# {slug}: {one-line title}

<!-- Copy to: 1_ideas/YYYY-MM-DD_{slug}.md -->

## Metadata
- **Date**: YYYY-MM-DD
- **Author**: {agent or human name}
- **Status**: draft | validated | in-progress | done | rejected

## Hypothesis

**Question**: What happens if we {specific change}?

**Prediction**: We expect {metric} to change by {amount} because {mechanism}.

**Falsification**: If {metric} does NOT change by at least {threshold}, the hypothesis is wrong.

## Approach

Brief description of the implementation. What files change? What's the core modification?

```python
# pseudocode or key snippet showing the idea
```

## Why this is novel

How does this differ from what's already in RESEARCH_JOURNAL.md? What paper/idea is it based on?

## Expected implementation effort

Small (< 1 hour) / Medium (1-4 hours) / Large (> 4 hours)

## Pueue label (for queuing)

```
Q: does {change} improve {metric}? H: expect {delta}
```

---

## Subagent Critique

<!-- Append subagent feedback here before implementing -->
<!-- Prompt used: "Is this idea sound? What are the failure modes? Is the hypothesis testable?" -->

**Rating**: 1-5 (1=bad, 5=excellent)

**Strengths**:
-

**Failure modes**:
-

**Verdict**: proceed / revise / reject

---

## Result (fill in after experiment)

**Commit**: SHORT_SHA
**Report**: [9_reports/YYYY-MM-DD_{slug}.md](../9_reports/YYYY-MM-DD_{slug}.md)
**Observed**: {metric} = {value} (expected {expected})
**Verdict**: confirmed / partially confirmed / refuted / inconclusive