⌑ I · The MechanismHow it actually works.
Sleep is not a single state — it's a structured sequence of four distinct phases cycling every 90 minutes across the night. Each phase serves a different biological function, and losing any one of them produces measurable downstream deficits regardless of total sleep duration.[1]
The four phases
N1 (light sleep, ~5%): the transitional phase between wakefulness and sleep. Muscle tone drops, breathing slows, awareness fades. Duration measured in minutes per cycle.
N2 (light sleep, ~45-55%): the largest single phase. Sleep spindles and K-complexes appear on EEG. Body temperature drops, heart rate slows. Contributes to motor memory consolidation and general recovery.
N3 (deep sleep / slow-wave sleep, ~15-25%): the most physically restorative phase. Growth hormone surges (~70% of daily GH release occurs here). Glymphatic clearance — the brain's waste removal system — runs at 2x waking rate, clearing beta-amyloid, tau, and other neurotoxic byproducts. Concentrated in the first third of the night.[2]
REM (dreaming sleep, ~20-25%): paradoxical state. Brain activity resembles wakefulness; body is atonic (paralyzed). Memory consolidation, emotional regulation, and creative pattern-matching happen here. Concentrated in the last third of the night — cutting sleep short by 90 minutes disproportionately deletes REM.[3]
A person who sleeps 8 hours but wakes 4 times, drinks alcohol, or takes benzodiazepines has "8 hours of sleep" that contains almost no deep sleep or REM. The polysomnogram will show N1 and N2 dominance with fragmented architecture. This is why fitness trackers measuring only duration systematically overstate sleep quality — and why simply "sleeping more" doesn't produce the expected benefits when architecture is destroyed.
⌑ II · The EvidenceWhat the research actually shows.
- All-cause mortality. Cappuccio et al. (2010) meta-analyzed 16 prospective cohort studies (n=1,382,999) and found a U-shaped mortality curve: sleeping less than 6 hours OR more than 9 hours was associated with increased all-cause mortality. The 7-8 hour range was optimal.[4]
- Cardiovascular disease. Same meta-analysis (2011 CVD subset): short sleep < 6 hours associated with 48% increased incidence of coronary heart disease. Mechanism involves sympathetic overactivity, inflammation, and metabolic dysregulation from insufficient sleep.[5]
- Testosterone. Leproult and Van Cauter (2011) demonstrated that 7 nights of 5-hour sleep reduced total testosterone by 10-15% in healthy young men — equivalent to 10-15 years of normal aging in one week. See the testosterone protocol.[6]
- Insulin sensitivity. Buxton et al. (2010) demonstrated one week of restricted sleep (5 hr/night) reduced insulin sensitivity by 20% in healthy adults. Restoration required extended recovery sleep.[7]
- Alzheimer's / cognitive decline. Xie et al. (2013, Science) demonstrated the glymphatic system — the brain's waste clearance — runs 60% more efficiently during sleep than wakefulness. Beta-amyloid clearance in particular is a sleep-dependent process, providing mechanistic support for the observed association between chronic short sleep and elevated dementia risk.[2]
- Immune function. Prather et al. (2015) exposed 164 healthy adults to rhinovirus and tracked infection outcomes. Individuals sleeping less than 6 hours were 4.2x more likely to develop clinical cold vs those sleeping 7+ hours.[8]
- Exercise adaptation. Sleep deprivation blunts every training adaptation studied: reduced protein synthesis response, impaired glycogen replenishment, elevated cortisol, reduced reaction time and power output. The most consistent finding is that sleep is the substrate through which training produces adaptation.[9]
⌑ III · The ProtocolWhat actually works.
Duration target
7-9 hours of actual sleep opportunity. Time in bed must exceed target sleep time (accounting for sleep latency and normal awakenings). A person who needs 8 hours of sleep should be in bed 8.5+ hours.[4]
Consistency
Sleep onset and wake time within ±30 minutes daily, including weekends. Chronic weekend shift (a.k.a. "social jetlag") produces measurable metabolic dysregulation even when total sleep is preserved.[10]
Light environment
Bright light exposure in the first 30 minutes of waking (ideally sunlight, minimum 10,000 lux) anchors the circadian rhythm and drives evening melatonin release ~14-16 hours later. Light avoidance in the 2-3 hours before bed — particularly blue-enriched light > 100 lux — preserves melatonin onset.[11]
Room temperature
Cool bedroom (60-67°F / 15-19°C). Sleep onset requires core body temperature to drop; warm environments delay sleep onset and reduce deep sleep percentage. This is the single most manipulable variable for sleep quality.[12]
Caffeine timing
Half-life 5-6 hours, quarter-life 10-12 hours. Caffeine consumed after ~2 PM measurably reduces deep sleep percentage even when subjective sleep onset feels normal. See the caffeine protocol for the decay curve.
Alcohol
Sedates for the first half of the night, fragments the second half. Chronic use suppresses REM sleep specifically. No dose has been shown to improve sleep architecture; all evidence points the other direction.[13]
⌑ IV · OptimizationBeyond the foundation.
Wind-down routine
60-90 minutes of dim light, low-arousal activity before intended sleep. Reading, low-intensity conversation, warm showers (which paradoxically COOL the body core through peripheral vasodilation). Screens are compatible if brightness is minimized and content is not emotionally activating.
Sleep tracking — informational, not prescriptive
Wearables (Oura, Whoop, Apple Watch, Garmin) estimate sleep stages from heart rate variability and movement. These estimates correlate roughly with polysomnography for duration and light/deep proportion but are unreliable for REM staging. Use tracker data to identify TRENDS (better/worse across weeks), not to make sleep decisions based on individual nightly scores.[14]
Magnesium (glycinate)
The only supplement with reasonable clinical evidence for sleep support in most populations. Glycine (the ligand in glycinate form) has independent sedative-hypnotic effects. Standard dose 200-400 mg 60-90 min before bed. See the magnesium protocol.
Melatonin — dose and timing matter
Endogenous nightly release is approximately 0.3 mg. Supplemental doses of 5-10 mg (the standard OTC format) are dramatically supraphysiological. Emerging evidence supports LOWER doses (0.3-1 mg) taken 4-5 hours before intended sleep to shift circadian phase in jet lag or shift work. Higher doses used as "sleeping pills" are less effective than marketed and can produce next-day grogginess.[15]
Cognitive Behavioral Therapy for Insomnia (CBT-I)
For chronic insomnia, CBT-I has stronger long-term efficacy than any sleep medication. Meta-analyses consistently show sustained improvement 6-12 months post-treatment vs medication response that ends when medication ends. First-line treatment per most sleep medicine guidelines.[16]
⌑ V · When to InvestigateSleep pathology.
- Sleep apnea. The most common under-diagnosed cause of poor sleep architecture. Loud snoring, witnessed apneas, morning headaches, unrefreshing sleep despite adequate duration, or unexplained daytime fatigue → home sleep study. Untreated OSA blunts testosterone, worsens cardiovascular risk, and destroys deep sleep specifically.[17]
- Restless legs syndrome. Uncomfortable leg sensations at rest, relieved by movement. Ferritin below 75 ng/mL is a common driver; check with a complete iron panel.
- Chronic insomnia. Difficulty initiating or maintaining sleep at least 3 nights weekly for 3+ months. CBT-I is the evidence-based first line. Prescription sleep aids should be short-term rescue only.
- Circadian rhythm disorders. Delayed sleep phase, advanced sleep phase, non-24-hour disorder. Distinct from insomnia — the total sleep is normal but timing is misaligned. Chronotherapy + light exposure protocols are the primary intervention.
Every other protocol in the Codex — creatine, testosterone optimization, protein intake, training frequency, longevity compounds — becomes less effective when sleep is compromised. Sleep is not one lever among many. It is the substrate through which every other lever works. Fix this before optimizing anything else.