Dependency graph
Evaluate a proposed change in the context of affected components and relationships.
PUBLIC · AI CHANGE INTELLIGENCE
AI Change Impact Engine
A lightweight reference implementation that turns proposed AI changes into an explicit, explainable release-risk forecast.
Publicv0.1.0Reference implementationApache-2.0
The problem
Why this project exists
A passing evaluation suite does not explain which dependent behaviors may move when an AI component changes. The engine makes dependencies, historical outcomes and behavioral cohorts explicit inputs to a release forecast.
Verified capabilities
What is actually implemented
These capability descriptions are intentionally limited to the evidence-backed implementation surface.
Historical evidence
Use prior outcomes as explicit risk features rather than hidden intuition.
Behavioral cohorts
Preserve cohort-aware evidence so aggregate scores do not erase behavioral differences.
Risk model
A dependency-free logistic model produces an inspectable forecast.
Release policy
Forecasting feeds a separate ALLOW / REVIEW / BLOCK policy decision.
Verification
Core, model and calibration tests plus CI provide executable evidence.
Architecture
How the system is structured
The diagram below represents the current verified architecture boundary; planned or environment-dependent components are described in the engineering notes rather than silently promoted.
Proposed Change
↓
Dependency Graph
↓
Historical Evidence
↓
Behavioral Cohorts
↓
Risk Model
↓
Release Policy
Evidence and authority
The core is intentionally dependency-light and deterministic.
The project does not present a probabilistic forecast as a guarantee.
Engineering deep dive
Decisions, trade-offs and boundaries
Technology names matter less than the engineering contracts they enforce.
The core is intentionally dependency-light and deterministic.
Forecast output preserves contributing factors for human review.
Calibration tests make uncertainty an explicit engineering concern.
The project does not present a probabilistic forecast as a guarantee.
Project status
Public · evidence-backed status
v0.1.0 working reference implementation; dependency graph, evidence, forecast and release-policy surfaces are verified.
ClassificationPublic implementation
EvidenceRepository implementation, tests and project records cross-checked
ClaimsCurrent, planned and limitations kept separate
DeploymentNot changed by portfolio integration
PUBLIC IMPLEMENTATION
Inspect the implementation directly.
Public implementation — source can be inspected directly.