Show Notes
Colin Lawlor called in from Berlin at nine at night, three days after leaving San Diego, worn down by a Heathrow layover. He runs a sleep company. He was tired, and he knew exactly why, down to the hormone.
Funding stage: Series A Plus. Sleep.ai, formerly SleepScore Labs, spun out of ResMed in 2016 with backing from Dr. Mehmet Oz and Pegasus Capital, has since built a reimbursement deal reaching 25 million insured members in Germany and normalized data across hundreds of wearables, the kind of institutional scale and distribution that reads as late-stage growth rather than an early raise.
That is the whole thesis of this episode. Sleep is not a soft topic. It is a measurement problem, a data problem, and now a business model, and Colin has spent a decade turning it into all three. Sleep.ai started life in 2016 as SleepScore Labs, spun out of ResMed with backing from Dr. Mehmet Oz and Pegasus Capital. It rebranded in 2025. Along the way it collected something close to a billion hours of real-world sleep data, ran more than 250 intervention studies, and became the first sleep technology in the world to be reimbursed by health insurers.
The founder relevance arrives fast. Colin describes landing at Heathrow exhausted and craving sugar and fat, and explains that this was not weakness. The hormones that govern appetite are produced during deep sleep. Short the sleep, short the hormones, and your body starts negotiating with you at the vending machine. Every founder who has ever eaten a gas station dinner at 11pm after a bad night has felt this mechanism without naming it.
Then he goes after the thing every one of your listeners is wearing. Put four wearables on one wrist and you will get four different answers about the same night. Sleep.ai's answer is to normalize all of them against polysomnography, the hospital sleep lab standard where a technician scores every thirty seconds of your night. That is expensive, slow, and almost nobody else has funded it at scale, which is precisely why it became a moat.
The line that will stick with you is the one about action. There is a decade of aggregated wearable data on this planet and the population is not sleeping better. Data without intervention is a graph you feel bad about. Sleep.ai sells the intervention layer to companies that already own the customer relationship, priced per monthly active user, integrated in about a day.
Colin also drops something most founders never see coming. Germany passed a prevention law requiring insurers to reimburse proven prevention tools, and Sleep.ai walked through that door first with Dein Schlaf. Regulation as a distribution channel. Not a phrase you hear at demo day.
There is a harder passage in here too, about his mother's cancer treatment and the stress he was carrying in the wrong direction. And a lovely detour where Ryan brings up Turiya, the fourth state of consciousness in Vedic philosophy, and Colin, running a company built entirely on quantifying sleep stages, says he has no problem believing the brain is doing something in all of them that we have not measured yet.
Every founder in your audience is making a trade they have not priced. This episode prices it.
Named Frameworks
The Consistency-First Sleep Stack
Colin's personal protocol, ranked by what he actually believes moves the needle.
- →Consistent bedtime and wake time nearly every day, the single highest-leverage change available to almost anyone.
- →Black coffee, three or four cups, all before noon, because caffeine after midday wrecks his night.
- →Running three or four times a week, a few miles, framed as heart maintenance rather than performance.
- →Gluten-free by medical necessity (celiac), with calcium supplementation to offset malabsorption.
- →Supplement stack chosen on evidence per item: turmeric, vitamin D3 with K2, fish oil, and creatine, the last one specifically for brain effects.
- →Melatonin used as a clock-shifting tool for travel, explicitly not as a nightly habit.
The Light Sandwich (Jet Lag Protocol)
Colin's own travel routine for resetting the clock across time zones.
- →Blue light filtering glasses on from evening, worn through the airport, the jet bridge, the cabin, the in-flight movie, and the brutally lit lavatory.
- →Airports and planes are an almost perfect design for circadian sabotage, bright light at every point where the body is trying to shut down.
- →On landing, glasses off and straight into as much natural daylight as possible, ideally ten to fifteen minutes.
- →Caffeine timing and short-term melatonin layered on top to nudge the clock.
The Gold Standard Normalization Layer
How Sleep.ai turns four different wearable answers into one comparable number.
- →Anchor everything to polysomnography: sixteen to twenty-six sensors, human-scored in thirty-second epochs.
- →Fund thousands of lab nights across healthy people, people with disorders, people with chronic disease, varied BMI, gender, and ethnicity, run across multiple labs.
- →Once the core measurement is validated, ingest data from any consumer wearable and score it against the same standard.
- →Result: an Apple Watch user, an Oura user, and someone wearing nothing at all can be compared on one scale.
All Data, No Action
Colin's diagnosis of the entire wearables category.
- →A decade of aggregated sleep data has not improved population sleep.
- →Bad data makes the problem worse, because you cannot manage what you cannot measure.
- →Good data is necessary and still insufficient.
- →The value sits in personalized intervention: what works for you specifically, tested and confirmed.
- →Test the interventions themselves at scale, which is why the company has run 250-plus studies across advice, CBT, supplements, pillows, mattresses, blue light filtering, and air purification.
Sleep as the Diagnostic Window
The underused half of the sleep story: an early warning system, not just a performance lever.
- →High-quality sleep drives performance, disease risk, and longevity, the well-known half.
- →The underused half: sleep is the cleanest eight-hour window into underlying health, before symptoms surface.
- →Company precedent, per Colin: early work predicting acute hospitalizations in congestive heart failure and acute events in COPD.
- →Independent precedent: SleepFM at Stanford, one night of lab data forecasting risk across 130 conditions.
Two-Lane Monetization
How a sleep data company actually makes money.
- →Lane one: per monthly active user pricing on the SDK and API, integrable in about a day, sold to health and wellness apps.
- →Lane two: R&D and validation services, where a company building a physical or digital sleep product pays for data access and full studies to prove it works.
- →The second lane feeds the first, because every study makes the recommendation engine sharper.
Founder Experiment
Run the Fourteen-Day Wake Anchor. Pick one wake time and hold it every single day for two weeks, including weekends, including the day after a bad night. Do not touch your bedtime, do not add a supplement, do not buy anything. Change one variable. Alongside it, log two numbers each evening on a scale of one to ten: how good your decisions felt today, and how hard your food cravings hit after 4pm. Two numbers, no app required. Add one optional constraint if you want to stress-test Colin's strongest claim: no caffeine after noon. At day fourteen, look at the craving column first. Colin's argument is that appetite is downstream of deep sleep, which means the fastest visible signal that your sleep is repairing is not how alert you feel, it is that the 9pm kitchen raid stops happening. If that column falls and your decision column rises, you have just replicated the mechanism on a sample size of one, for free, and you now know it applies to you specifically rather than to a population average. Founders who travel: run the Light Sandwich on your next trip and log the same two numbers on landing day and the day after.
Glossary
Polysomnography (PSG)
The overnight hospital sleep study. Sixteen to twenty-six sensors, all-night signal capture, then a certified human scores every thirty seconds of the night for wake, sleep, and sleep stage. The reference standard everything else is measured against.
Slow-wave sleep (SWS)
Also called deep sleep. The stage where respiration slows and deepens, body movement drops, and the brain runs its heaviest maintenance. Associated with glymphatic clearance of metabolic waste including proteins implicated in Alzheimer's.
REM sleep
The dreaming stage. The brain is highly active while the body is effectively paralyzed, a safety mechanism that stops you from acting out dreams. Breathing becomes irregular as the emotional content of the dream shifts.
Hypnic jerk
The falling sensation and sudden twitch as you drop into early sleep.
Chronotype
Your biological bias toward mornings or evenings, morning bird, night owl, or somewhere between. Colin's argument is that prescribing 6am workouts to a night owl is prescribing failure.
Circadian rhythm
The roughly 24-hour internal clock, entrained primarily by light exposure, that governs sleep timing, hormone release, and metabolism.
Chronotherapy
Timing a medical treatment to the patient's circadian rhythm. Evidence is genuinely mixed.
HRV (heart rate variability)
Beat-to-beat variation in heart rate, used as one of several proxies for sleep stage and recovery when EEG is not available.
EEG (electroencephalography)
Direct brain-wave measurement. The purest way to stage sleep, and the one nobody wants to wear to bed.
Biomotion and sonar sensing
Non-contact measurement using a device's speaker and microphone to detect breathing and movement across a room, no wearable required.
Sleep apnea
Repeated breathing interruption during sleep. Common, dangerous, and treatable.
Chronic insomnia
Persistent difficulty falling or staying asleep. Colin cites 800 million people worldwide, mostly untreated, though global prevalence estimates vary widely by definition.
Restless leg syndrome (RLS)
An urge to move the legs, worse at rest and at night. Colin has it and puts prevalence around 8%, consistent with commonly cited ranges.
Narcolepsy and cataplexy
A neurological disorder of sleep-wake regulation, and the sudden muscle weakness that can accompany it.
CBT-I
Cognitive behavioural therapy for insomnia. The first-line non-drug treatment, and one of the intervention categories Sleep.ai has studied.
Foundation model
A large model pretrained on raw signal data that can then be adapted to many downstream tasks. SleepFM is one, trained on polysomnography.
MAU pricing
Monthly active user pricing. Sleep.ai's core API revenue model.
SDK
Software development kit. The package a partner drops into their app to connect to Sleep.ai's APIs.
Q&A: What Founders Ask After This Episode
What is Sleep.ai and what does the company do?
Sleep.ai is a sleep intelligence platform that sells sleep measurement and personalized sleep improvement to other companies through APIs and SDKs. It was founded in 2016 as SleepScore Labs, spun out of ResMed, and rebranded to Sleep.ai in 2025. It is led by founder and CEO Colin Lawlor and headquartered in Carlsbad, California, with offices in Dublin and Berlin.
Why do different sleep wearables give different results for the same night?
Because most consumer wearables have not funded thousands of nights of polysomnography data collection to train and validate their algorithms against the clinical gold standard. The sensors themselves are usually fine; many consumer devices simply overestimate total sleep time. Sleep.ai normalizes data from hundreds of wearables against the same standard so results are comparable.
How much does losing deep sleep increase dementia risk?
A Framingham Heart Study analysis published in JAMA Neurology found that each percentage-point decrease in slow-wave sleep per year in adults over 60 was associated with a 27% increase in dementia risk and a 32% increase in Alzheimer's disease risk, across 346 participants followed for 17 years.
Can AI predict disease from a single night of sleep?
Stanford researchers published SleepFM in Nature Medicine in January 2026. Trained on close to 600,000 hours of polysomnography from roughly 65,000 people and linked to decades of electronic health records, it identified 130 conditions predictable from one night of sleep data, with the strongest performance in cancers, pregnancy complications, circulatory disease, and mental health disorders.
How does a sleep API company make money?
Sleep.ai runs two revenue lines: a per monthly active user SaaS fee for partners embedding its SDK and APIs into health and wellness apps, and a research and development service where companies building sleep products pay for data access and full validation studies.
Is sleep coaching covered by health insurance anywhere?
Yes, in Germany. Sleep.ai's Dein Schlaf launched in 2024 as the world's first reimbursed sleep improvement app under Germany's preventative health framework. As of March 2026, participating statutory insurers including Barmer, AOK Niedersachsen, IKK Brandenburg und Berlin, and Bergische Krankenkasse reach about 25 million insured members.
What is the best way to fix jet lag according to a sleep company CEO?
Colin Lawlor wears blue light filtering glasses from the evening onward through the entire flight, including trips to the aircraft lavatory, then removes them on arrival and seeks ten to fifteen minutes of natural daylight to signal wake time. He layers caffeine timing and short-term melatonin on top, using melatonin only to shift the clock rather than as a nightly habit.
How long should a nap be?
Colin Lawlor cites research supporting either roughly 20 minutes or a full sleep cycle of about 90 minutes, taken early enough in the day. Naps running two to three hours tend to interfere with the following night.
Why should high schools start later?
Teenagers have a delayed circadian phase and are biologically inclined to sleep and wake later. Colin Lawlor has campaigned with others in the sleep community for start times no earlier than 8:30am.
What is Sleep Sense by Sleep.ai?
A technology that identifies sleep and wake boundaries from ordinary smartphone signals, extending sleep measurement to nights when a user's wearable is not worn, or to users who own no wearable at all.
I sleep five hours and feel fine. Am I actually fine?
Almost certainly not. A very small number of people carry a short sleep gene and appear to function well on little sleep with no obvious short-term cost, but we have not observed enough of them for long enough to know the long-term picture. Colin doesn't prescribe seven hours as a universal number, because the evidence doesn't support one. His actual advice is unglamorous and achievable: claw back minutes. If you're at six hours, get to six hours and two.
URLs Mentioned in the Episode
- Sleep.aihttps://www.sleep.ai
- Sleep APIhttps://www.sleep.ai/sleep-api/
- Ryan Estes on LinkedInhttps://www.linkedin.com/in/estesryan/
- AI for Foundershttps://aiforfounders.co
- Inbox Alchemyhttps://inboxalchemy.co/
- AI Native Studenthttps://ainativestudent.com/
- Momentoushttps://crrnt.app/MOME/8RDrnXDd
- Taelorhttps://taelor.style/
- Get Howdihttps://gethowdi.com/
