Skip to content

0al-spec/Metrics

Repository files navigation

Metrics

Metrics for AI-Driven Development under CC BY 4.0 license.

Papers

  1. SIB - Specification-Implementation Balance - A non-normative diagnostic framework for observing the balance between specification (intent), implementation (code), and AI inference cost in modern AI-assisted software development. Rather than proposing optimization targets or productivity metrics, the framework is designed to expose structural and economic shifts in the process of intent-to-implementation conversion.
  2. SIB Full Metrics - An extended draft framework covering pre-implementation, process, structural, and retrospective defect metrics.
  3. SIB Economic Observability - A non-normative economic observability lens for inference, verification, build cost, and structural cost proxies.

Metric Packs

See METRIC_PACKS.md for the plugin-style metric pack model, authority states, run artifact sketches, and interpretation guardrails used when SpecGraph consumes these metrics as read-only diagnostic overlays.

The draft Idea-to-Spec Lifecycle Telemetry pack defines objective observability signals for product idea-to-spec workflows, including clarification load, answer materialization, ontology grounding, candidate repair, temporal progress, workflow friction, approval readiness, and read-model publication state. Its initial machine-readable report contract lives in schemas/idea_maturity_metrics_report.schema.json. The validator output contract lives in schemas/idea_maturity_metrics_validation_report.schema.json, and the stable CLI/compatibility rules are documented in VALIDATOR_CONTRACT.md. Example report invariants can be checked without third-party dependencies:

python3 scripts/validate_idea_maturity_examples.py
python3 scripts/validate_idea_maturity_examples.py path/to/report.json
python3 scripts/metrics.py validate idea-maturity path/to/report.json \
  --output runs/idea_maturity_metrics_validation_report.json

scripts/metrics.py is the stable CLI entrypoint for sibling repositories. It keeps Metrics as the source of truth for metric-pack validation while allowing producers such as SpecGraph to emit structured validation reports without copying validator logic.

Building

Each paper is in its own directory with a Makefile:

cd SIB
make build   # Compile to PDF
make view    # Compile and open PDF

License

CC BY 4.0 (Creative Commons Attribution 4.0 International)

About

Metrics for AI Driven Development under CC BY 4.0 licence

Resources

License

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors