Experiments are active tests with a hypothesis and measured outcome. → Protocols are the stable system experiments can change.
H
Hypothesis
Every experiment starts with a falsifiable hypothesis. What specifically do I expect to change, and by how much?
P
Protocol
Defined intervention, duration, and measurement. Baseline established before starting. No mid-experiment changes.
D
Data
Results from 26 live data sources — sleep, HRV, glucose, body composition, habits. Compared against pre-experiment baseline.
Experiments that prove out become Protocols. Failed experiments are published in full in Discoveries.
Currently running
Loading…
Launch roadmap
Week 1
First experiment launches. Top-voted from library.
Week 2-3
Second batch launches. Results from Week 1 published.
Month 2+
First graduations to Protocol. Community experiment suggestions open.
Start here — 3 recommended experiments
Experiment library — vote on what I test next
Loading library…
How the library grows

I've been running informal experiments my whole life — reacting to a book I read, a podcast I heard, something a friend swore by. I just never applied the scientific method. Now I have the data to actually notice if something changes. Our engine scans research journals and podcasts to surface new experiment candidates based on sourced evidence. The Board reviews each one for safety and testability before it enters the library. Every experiment traces back to a source.

Monitored sources
PubMed Alerts Huberman Lab The Drive (Attia) Examine.com ISSN Position Stands Cochrane Reviews FoundMyFitness Barbell Medicine
Suggest an experiment

Have an idea for something I should test? The Board reviews every submission for safety and testability.

Looking for action, not science? Challenges has short-term provocations born from your data — no hypothesis required.
On the methodology

These are N=1 experiments — single-subject self-experiments with no control group. The scientific limitations are real: no blinding, no randomization, seasonal confounders, regression to the mean. Results apply to me specifically and may not generalize.

What makes them useful anyway: consistency of measurement. With 26 live data sources running continuously, the baseline data is dense enough that meaningful signal can emerge from even a single-person study — particularly for outcomes with low day-to-day variance like resting HRV and fasting glucose.

The library experiments are graded by evidence tier: ●●● Strong (multiple RCTs), ●● Moderate (observational studies), ● Emerging (preclinical or anecdotal). And each experiment is classified as Measurable (has a biomarker endpoint from my 26 data sources) or Behavioral (compliance tracking is the outcome).

The record — completed experiments
No results yet
Every experiment — including the failures — will be published here in full. First results expected by mid-April.