21 lines
1.2 KiB
Markdown
21 lines
1.2 KiB
Markdown
---
|
|
version: "0.1.2"
|
|
level: copilot
|
|
processes:
|
|
design: pair
|
|
implementation: copilot
|
|
testing: copilot
|
|
documentation: copilot
|
|
review: none
|
|
deployment: none
|
|
---
|
|
|
|
This format is based on [AI-DECLARATION.md](https://ai-declaration.md/en/0.1.2).
|
|
|
|
## Notes
|
|
|
|
- The Rust port of Paul Uszak's Java ent3000 tests was made with OpenAI's ChatGPT, which provided guidance, debugging help, and inspiration throughout the development process.
|
|
- The scoring was designed together with Anthropic's Claude (Claude Code) and implemented by it. This covers calibrated p-values, the combined score, the survival depth, the calibrator, and the score in the CLI, C and Python bindings. I set the purpose (a health score for a source), chose the method (combined p-value plus QNGmeter-style survival depth) and the calibration approach (empirical), and approved the plan and each step.
|
|
- `src/calibration/table.rs` is not written by hand or by AI. It is generated by `examples/calibrate.rs` from simulated random data.
|
|
- The calibration was checked against fresh random data, up to millions of samples per size (see the Accuracy section of the readme). The raw p-values of all tests still match the Java original on the three test files.
|
|
- No AI was used for code review or deployment.
|