Completed research · Comparative study
When does a packaged capability add value?
Both implementations met the authored contracts. A later audit found a validation defect that favored the simpler catalogue implementation.
- Work completed
- Result posted
Method
Two bounded operations—MIDI passage extraction and catalogue reconciliation—were each compared with a conventional implementation under the same complete specification. Forty-four distinct synthetic cases produced 88 implementation-case executions; expected refusals counted as correct outcomes.
Findings
- Each implementation initially passed 22 of 22 cases in its family. Ten separate positive and corrupted-output checker controls are outside the 88 executions.
- A subsequent code-informed malformed-CSV probe was incorrectly accepted by the original data candidate and correctly refused by the standard-library alternative. Frozen results were retained; this was not a blinded additional case.
- The simpler catalogue implementation was selected for follow-on work. No additional functional benefit over the conventional implementations was demonstrated.
Limits of the finding
- Cases, implementations and assessments were authored within one model family. The full domain contract was supplied equally.
- These results establish neither general reliability nor statistical equivalence, measured time savings, customer demand or formal Machine Edition qualification.
Evidence reviewed
Primary records reviewed
- Bounded capability comparison v0.1: frozen run summary, checker controls and separate evidence audit, November 9, 2026.
Next question
Which recurring task leaves a useful gap after a maintained, reusable alternative receives full credit?