AAT/SFT Research Notes

ArchSig Roadmap

Roadmap status stays separate from theorem status.

The roadmap tracks tooling phases, implemented pipelines, schema and validator coverage, advisory reports, empirical feedback, planned gaps, and standardization targets. It is mutable planning, not mathematical proof status.

Current snapshot

ArchSig is beyond a minimal AAT diagnostic tool. It includes Lean and Python graph extraction, policy and runtime evidence, AIR / Feature Extension Report, PR / CI governance artifacts, empirical feedback artifacts, Architecture Dynamics artifacts, and a bounded SFT forecasting pipeline.

  • Implemented pipeline CLI and workflow paths that generate artifacts from real inputs.
  • Schema + validator Artifacts with fixed shape and validation, but limited inference from real inputs.
  • Empirical / operational Dataset, calibration, feedback, and outcome records that do not imply causal theorem status.

Remaining gaps

The roadmap separates product surface from planned gaps. The current implementation is useful for bounded observation, review, SFT projection, and feedback artifacts, but several adapters and calibration paths remain outside the product surface.

  • Extractor and framework adapters Call graph, data dependency, dynamic import, plugin loading, and framework convention handling remain adapter work.
  • SFT input adapters Markdown PRD / Spec input is implemented; GitHub Issue JSON and AI proposal JSON adapters remain future work.
  • Calibration Real dataset calibration, organization policy tuning, and operational threshold management require local feedback data.
  • Formal boundary Tool output, forecast output, and empirical correlation do not become Lean theorem claims without explicit formal bridges.

B0-B13 phases

The phase list moves from AIR boundary and static reports through runtime, semantic, generated-change, repair, empirical, CI, policy ecosystem, schema, operational feedback, dynamics, and SFT forecast work.

  • B0-B5 AIR, Feature Extension Report, runtime / semantic integration, generated-change provenance, repair and synthesis prototype.
  • B6-B10 Empirical validation, PR review integration, extractor / policy ecosystem, schema compatibility, and operational feedback loop.
  • B11-B13 Architecture Dynamics tooling, SFT forecasting MVP, and implemented SFT forecaster pipeline.

SFT forecasting

B12 fixed ArtifactDescriptor, OperationSupportEstimate, ForecastConeSkeleton, ConsequenceEnvelope, and ForecastCalibrationHook as schema + validator artifacts. B13 connects real Markdown PRD / Spec / Issue / AI proposal input to the bounded pipeline.

B13 pipeline
real PRD / Spec / Issue / AI proposal
  -> ArtifactDescriptor builder
  -> OperationSupportEstimate generator
  -> ForecastConeSkeleton generator
  -> ConsequenceEnvelope generator
  -> end-to-end sft-forecast command

Standardization targets

Standardization targets include artifact schema naming, claim boundary vocabulary, measurement boundary vocabulary, non-conclusion preservation, schema compatibility policy, example fixtures, report consumer contract, forecast input / output contract, and forecast boundary vocabulary.

Roadmap progress does not create a Lean theorem claim. It records the implementation and validation surface of the tooling layer.