# Experiment: EXP-<paper>-<NN>-<slug>

**Date started:** YYYY-MM-DD
**Paper:** P1 | P2 | P3
**Status:** designing | running | analyzing | done | abandoned

## Question
What single question does this experiment answer? Phrase it so the answer is a number, a yes/no, or a short qualitative finding — not "explore X."

## What would falsify the answer I expect
This is the most important field on the page. If you cannot answer this, you do not yet have an experiment, you have a hope.

> Example: "I expect RMSE < 5%. If RMSE > 10%, the spring stiffness model is wrong and I need to characterize each spring individually before assembling the kit."

## Setup
- Hardware / software / dataset / participants
- Exact versions, exact random seeds, exact prompts (for LLM work)
- Cost in time and money

## Procedure
1. Step
2. Step
3. Step

(Anyone — including future-you in 6 months — should be able to reproduce this from the procedure alone.)

## Raw outputs
Path to the raw data in the repo. Never paste numbers here without also committing the raw file.

```
data/2026-MM-DD/run01.csv
data/2026-MM-DD/run01-photo.jpg
```

## Analysis
What did the numbers say. One paragraph.

## Conclusion (one sentence)
The single sentence that would go in the paper if this were the only experiment that ever ran.

## What this tells me to do next
Forward link to the next experiment ID, or "ship to draft."