Joe's monthly health checkpoint for August 2026, written by Claude from the Apple Watch, Withings scale, Tonal, CPAP machine, food log, and daily notes. Joe's own words are quoted where they appear. It describes what happened and does not prescribe. Companion to Health Insights.
August was the best month of the weight cut so far. The scale dropped about four and a half pounds on the monthly average, estimated body fat fell more than a full point, and both changes are well outside the normal month-to-month wobble. At the same time Joe lifted a third more total weight on Tonal than in July and set a new high on Tonal's strength score. Fat is coming off and strength is going up, which is the shape the plan is built around.
The other big event was the CPAP machine, which arrived on August 19 to treat the sleep apnea diagnosed in May. When the mask is on, it works: the machine's own apnea count averaged 2.6 events an hour, comfortably under the treatment target of 5. The problem is hours. The mask was on for about 3.1 hours a night on average, roughly half of each night's sleep, and cleared the 4-hour insurance bar on only 4 of 15 nights. Joe could not feel the difference from the inside; his morning notes read "normal" even on nights the mask was on for twenty minutes.
Everything else was steady. Sleep held at six and three-quarter hours, recovery numbers were normal for him, the weekend protein dip came back once every weekend was logged, and a run of small aches came and went, ending with a right-foot flare that cleared by September 2.
All three data exports were fresh on September 4: Apple Health, Withings, and for the first time the ResMed myAir export from the CPAP machine. Food was logged all 31 days, the first complete month since April. Sleep was recorded all 31 nights, with one night excluded as a watch-off gap (August 12, in Ojai). Two things sit inside the month and color parts of it: a nine-day prednisone taper for the sinuses (August 5 to 13) and a two-week antibiotic course (August 5 to about 18). Where those matter, the section says so.
| Aug 2026 | Jul 2026 | 12-month avg | |
|---|---|---|---|
| Weight (lb) | 173.1 | 177.5 | 187.4 |
| Body fat %, estimated | 37.0 | 38.2 | 40.2 |
| Fat mass (lb), estimated | 64 | 68 | — |
| Visceral fat (Withings scale) | 3.78 | 3.97 | 4.36 |
The body fat and fat mass numbers are the Withings scale reading plus a fixed correction to the DEXA scan Joe had on June 11. A DEXA scan is the gold-standard body composition measurement; the home scale reads about 15 points low, so the raw scale number is not shown. The corrected figure is an estimate, and the trend is more trustworthy than the exact level.
Weight went from 175.7 on August 3 to 170.7 on August 31, and hit 168.1 on September 4, the lowest reading of the year. That is a little over a pound a week. The prednisone in early August holds water and pushed the scale around for about ten days; taking those readings out changes the monthly slope by almost nothing.
Against the June DEXA baseline (181.9 lb, 38.4% body fat, 69.8 lb of fat, 106.5 lb of lean mass), the estimate now sits at about 170 lb, 36.4% body fat, and 62 lb of fat, so roughly 8 lb of fat gone in twelve weeks. Lean mass cannot be measured between scans; Tonal's strength score is the stand-in, and it rose from 514 to 519 during the month. The next DEXA is due around mid-October and is not booked yet. Joe's read on August 31 was that booking it is not urgent.
The long-term target, set after the June scan, is body fat in the low twenties and a weight around 150 to 160 lb, with lean mass going up rather than down.
Joe trained on most days: 10 strength sessions (9 on Tonal and one hotel session), 7 boxing sessions, 2 short runs, 3 Peloton rides, a five-minute jump-rope test of the Achilles, and 22 logged walks covering 29 miles. July had 12 strength, 7 boxing, 1 run, and 18 walks. Frequency held through a trip week and a medication week.
The Tonal sessions got much bigger. Nine sessions moved 84,200 lb across 228 working sets, against 63,000 lb across 158 sets in July, so a third more total work in fewer sessions. The two biggest sessions of the summer were August 17 and August 31, and the strength score ticked up on the second one. Tonal's score splits into upper (621 to 630), core (567 to 572), and lower (353 to 356). Lower body has sat in a 353 to 358 band all year while the other two climb, which is the long-standing gap in the program.
The balance of the work shifted toward pulling and the lower body compared with June: 44% pushing, 31% pulling, 25% lower body. In July the deadlift and hip-thrust family had disappeared entirely; in August they came back (Romanian deadlift, conventional deadlift, sumo deadlift, plus two new hip and lunge movements). Squats and rows progressed, bench press and its variants moved up a little, and three lifts slipped (triceps extension, chin-up, biceps curl). Tonal's strength estimates are on its own cable-load scale, so treat them as trend, not barbell numbers.
Boxing was the same engine as June and July. Seven sessions, most of them 70 to 120 minutes, averaging 140 beats per minute apart from one deliberately light session on August 25 when the right arch started hurting mid-workout. The two longest sessions (August 7 and 11) burned around 1,170 to 1,190 calories each. The goal of doing the same work at a lower heart rate needs more months to read.
The aches this month lined up with the dense stretches. August 3 to 7 was five straight training days ending with boxing and a leg session back to back, and a hip, a calf cramp, and lower-back tightness showed up on days two, four, and five of it, then faded. A big leg day on August 16 produced two days of heavy soreness. The end-of-month volume records came while the foot was sore and did not make it worse.
Protein averaged 136 g a day (July: 135), and calories averaged 1,724 (July: 1,776). Four days reached the 160 g target, seven reached 150 g, and five fell under 120 g.
The pattern that keeps showing up is weekends. Weekday protein averaged 143 g; weekend protein averaged 121 g. Four of the five low-protein days were weekend days, including both days of a trip on August 8 and 9. Weekend calories were also lower than weekdays (1,656 vs 1,756), so the weekend problem is missing protein rather than extra food. The one clean weekend was August 15 and 16. In July it had looked like the weekend gap had closed, but that read came from only five logged weekend days. With all ten August weekend days logged, the gap is back.
The prednisone stretch pushed appetite up by about 150 calories a day, which the daily log predicted at the time, and intake fell back once the taper ended. The second half of the month averaged 1,631 calories against 1,823 in the first half, which answers July's question about whether 1,900 a day was becoming the new normal. It was not.
Intake of about 1,700 calories against an estimated maintenance of about 2,250 predicts roughly a pound a week of loss, and the scale delivered 1.1 to 1.25. The food log and the scale have agreed for three months now. The Apple Watch's own burn estimate for the month was 2,800 calories a day, about 24% too high, and it is no longer used for planning.
Two smaller notes. Hard boxing mornings produce both extremes: two of the 160-plus days were boxing days with a full lunch and dinner, and the 92 g, 1,205-calorie day on August 4 was a boxing day where the post-session hunger never showed and lunch got skipped. And on August 26 Joe named a pattern of his own: "feel like I get constipated when i overload on the pure protein drink or bars," logged the morning of a 60 g single-sitting protein shake.
Joe slept an average of 6.7 hours a night on the 30 valid nights, with a median of 6.8 (July: 6.6 and 6.6). Night-to-night variation tightened slightly. Deep sleep held at 43 minutes and REM at 102 minutes, so no change in sleep quality on the watch. Two genuinely short nights stayed in the numbers: August 11, when he woke at 3 a.m. and could not get back to sleep, and August 28, a 4:50 a.m. boxing morning. Nights before boxing averaged 5.5 hours against 7.1 otherwise, the third month with that shape.
On the morning check, 23 of 28 mornings were "normal," four were "not good," and one was "really good." The four rough mornings each had a cause: a calf cramp on day two of the steroid, worry about an early airport run, the 3 a.m. wake, and a night of fighting the CPAP mask. The morning rating tracks how long he slept and does not track breathing disturbances, which matches what the sleep study showed in May: Joe cannot feel the apnea.
The Apple Watch's breathing-disturbance count kept drifting down (only 10% of nights read elevated, from 28% in July and 71% in May), but the drop happened in the week before the CPAP machine arrived, so it cannot be credited to the machine. From here, the machine's own numbers are the airway measurement and the watch is background.
The machine is a ResMed AirSense with an AirFit P30i mask, which is a nasal-pillows mask (small cushions at the nostrils, nothing over the mouth). It arrived August 19. The table shows every night the export covers.
| Night of | Hours on | Apnea events/hr (AHI) | Mask leak (95th pct, L/min) | Mask on/off | Joe's morning read |
|---|---|---|---|---|---|
| Aug 19 | 3.9 | 2.0 | 31 | 2 | "went well" |
| Aug 20 | 5.3 | 5.6 | 35 | 2 | normal |
| Aug 21 | 4.1 | 2.2 | 31 | 2 | "trouble with my cpap mask that was waking me up" |
| Aug 22 | 2.5 | 0.8 | 36 | 2 | normal |
| Aug 23 | 3.2 | 2.5 | 22 | 2 | normal |
| Aug 24 | 3.2 | 3.4 | 36 | 3 | normal |
| Aug 25 | 6.4 | 2.6 | 37 | 5 | "woke up a bunch of times messing with the cpap mask" |
| Aug 26 | 0.0 | — | — | 0 | "forgot to wear cpap mask," sleep normal |
| Aug 27 | 2.2 | 1.3 | 29 | 3 | normal |
| Aug 28 | 0.9 | 1.1 | 20 | 2 | normal |
| Aug 29 | 2.7 | 5.5 | 35 | 2 | normal |
| Aug 30 | 0.9 | 3.2 | 34 | 4 | normal |
| Aug 31 | 0.3 | 0.0 | 16 | 3 | normal |
| Sep 1 | 4.5 | 2.6 | 28 | 2 | normal |
| Sep 2 | 3.3 | 3.6 | 26 | 6 | normal |
Three things stand out.
The pressure works. AHI is the number of apnea events per hour while the mask is on; under 5 counts as treated. Joe averaged 2.6, and 12 of 14 nights were under 5.
The hours fall short. The machine ran an average of 3.1 hours a night, against about 6.7 hours of total sleep on the same nights, so it was on for roughly half of each night. Four nights cleared four hours. The last five nights of August ran between twenty minutes and two and a half hours. After a consultant visit on September 1 the next three nights ran 4.5, 3.3, and 3.2 hours. The insurance rule behind the $653 payment is four hours a night on 21 of 30 consecutive nights. The myAir export's own compliance card shows 4 of 21 for the window that runs August 19 to September 17, and with 14 nights left in that window the most it can reach is 18. Whether a later 30-day window counts is a question for the clinic.
The mask is fighting him. Leak at the 95th percentile averaged 30 liters a minute (ResMed treats 24 as high), and the mask came off and went back on two to six times a night. Joe went into the fitting expecting a full-face mask because he cannot breathe comfortably through his nose, and the August 19 CT scan confirmed why: a deviated septum and blocked sinus openings on both sides. A mask that only covers the nose has to win against that nose every night, and short hours, high leak, and repeated removals are what losing that fight looks like. Mask type is the first thing to raise with the clinic, ahead of pressure settings. What was decided at the September 1 visit is not in the log.
One more observation for anyone reading the watch data alongside the machine: on CPAP nights, more hours on the machine went with a higher Apple breathing-disturbance score, the opposite of what you would expect if the watch were measuring the airway. The likeliest explanation is that the watch reads pressurized breathing as disturbance. On CPAP nights, ignore the watch's breathing number.
The month opened with the tail of a run of small, rotating aches (left knee, right forefoot, left hip, left calf, right lower back) that appeared one after another between July 27 and August 7, none of them in the same place twice, each settling within a day or two. They landed on heavy training days during a calorie deficit and read as accumulated load, and the run closed on August 9.
The right foot was different. Arch pain arrived mid-boxing on August 25, cleared over three rest days, came back at the end of the day on August 28 after a hard boxing morning, and then got worse through two days with no training at all, spreading to the top of the foot and the ankle by August 30 and bothering him on a Peloton ride that afternoon. Pain that worsens under rest does not read as training load. It improved on August 31, was gone on September 2, and held through hill sprints on September 3 and hard boxing on September 4. The unexamined variable is shoes: Joe had concluded on August 17 that his everyday pair was worn out, replaced it on August 24 with a New Balance Numeric 933 (a skate shoe, built flat), and the arch pain started the next day. Nothing has been raised with a doctor.
The walking-asymmetry number the watch produces stayed inside its normal range and ran higher on boxing days, as it always does. Nothing to flag there.
Evening ratings averaged energy 7.2, soreness 6.2, and focus 7.0 out of 10, essentially identical to July. The three low-energy days were August 11 (the 3 a.m. wake, hard boxing, and a two-hour drive), August 14 (4 a.m. boxing on the sixth straight training day), and August 27, when a strange pain on swallowing had been "bothering me all day." The two highest soreness days were August 18 (two days after the big leg session) and August 25 (the arch).
A correlation flagged in July, between a higher heart-rate variability reading and better evening energy, did not show up again in August. One month of signal and one month of nothing means it is not a pattern yet. Heart-rate variability, or HRV, is the beat-to-beat variation in heart rhythm; higher is better and means the body is recovered rather than stressed. Joe's August average of 33 ms sat near the top of his own yearly range.
The sinus workup moved. Dr. Mazza's antibiotic and prednisone course ran August 5 to 18, the CT scan on August 19 found chronic sinusitis with blocked sinus openings on both sides and a deviated septum (and no polyp, despite the referral history), and the September 2 follow-up settled on surgery to clear the blockage. There is no surgery date, and it will not happen before the October checkpoint. The prednisone did what steroids do for nine days (disrupted sleep, a headache, more appetite, a bouncing scale) and left nothing behind.
Peptides (BPC-157/TB-500 and GHK-Cu) were taken every weekday except August 12, when they were left at home on a trip. The GHK-Cu, taken partly for psoriasis, got its first real before-and-after: Joe noted on August 21 that "in the time I was off I was definitely seeing psoriasis in my beard and having less flair while I am on it." The improvement while back on overlaps with the prednisone, so September skin is the clean read.
The weekly Zepbound shot (10 mg, a GLP-1 medication for weight loss) is logged on July 31, August 15, August 21, and August 30. Nothing is logged between July 31 and August 15, a fifteen-day gap, so either a shot went unrecorded around August 7 or 8 or a week was skipped. It matters for two of the questions below.
A pain on swallowing appeared on the evening of August 26, localized to where the left side of the neck meets the chest, tender to the touch, triggered only by swallowing including water, with no lump, no ache at rest, and no change with breathing or exertion through a hard boxing session. It improved from day three and was gone by August 29, the course a muscle strain would take. Joe's call to watch it through the weekend rather than seek same-day care was consistent with the data. If it recurs, it goes to a doctor.
Joe also started an experiment on August 13, dropping the bedroom temperature automatically around 2 a.m. to test whether it stops a recurring 2 a.m. wake. The nightly totals cannot score it; it needs the raw sleep segments, and it overlaps with the CPAP start.
Four questions have been running through these reviews. Here is where each stands.
How many calories does Joe actually burn? Settled. For three months the food log and the scale have implied a maintenance level of about 2,250 calories a day (August: 2,270 to 2,350 depending on method), while the Apple Watch claims about 2,800 to 2,900. The Watch is about 24% high for Joe and is retired from planning. Every target now runs off 2,250. Joe confirmed this as a settled finding on September 4.
Do the peptides soften the recovery hit from the weekly Zepbound shot? Not answered, and August showed why it is hard. Earlier data suggested resting heart rate and HRV looked better on the day after a shot when peptides were on board. In August that difference turned out to be an artifact: the day after a shot is always a weekend, and August's weekends simply had better recovery numbers than June's, peptides or not. Within each period, there was no day-after cost to speak of. Peptides also did not prevent the rotating aches or the foot flare. Joe was not ready to close this one; the next scheduled off-cycle, around mid-October, is the cleaner test.
Will treating the apnea lift testosterone and daytime energy? Treatment has started but is not yet at a level that can answer it. The bar written before the data was good adherence (four-plus hours on at least 70% of nights) plus a treated AHI under 5. The AHI half is met; the hours half is at 27%. A testosterone draw in October or November on these hours would read as an untreated result, and a flat number would prove nothing either way. That branch was anticipated: if treatment starts late or thin, the right move is to wait, not to default to testosterone therapy.
Did moving the Zepbound shot to Friday deepen the weekend protein dip? Still untested, because every shot has landed on a Friday, Saturday, or Sunday, so there has been no midweek shot to compare. August did lift the question back off the shelf: July's weekends had looked fine on five logged days, but August's full ten show the 22 g gap. A thin hint in the direction of the mechanism: the two weekends that fell one and two days after a shot averaged around 110 g of protein, while the weekend where the shot came Saturday night and the weekend eight days past a shot averaged 138 to 148 g. Two against two, so a hint and no more.
The CPAP hours are the month's open question. Three things need answers from the clinic: whether the nasal-pillows mask is the right interface against a nose the CT shows is blocked, whether a swap can happen before the no-cost window closes around September 9, and whether the insurer allows a later 30-day compliance window. On these hours, the October testosterone draw probably moves to November, per the plan already written in Health.
Was there a Zepbound shot between July 31 and August 15? None is logged.
The right foot has been clean since September 2, with the shoes untested as a cause and no doctor involved. Whether the skate shoe stays the everyday shoe is Joe's call.
Not yet done: the September peptide labs (ordered July 20, not drawn), the extended testosterone/PSA lab order, and the October DEXA and urology booking, the last two not urgent per Joe.
The 2 a.m. wake experiment can be scored from the raw sleep data if it is wanted; otherwise it waits until CPAP hours settle.
Written for Joe and the people he shares it with. Numbers are from consumer devices and a home scale except where a clinical test is named.