Methodology
Sources & Methodology
Last updated: 9 April 2026
This page describes where Milligram's compound data and AI-generated information comes from. Milligram is an informational tracking tool — it does not provide medical advice, diagnosis, or treatment.
1. About Milligram AI
Milligram AI is powered by Google's Gemini large language model. It draws from the app's built-in compound database, published pharmacokinetic research, and manufacturer documentation to answer questions about compounds, dosing, reconstitution, and protocols. Responses are framed as general educational information, not personalised medical advice.
2. Compound Database
The app includes a database of 103 compounds with the following data for each:
- Pharmacokinetic parameters (half-life, bioavailability, absorption rate constants, volume of distribution)
- Standard dosing protocols (dose ranges, routes, frequencies, cycle durations)
- Mechanism of action summaries
- Expected timelines (onset of effects, time to steady state)
- Storage and reconstitution information
- Common stacking combinations
This data is derived from published pharmacokinetic studies, manufacturer prescribing information, and established clinical dosing guidelines. Pharmacokinetic calculations (level buildup curves, steady-state projections) use standard one-compartment and two-compartment mathematical models.
3. Reference Categories
Pharmacokinetic Research
- PubMed — National Library of Medicine database of biomedical literature
- Google Scholar — Academic literature search across disciplines
- DrugBank — Comprehensive drug and pharmacokinetic data resource
Regulatory & Manufacturer Data
- FDA Drugs@FDA — FDA-approved drug products with prescribing information
- DailyMed — FDA label and prescribing information for marketed drugs
- European Medicines Agency — EU-approved medicines with assessment reports
Notable Prescribing References
- Semaglutide (Ozempic/Wegovy) — FDA prescribing information
- Tirzepatide (Mounjaro/Zepbound) — FDA prescribing information
- Testosterone Cypionate — DailyMed label
4. AI Methodology
When you ask Milligram AI a question, your message and relevant context (age, sex, goals, active compounds) are sent to Google's Gemini API. Gemini generates a response drawing from its training data, which includes published medical and scientific literature. The app also injects compound-specific data from its built-in database into the AI's context to improve accuracy for dosing, reconstitution, and protocol questions.
Milligram AI does not generate formal per-response citations because the underlying language model synthesises information from its training data rather than retrieving specific papers. The reference links above represent the categories of sources that inform both the compound database and the AI's training knowledge.
5. Pharmacokinetic Model
Level buildup curves and steady-state projections use standard compartmental pharmacokinetic models:
- One-compartment model: C(t) = (F × D × ka) / (Vd × (ka - ke)) × (e-ket - e-kat)
- Multi-dose superposition: Individual dose curves are summed to project steady-state accumulation
Parameters (half-life, bioavailability, absorption rate, volume of distribution) are population averages derived from published clinical pharmacokinetic studies. Individual responses vary based on body composition, metabolism, injection technique, and other factors.
6. Limitations
- AI-generated responses may contain inaccuracies. Verify important information independently.
- Compound data is reviewed but has not been independently peer-reviewed.
- Pharmacokinetic projections are approximations based on population-average parameters, not personalised predictions.
- Milligram does not have access to your medical history, lab results, or health records.
- This app is not a substitute for professional medical advice. Always consult a qualified healthcare provider before starting, changing, or stopping any compound or protocol.
7. Contact
Questions about our sources or methodology can be directed to enzoanderson2010@icloud.com.