Transfer project / Computing
A small search engine
Build and test an index over a local text collection, then explain the invariants and cost tradeoffs that make retrieval trustworthy.
- Candidate
- 0.1-foundations-candidate
- Checkpoints
- 4 local records
- Review
- Independent review pending
Milestones
- 01
Specify tokenization, normalization, and document identity before writing the index.
- 02
Build an inverted index and trace one document through every state transition.
- 03
Add queries with zero, one, and many matches plus malformed input cases.
- 04
Measure build and query behavior as the corpus grows.
Deliverables
- Source code, a tiny licensed corpus, and a deterministic test command.
- A state/invariant table for ingestion and query phases.
- A cost note distinguishing measured behavior from asymptotic expectation.
Evidence requirements
- At least one unhinted trace and one transfer explanation.
- A failure fixture that would catch an index/document mismatch.
Reproducibility checks
- Rebuilding from the same corpus produces the same normalized index.
- Every returned document can be traced to an inspectable posting entry.
No grade, ranking, certificate, or public submission. Checkpoints and notes stay in this browser. Recording a checkpoint asserts only that you created inspectable evidence; it does not certify its quality or human review.