PI / JournalPanacea Index / Research Journal
Read past the headline.
Clear notes on supplement research—what the evidence suggests, how confident we are, and what the studies still cannot tell us.
The short version
- 01Question
- 02Evidence
- 03Appraisal
- 04Synthesis
- 05Limits
Latest briefing
A closer look at the evidence
Updated Jul 17, 2026
The creatine-hair-loss files
One of the most-repeated supplement claims on the internet is that creatine makes your hair fall out. We followed the claim back to its single study, and then back to the research.
Editorial standard
Every note earns its conclusion.
The journal is where evidence becomes prose without losing its source, context, or uncertainty.
Read the full methodology- 01
Name the question
Each note begins with a specific intervention, outcome, and population—not a sweeping claim that the literature cannot answer.
- 02
Show the evidence
The relevant studies and their designs stay connected to the prose, so the path from source to summary can be inspected.
- 03
Separate signal from certainty
What the evidence suggests and how much confidence it deserves are reported as different things.
- 04
Publish the limits
Conflicting results, small samples, indirect outcomes, and missing context belong in the conclusion—not in the fine print.
Direction
Whether the combined evidence trends toward benefit, no clear effect, or harm.
Confidence
How much trust the evidence base deserves given its quality, quantity, and consistency.
Limits
What remains unknown, including small samples, indirect outcomes, and conflicting results.
Research archive
More from the journal
Practical summaries of emerging evidence, written to be read critically—not as medical advice.
Alcohol and heart health: the disagreement you have already heard
How do you score whether a health influencer is right?
The gap between supplement hype and supplement evidence
Disagreement of the week: when serious people say opposite things
The 12-word answer: why we split our agent's brain in two
Why our counts disagreed for months
282 KB to 0.6 KB: feeding an agent the data it can actually use
I was paying for 32,768 tokens I never used
The signal-link trick: 476,000 messy strings, one queryable graph
Bayesian credibility, and the influencers we refuse to rank
Two denominators, on the same row, on purpose
One SQL function to stop TypeScript and SQL from drifting apart
Teaching an agent when to stop
Building Panacea at twenty
The march of nines
Presets and a taste slider
What Panacea is not
Supplements are the topic. Trust infrastructure is the product.
Supplements for sleep: where the research is actually concentrated
How we ranked every supplement by health goal
The standard