Blueprint

MacroFactor logs calories and macros for everything I eat, but vitamins and minerals only for foods from its own database. A label I typed in, a konbini bento, a bowl from Matsuya: none of them carry any. Averaging what’s there would make me look short of nearly everything.

So the diet review weighs each nutrient by its coverage, the share of my calories from foods that report it, and calls a shortfall low only when at least 80% of those calories are covered. Below that, the honest answer is “can’t tell”. On MacroFactor’s numbers alone, that was the answer for 19 of 28 nutrients.

To fill the gaps, Claude estimates each food MacroFactor leaves blank: from its ingredients, or for packaged food, from the maker’s nutrition label (栄養成分表示), found with a web search. Code checks each estimate’s calories against the logged entry and its values against sane ranges. So far that’s 91 foods, 31 of them estimated with high confidence. A low-confidence estimate counts only provisionally and asks me a question in the inbox (“sauce only, or coated pieces?”); my answer re-estimates the food.

What 17 days of logged food can tell

Each nutrient as a share of its daily target (US reference intakes, men 19–30), 29 Sept to 26 Oct. Hollow: MacroFactor's numbers alone. Filled: with the AI estimates.

With estimates MacroFactor alone

0 of 28 can't tell · 11 low · 2 over the limit 19 of 28 can't tell · 0 low · 1 over the limit

0%50%100%150%200% VerdictCovered

Vitamins

Vitamin A low can't tell 28% + 72% 28%

Vitamin C 438% 415% met met 28% + 72% 28%

Vitamin D low can't tell 12% + 88% 12%

Vitamin E low can't tell 36% + 64% 36%

Vitamin K met met 28% + 72% 28%

Thiamin (B1) met can't tell 48% + 52% 48%

Riboflavin (B2) 267% 228% met met 54% + 46% 54%

Niacin (B3) 389% met met 50% + 50% 50%

Pantothenic acid (B5) met can't tell 46% + 54% 46%

Vitamin B6 446% 374% met met 56% + 44% 56%

Folate low can't tell 47% + 53% 47%

Vitamin B12 743% 604% met met 25% + 75% 25%

Choline low can't tell 35% + 65% 35%

Minerals

Calcium low can't tell 51% + 49% 51%

Copper met can't tell 46% + 54% 46%

Iron met can't tell 48% + 52% 48%

Magnesium low can't tell 47% + 53% 47%

Manganese met can't tell 46% + 54% 46%

Phosphorus met can't tell 47% + 53% 47%

Potassium low can't tell 51% + 49% 51%

Selenium 201% met met 46% + 54% 46%

Sodium over the limit met 70% + 30% 70%

Zinc low can't tell 46% + 54% 46%

Fats, fibre and limits

Fibre low can't tell 49% + 51% 49%

Omega-3 (ALA) low can't tell 33% + 67% 33%

Saturated fat a limit, no target within the limit can't tell 51% + 49% 51%

Added sugar a limit, no target over the limit over the limit 7% + 93% 7%

Caffeine a limit, no target within the limit can't tell 4% + 96% 4%

MacroFactor aloneWith estimatesTarget

Source: Blueprint's diet review, run twice on the same food log. "Covered": the share of calories from foods that report the nutrient, measured + estimated. Below 80%, a shortfall can't be called low.

See the numbers

Nutrient	MacroFactor alone	With estimates	Covered
Vitamin A	18%, can't tell	34%, low	28% measured + 72% estimated
Vitamin C	415%, met	438%, met	28% measured + 72% estimated
Vitamin D	28%, can't tell	47%, low	12% measured + 88% estimated
Vitamin E	24%, can't tell	48%, low	36% measured + 64% estimated
Vitamin K	107%, met	131%, met	28% measured + 72% estimated
Thiamin (B1)	86%, can't tell	112%, met	48% measured + 52% estimated
Riboflavin (B2)	228%, met	267%, met	54% measured + 46% estimated
Niacin (B3)	196%, met	389%, met	50% measured + 50% estimated
Pantothenic acid (B5)	72%, can't tell	144%, met	46% measured + 54% estimated
Vitamin B6	374%, met	446%, met	56% measured + 44% estimated
Folate	77%, can't tell	97%, low	47% measured + 53% estimated
Vitamin B12	604%, met	743%, met	25% measured + 75% estimated
Choline	30%, can't tell	66%, low	35% measured + 65% estimated
Calcium	36%, can't tell	50%, low	51% measured + 49% estimated
Copper	59%, can't tell	110%, met	46% measured + 54% estimated
Iron	93%, can't tell	123%, met	48% measured + 52% estimated
Magnesium	32%, can't tell	64%, low	47% measured + 53% estimated
Manganese	85%, can't tell	146%, met	46% measured + 54% estimated
Phosphorus	74%, can't tell	168%, met	47% measured + 53% estimated
Potassium	38%, can't tell	68%, low	51% measured + 49% estimated
Selenium	111%, met	201%, met	46% measured + 54% estimated
Sodium	124%, met	168%, over the limit	70% measured + 30% estimated
Zinc	35%, can't tell	71%, low	46% measured + 54% estimated
Fibre	34%, can't tell	42%, low	49% measured + 51% estimated
Omega-3 (ALA)	29%, can't tell	47%, low	33% measured + 67% estimated
Saturated fat	—, can't tell	—, within the limit	51% measured + 49% estimated
Added sugar	—, over the limit	—, over the limit	7% measured + 93% estimated
Caffeine	—, can't tell	—, within the limit	4% measured + 96% estimated

With the estimates, every nutrient gets a verdict, and the picture changes. Eleven come out low, from vitamin A to omega-3. Sodium, which looked fine at 124% of its target, is over its limit at 168%.

The estimates have a twin that checks the numbers MacroFactor does have. It started when I noticed a can of Monster logged with ten times its real sodium. Now, after every import, the biggest sources of sodium, saturated fat and added sugar are checked against their labels: 21 checks so far, and 6 corrections. One energy drink was logged with 568 mg of sodium per 355 ml, against 60 on its label; a cup noodle with 0.6 g of saturated fat had 11.5.

The yardstick is a younger body: men of 20. A switch on the chart puts me next to men my own age instead. They’re US men, since I’m American, from NHANES, the CDC survey that examines and takes blood from a sample of Americans weighted to stand for the whole country.

The tool reads NHANES’s public files directly, 61 of them and 99 MB in SAS’s old transport format, and works out weighted percentiles for every age from 20 to 77, each pooled with the two years either side. Checked against the CDC’s own published tables for men in their twenties, the sample size matches exactly, at 801 men, and the BMI percentiles agree within 0.1.

NHANES doesn’t measure what WHOOP does, so heart-rate variability, resting heart rate and breathing rate overnight come from published wearable studies instead. Those were Fitbit users, so they’re marked approximate.

Next to US men of 2026, my age

18 markers measured so far. Bands: where the middle 95% and the middle half of men fall; the tick is the median. The dot is me, at 26.

Compare with Men of 20 Men my age (26)

Better than most men of 20 on 4 of 12 markers; for the other 6, the middle is best Better than most men of 26 on 5 of 12 markers; for the other 6, the middle is best

Furthest ahead of men of 20: systolic blood pressure (85%), HDL (84%), resting heart rate (83%) and HRV (65%). Furthest behind: AST (23%), γ-GTP (24%), diastolic blood pressure (36%), VO2max (38%), waist (38%) and ALT (39%). Percentages: the share of them I'm better than.

Furthest ahead of men of 26: systolic blood pressure (85%), resting heart rate (82%), HDL (81%), HRV (77%) and LDL (57%). Furthest behind: AST (23%) and γ-GTP (35%). Percentages: the share of them I'm better than.

Heart & fitness

VO2max 43 ml/kg/min better than 38% better than 50% higher is better · typical of 28

HRV, overnightapprox. 82.3 ms better than 65% better than 77% higher is better · typical of men younger than 20

Resting heart rate, overnightapprox. 54.6 bpm better than 83% better than 82% lower is better

Respiratory rate (sleep)approx. 15.6 breaths/min mid-range (52nd percentile) mid-range (53rd percentile) the middle is best

Blood pressure

Systolic blood pressure 107 mmHg better than 85% better than 85% lower is better · typical of men younger than 20

Diastolic blood pressure 69 mmHg better than 36% better than 49% lower is better · typical of 26

Body

Weight 78.8 kg mid-range (53rd percentile) mid-range (44th percentile) the middle is best

BMI 25.4 kg/m² mid-range (52nd percentile) mid-range (45th percentile) the middle is best

Waist circumference 93 cm better than 38% better than 45% lower is better · typical of 28

Cholesterol & fats

LDL cholesterol 95 mg/dL better than 46% better than 57% lower is better · typical of 21

Triglycerides 84 mg/dL better than 45% better than 49% lower is better · typical of 27

HDL cholesterol 60 mg/dL better than 84% better than 81% higher is better

Blood sugar

Fasting glucose 63 mg/dL lower than almost all lower than almost all the middle is best

Liver & kidneys

ALT (GPT) 25 U/L better than 39% better than 45% lower is better

AST (GOT) 29 U/L better than 23% better than 23% lower is better

γ-GTP (GGT) 25 U/L better than 24% better than 35% lower is better

Blood & inflammation

Hemoglobin 15.1 g/dL mid-range (42nd percentile) mid-range (44th percentile) the middle is best

Red blood cells 512 ×10⁴/µL mid-range (50th percentile) mid-range (56th percentile) the middle is best

Middle 95%Middle halfMedianMe

Source: NHANES (CDC), weighted percentiles for US men within two years of each age; labs from my 2026 workplace health check, the rest from WHOOP, MacroFactor and a tape measure. "Approx.": wearable norms from published Fitbit studies, which NHANES doesn't measure. Not measured yet: 10 more, from ApoB to grip strength.

See the numbers

Marker	Me	vs men of 20	vs men of 26	Median at 20
VO2max (estimated)	43 ml/kg/min	better than 38% of 20-year-old men	better than 50% of men aged 26	45.4
HRV (RMSSD, overnight)	82.3 ms	better than 65% of 20-year-old men	better than 77% of men aged 26	66
Resting heart rate (wearable)	54.6 bpm	better than 83% of 20-year-old men	better than 82% of men aged 26	61.8
Respiratory rate (sleep)	15.6 breaths/min	mid-range for 20-year-old men (percentile 52)	mid-range for men aged 26 (percentile 53)	15.5
Systolic blood pressure	107 mmHg	better than 85% of 20-year-old men	better than 85% of men aged 26	117
Diastolic blood pressure	69 mmHg	better than 36% of 20-year-old men	better than 49% of men aged 26	65.3
Weight	78.8 kg	mid-range for 20-year-old men (percentile 53)	mid-range for men aged 26 (percentile 44)	76.8
BMI	25.4 kg/m²	mid-range for 20-year-old men (percentile 52)	mid-range for men aged 26 (percentile 45)	25.1
Waist circumference	93 cm	better than 38% of 20-year-old men	better than 45% of men aged 26	86.4
LDL cholesterol	95 mg/dL	better than 46% of 20-year-old men	better than 57% of men aged 26	92
Triglycerides	84 mg/dL	better than 45% of 20-year-old men	better than 49% of men aged 26	78
HDL cholesterol	60 mg/dL	better than 84% of 20-year-old men	better than 81% of men aged 26	47
Fasting glucose	63 mg/dL	lower than almost all 20-year-old men	lower than almost all men aged 26	96
ALT (GPT)	25 U/L	better than 39% of 20-year-old men	better than 45% of men aged 26	21
AST (GOT)	29 U/L	better than 23% of 20-year-old men	better than 23% of men aged 26	22
γ-GTP (GGT)	25 U/L	better than 24% of 20-year-old men	better than 35% of men aged 26	17
Hemoglobin	15.1 g/dL	mid-range for 20-year-old men (percentile 42)	mid-range for men aged 26 (percentile 44)	15.3
Red blood cells	512 ×10⁴/µL	mid-range for 20-year-old men (percentile 50)	mid-range for men aged 26 (percentile 56)	512

Against men of 20, I’m ahead of most on four of the twelve markers that have a better direction; against men my age, on five. The liver enzymes AST and γ-GTP stay weak either way, and WHOOP’s VO2max estimate of 43 is typical of a man of 28. A fasting glucose of 63 is lower than almost anyone’s, but glucose is a marker where the middle is best, so it’s flagged to watch rather than counted as a win.

I didn’t build Blueprint only to live longer. My weekdays are teaching, then editing for a client and building things like this in the evening, and all of it runs on the same few hours of sleep, the same meals, and whatever time is left.

Sleep is the biggest lever. In a 2003 study, people allowed six hours in bed a night for two weeks built up cognitive deficits equal to up to two nights without any sleep, and they were largely unaware of it: how sleepy they felt hardly changed. Across five countries, RAND estimated that a worker sleeping under six hours loses about six more working days a year than one sleeping seven to nine, and put Japan’s yearly loss at up to 2.9% of its GDP.

That was me. In the month before the first plan, I slept 5.5 hours a night and was usually asleep at about half past one. Over the 15 nights since, with a bedtime set each week, I’ve averaged 6.3 hours and been asleep by about ten past midnight, and recovery has gone from 49% to 67%.

The month before the first plan, and the 15 nights after

Every night WHOOP recorded, as a bar from falling asleep to waking, dated by the morning. Grey: before the first plan on 29 Sept. Orange: under a plan.

22:0000:0002:0004:0006:0008:0010:00

First plan median 01:28 median 00:10

31 Aug7 Sept15 Sept22 Sept29 Sept6 Oct13 Oct

Asleep a night

5.5 h→6.3 h

Asleep by (median)

01:28→00:10

Recovery

49%→67%

HRV

68 ms→83 ms

Nights of 7 h or more

3 of 29→5 of 15

Source: WHOOP export through 14 Oct. Before: 31 Aug–29 Sept, 29 nights recorded; after: 30 Sept–14 Oct. Off the chart: 19 Sept (13:05–17:40) and 25 Sept (17:24–04:55), both kept in the averages. One person over a few weeks, not a trial.

See the numbers

	Before	After
Asleep a night	5.5 h	6.3 h
Asleep by (median)	01:28	00:10
Recovery	49%	67%
HRV	68 ms	83 ms
Nights of 7 h or more	3 of 29	5 of 15
Sleep consistency (WHOOP)	57%	74%

Scheduling saves the deciding. Each week, every session, meal and errand gets a time and a place once, when the plan is made. A 2006 meta-analysis of 94 tests found that plans saying when and where something will happen raise follow-through, with an average effect of d = 0.65, medium to large. The capacity check matters as much, because paid work goes into the same week as training and food, so an overbooked week shows up before it starts. When W42 was planned, the fit flagged that a 12-hour edit had only 7.2 hours of room before its Friday deadline. I’d already finished that one, but the check runs every week.

Food is planned for time as well as nutrition. W42’s shopping and batch cooking came to one 30-minute shop and two prep sessions, 2 hours 40 minutes for the week, and even in an exam week Thursday and Friday kept more than four free hours each after training. In a survey of 19,803 employees at three large US companies, those with unhealthy diets were 66% more likely to report lost productivity at work, and occasional exercisers 50% more likely. That’s self-reported and a correlation, and the same study found the biggest factor was the work itself, too much of it in too little time, which is the case for the capacity check.

What Blueprint doesn’t measure yet is the work itself. Sleep, recovery and the schedule are tracked every day; my output isn’t. So the link to productivity here rests on the research and on how the weeks have felt, and 15 nights is a short window. Measuring the work is the next piece.