The research behind every line of code.

Mazzetti et al., 2000 — Medicine & Science in Sports & Exercise

n = 20 moderately trained men, 12-week resistance-training protocol

Finding: Directly supervised resistance training produced significantly greater increases in maximal squat and bench press strength than the same program performed unsupervised. Same exercises, same weeks — the difference was a coach watching and adjusting.

Why it matters for LIM: Most people can't afford a coach in the room every session. LIM's thesis: an AI coach that reviews every logged session and adjusts the next one recovers a meaningful share of that supervision effect at app pricing.

Burke, Wang & Sevick, 2011 — Journal of the American Dietetic Association (systematic review)

22 studies of self-monitoring in weight loss

Finding: The review found a consistent, significant association between self-monitoring (of diet, exercise, or weight) and weight-loss outcomes across the literature. People who track, lose; people who stop tracking, stall.

Why it matters for LIM: A coach who prompts you to log, notices when you go quiet, and follows up is an adherence tool, not a luxury. That's why check-ins are built into the daily loop.

Javaloyes et al., 2019 — International Journal of Sports Physiology and Performance

n = 17 well-trained cyclists, 8-week protocol

Finding: Cyclists whose weekly training was prescribed from daily HRV readings improved 40-minute time-trial performance and peak power, while the group on a predefined traditional plan showed no significant improvement. Same athletes' level, same period — the variable was whether the program listened to the body.

Why it matters for LIM: LIM reads your HRV trend from Apple Health and pulls load back when your recovery data says to, before you walk into the gym.

Kiviniemi et al., 2007 — European Journal of Applied Physiology

n = 26 healthy men, 4-week endurance protocol

Finding: Training guided daily by individual HRV produced greater improvements in maximal running velocity and comparable-or-better VO2max changes versus a predefined training program.

Why it matters for LIM: Autoregulation isn't a biohacker fad — it's been in the literature for nearly two decades. Most apps still ship fixed templates because reading the body daily is operationally hard. That's the part we automated.

Kecklund & Axelsson, 2016 — BMJ (review)

Narrative review of the shift-work and insufficient-sleep literature

Finding: Shift work and the short, mistimed sleep that comes with it are associated with elevated risk of obesity, type 2 diabetes, and cardiovascular disease, with circadian disruption of metabolism as a central mechanism.

Why it matters for LIM: Shift workers need MORE program adaptation, not less. Fixed-template apps and a 50-minute Saturday trainer session both fail this population on first principles.

Vyas et al., 2012 — BMJ (meta-analysis)

34 studies, over 2 million participants

Finding: Shift work was associated with a 23% higher risk of myocardial infarction and a 5% higher risk of ischaemic stroke versus day workers.

Why it matters for LIM: This is why LIM was built. Jake worked hospital security night rotations and watched this play out in real bodies — including his own, at 308 lbs. The wedge is real.

Mageau & Vallerand, 2003 — Journal of Sports Sciences

Motivational model of the coach-athlete relationship (review)

Finding: The quality of the coach-athlete relationship — a coach who knows the athlete's context and supports their autonomy — is a key driver of sustained motivation, beyond the technical content of the training itself.

Why it matters for LIM: Our thesis, built on that: a coach who actually remembers your injury history, your schedule, and what you said last Tuesday keeps you in the game longer than a marginally better spreadsheet. A traditional trainer holds that context for a handful of clients; LIM holds it for every client, permanently. We haven't seen an RCT on AI coach memory yet — when one lands, HERMES will find it.

Helander et al., 2014 — Journal of Medical Internet Research

Usage data from ~189,000 downloads of a photo-based food-logging app

Finding: The overwhelming majority of users abandoned dietary self-monitoring almost immediately — only a small fraction remained active users, and sustained logging was rare. Friction, not intent, is where logging dies.

Why it matters for LIM: Accuracy doesn't matter if you stop logging. LIM's barcode + photo + voice logging exists because of this failure mode — make it 5 seconds or it dies.