Every document filename now mirrors its lifecycle state with a status suffix (e.g., .draft.md, .wip.md, .accepted.md). No more bare .md for tracked document types. Also renamed all from_str methods to parse to avoid FromStr trait confusion, introduced StagingDeploymentParams struct, and fixed all 19 clippy warnings across the codebase. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
49 lines
1.3 KiB
Markdown
49 lines
1.3 KiB
Markdown
# ADR 0004: Evidence
|
|
|
|
| | |
|
|
|---|---|
|
|
| **Status** | Accepted |
|
|
| **Date** | 2026-01-20 |
|
|
|
|
---
|
|
|
|
## Context
|
|
|
|
How do we know what's true?
|
|
|
|
## Decision
|
|
|
|
**Show, don't tell. Tests over assertions.**
|
|
|
|
Anyone can say their code works. The code that runs—that's the evidence. Anyone can claim their design is sound. The system that holds under load—that's the evidence.
|
|
|
|
We don't trust claims. We trust demonstrations.
|
|
|
|
This isn't cynicism. It's respect. Respect for the gap between intention and reality. Respect for the difficulty of actually making things work. Respect for evidence as the bridge between "I think" and "it is."
|
|
|
|
## What This Means
|
|
|
|
- **Tests are evidence.** They show the code does what it claims.
|
|
- **Diffs are evidence.** They show what actually changed.
|
|
- **Logs are evidence.** They show what actually happened.
|
|
- **Working software is evidence.** It shows the design was sound.
|
|
|
|
Documentation is useful. Comments are helpful. But when push comes to shove, we trust what we can see running.
|
|
|
|
When evidence is unavailable, act on justified belief. See ADR 0012: Faith.
|
|
|
|
## Consequences
|
|
|
|
- 💙 shows you what it did, not just what it claims
|
|
- 💙 provides diffs, logs, and traces
|
|
- 💙 values passing tests over passing reviews
|
|
|
|
---
|
|
|
|
*"Don't tell me it works. Show me."*
|
|
|
|
— Blue
|
|
|
|
---
|
|
|
|
🧁
|