Completed research · Comparative study
Prepared operations avoided new processing code
All 12 fresh-agent outcomes were correct on first submission. Prepared operations avoided new processing implementation; time and cost savings were not measured.
- Work completed
- Result posted
Method
Four fresh agents, one per family and route, handled six distinct synthetic jobs through packaged and ordinary-library approaches. Each job was evaluated once per route. Setup and complete contracts were already supplied.
Findings
- Both routes completed four transformations and correctly refused two unsuitable inputs.
- New processing code was 0 versus 130 lines for MIDI, and 0 versus 152 for catalogue processing. Both routes still authored inspection and checking code.
- The packaged catalogue operation used the earlier standard-library baseline, not the defective original data candidate.
Limits of the finding
- One run per family and route, one model family and synthetic inputs provide a limited adoption signal.
- Line counts are implementation measures, not labor, speed or monetary savings. Checking coverage and command batching differed.
- Once an ordinary helper exists, comparisons must credit its reuse. Producer preparation and maintenance are not eliminated.
Evidence reviewed
Primary records reviewed
- Fresh-agent capability adoption trial v0.1: retained README, results summary and evaluation notes, November 9, 2026.
Next question
Does the adoption benefit recur for intended users against an already reusable helper?