiol. 2007;101(6):743–751. — Healthy, moderately fit men. Letting each morning’s HRV choose hard or easy days improved fitness more than a fixed plan. Small study.
15. Kutscher S. Plate discipline across the 2012 MLB season. Presented at SLEEP 2013, Baltimore; summarized by the American Academy of Sleep Medicine, May 30, 2013. — Conference abstract, not a journal paper. Swinging at pitches outside the strike zone rose as the 2012 season went on; September was worse than April for 24 of 30 teams.
16. Winter WC. Sleepiness and career longevity in Major League Baseball players. Presented at SLEEP 2013, Baltimore; same AASM summary. — Conference abstract, not a journal paper. Pre-season sleepiness scores in 80 players predicted who was still in the majors three seasons later.
17. Dublin B. Personal blood BHB log (first reading of the day since Sep 16), Polar sleep and overnight recovery data with nightly self-ratings (Jan 2025 – Oct 2026; deep/REM minutes dictated from the Polar app), and daily cognitive tests (digit span since Aug 17; reaction time, symbol coding and word fluency since Sep 28), through Oct 7, 2026. Unpublished N-of-1 dataset. — My own record. One person, not blinded; the reaction-time test is a home laptop version, not a lab device.