Public reference
Comparison guide
Choose a context approach by the work it must support. No single category is best for every task.
| Dimension | CoreStory | General retrieval over files | Wiki or documentation system | IDE assistant alone |
|---|---|---|---|---|
| Primary unit | Typed claims and relationships tied to code and governed context | Retrieved passages | Authored pages | Current conversation and selected files |
| Persistence | Shared intelligence model across sessions | Depends on the index and refresh process | Persistent, but manually maintained | Usually session- or workspace-scoped |
| Evidence | Claims can carry provenance, anchors, and epistemic status | Often points back to source passages | Depends on authoring practice | Depends on the model, tools, and citations exposed by the client |
| Best fit | Complex systems where agents and people need reusable, governed understanding | Finding relevant text quickly | Human-authored guidance and decisions | Interactive coding with local context |
Interpretation
These are architectural differences, not benchmark results. Products within each category vary, and effective systems often combine them. Evaluate exact model versions, tools, context, cost, latency, reliability, and task outcomes under your own conditions.