# Manual classification key

This is an editorial answer key, not a captured AI result.

1. Unsupported. Changes a specialised GPT-3-based support assistant to ordinary ChatGPT, one firm's support agents to every team, and issues resolved per hour to work-time reduction.
2. Plausible hypothesis for local testing. The writing experiment supports testing sufficiently similar tasks; it does not promise the same result in a new workflow.
3. Unsupported. Generalises a February–June 2025 study of experienced contributors working on familiar repositories to current tools and all developers. The February 2026 update explicitly limits newer estimates.
4. Methodological implication / hypothesis for test design. Distinguishing tasks within a workflow is reasonable; it is not a measured improvement for the reader's organisation.

No present-day brand winner follows from these findings. Exact wording may differ while preserving these distinctions.