MULIO · Methodology
1,105 sources. 8 operators. Two agents.
The page for the skeptical reader, the senior engineer, and the AI search crawler. Every claim MULIO makes is sourced here.
§1 · The corpus
1,105 sources. Curated, not crawled.
Every source is hand-reviewed before it enters the index. Sources are versioned and re-validated quarterly. The corpus is the upper bound of what the Strategist can cite — if a claim cannot be sourced, the claim does not appear.
- Peer-reviewed journals
- NIH PubMed entries
- Sleep
- Nature Human Behaviour
- JAMA Internal Medicine
- Cell Metabolism
- Expert podcasts
- Huberman Lab
- The Drive (Peter Attia)
- Found My Fitness
- Lex Fridman
- Naval
- Government / institutional health
- NIH
- CDC
- NICE
- WHO
- Books (primary texts)
- Why We Sleep
- Outlive
- Deep Work
- Driven to Distraction
§2 · The operators
Eight operators. Full credits.
MULIO doesn’t invent protocols. It runs the ones that already work — drawn from operators with decades of clinical, academic, and lived domain experience. Each entry below names what the operator is best known for and which specific protocols MULIO pulls from their work.
Andrew Huberman
Circadian, dopamine, focus
Protocol — Morning sun exposure within 30 minutes of waking, 5–10 minutes.
Stanford neurobiology professor. Host of Huberman Lab podcast (250+ episodes). MULIO uses his protocols on circadian anchoring, dopamine hygiene, and pre-sleep wind-down — including the full set of light-exposure timing rules that anchor every This Week schedule.
Peter Attia
Longevity, cardio, metabolism
Protocol — Zone 2 cardio, 180 minutes per week, conversational pace.
Physician, longevity researcher, author of Outlive. Host of The Drive podcast. MULIO uses his Zone 2 cardio prescription, his VO2 max progression structure, his last-meal-before-sleep cutoff, and his framework on metabolic flexibility as inputs to the weekly diagnostic.
Andy Galpin
Training, recovery, performance
Protocol — 4-week mesocycles. Heavy compound lifts twice per week. Deload every fifth week.
Professor of human performance at California State University, Fullerton. MULIO uses his mesocycle structure, his deload cadence, his sleep-and-strength interaction model, and his protocols around protein dosing and pre-workout caffeine timing.
Matthew Walker
Sleep
Protocol — 8-hour sleep window. Same wake time every day, including weekends.
Professor of neuroscience at UC Berkeley. Author of Why We Sleep. Founder of the Center for Human Sleep Science. MULIO uses his sleep-window prescription, his caffeine half-life math, his anti-jetlag protocols, and his case for consolidating wake time as the strongest single circadian lever.
Cal Newport
Deep work, attention
Protocol — Deep-work blocks of 90 minutes minimum. Phone in another room.
Professor of computer science at Georgetown. Author of Deep Work and Digital Minimalism. MULIO uses his block-based scheduling, his phone-discipline prescription, his weekly review structure, and his framing of attention as a finite, allocatable resource.
Russell Barkley
ADHD, executive function
Protocol — Externalize working memory. Visible cues at the point of action.
Clinical psychologist. Foremost ADHD researcher of the modern era. Author of Taking Charge of Adult ADHD. MULIO is shaped by his framing of ADHD as an executive-function disorder rather than an attention deficit — the reason notifications fire AT each block (not before, not after) is his.
Susanna Søberg
Cold exposure, contrast therapy
Protocol — 11 minutes per week of deliberate cold, distributed across multiple sessions.
Danish researcher in metabolism and thermoregulation. MULIO uses her 11-minutes-per-week prescription, her habituation curve for cold tolerance, and her separation-window rules for cold exposure relative to strength training and sleep.
Robert Sapolsky
Stress physiology
Protocol — Stress is a tool. Recovery is the protocol. Avoid sustained 24-hour cortisol elevations.
Professor of biology and neurology at Stanford. Author of Why Zebras Don’t Get Ulcers and Behave. MULIO uses his framing of acute vs chronic stress, his cortisol-curve dynamics, and his rule that recovery is a protocol — not a default state.
§3 · The pipeline
Two agents. Two passes. Two checks.
MULIO does not run a single LLM call. The split is structural — it forces the system to argue with itself before it answers you. The Diagnostician describes. The Strategist prescribes. Neither role bleeds into the other.
- Agent 01 · Reads
Diagnostician
Receives the week’s daily entries, the Apple Health snapshot if granted, and the prior-week diagnostic summary. Surfaces three to five named patterns. Writes a one-page diagnosis. Does NOT prescribe — its only job is to describe what happened.
- Agent 02 · Writes
Strategist
Receives the diagnosis. Pulls protocols from the 1,105-source corpus matched to the diagnosis. Writes a one-page strategy for the week ahead. Hands the strategy to the Playbook Composer to build the next This Week. Does NOT see the raw daily entries — only the diagnosis.
§4 · What MULIO does not do
Five things, named clearly.
- 01
MULIO does not train on user data.
- 02
MULIO does not give medical advice.
- 03
MULIO does not promise outcomes.
- 04
MULIO does not run on hustle culture, toxic positivity, Buffett quotes, or "unlock your potential" language.
- 05
MULIO does not market with fake testimonials, fake press logos, or fake user counts.
Voice contract is in the codebase, not in marketing slides. We break voice contract → the voice contract page on /privacy breaks publicly. We don’t.
§5 · Crisis detection
MULIO is not a therapist. The crisis path is logged.
When the Diagnostician detects markers consistent with suicidal ideation, self-harm, or psychiatric crisis, the report execution is halted. The Strategist is bypassed. The user is routed to immediate professional resources before any strategy is generated.
In the United States: 988 Suicide & Crisis Lifeline. In the United Kingdom: Samaritans (116 123). Equivalent regional line in every supported country.
The crisis path is logged, audited, and reviewed quarterly. The detector is calibrated against published clinical screening instruments — not vibes.
§6 · Security
How your data flows. Every step named.
Journal entries are encrypted at rest using Evervault end-to-end encryption. Plaintext exists only briefly on our backend, in memory, at the moment of an AI call. Evervault itself never sees plaintext.
All data in transit uses TLS 1.2+. Firebase App Check is enabled to prevent unauthorized API access. Sign-in is via Apple — we never see your Apple password.
Anthropic SDK, not Anthropic the company. MULIO’s backend uses an open-source SDK published by Anthropic to communicate with DeepSeek’s API, because DeepSeek’s API speaks the same wire format. Anthropic the company is not in our data path. Your journal text never touches Anthropic’s servers.
Full disclosure including processing location, jurisdictional implications, retention table, GDPR rights, and the children’s-privacy COPPA section is at /privacy.
The methodology is in the code.
Download MULIO. Log a week. Sunday morning the Diagnostician writes a one-page brief on what worked and what broke. Monday morning the Strategist’s playbook is on your lock screen.
