01 · Comparison
Five decision criteria distinguish prompt-led exploration from delegated implementation inside explicit plan, review, and verification boundaries.
5 decision criteria · first-person editorial guidance
02 · The two modes
Prompt-led exploration with lightweight structure.
Delegated implementation inside explicit plan, review, and verification gates.
03 · Decision criteria
| Criterion | Vibe coding | Agentic coding | Result |
|---|---|---|---|
| Exploration friction | Start with a prompt and discover the shape while working. | Define enough context and acceptance criteria before delegation. | Favours Vibe coding |
| Setup cost | Low for a disposable experiment. | Higher because repository context and checks need owners. | Favours Vibe coding |
| Repeatability | Depends heavily on the conversation and operator memory. | The plan, repository guidance, and checks can be reused and reviewed. | Favours Agentic coding |
| Reviewability | Review often happens continuously inside the same conversation. | The delegated change can be inspected against a named plan and acceptance criteria. | Favours Agentic coding |
| Production delivery | Needs an additional structured review and verification pass. | Designed to produce a bounded change with verification evidence before release. | Favours Agentic coding |
My recommendation · reviewed Jul 13, 2026
The result is 3–2 because these criteria ask how work moves from an idea into a reviewable repository change. Vibe coding wins the first minutes of an uncertain problem. Agentic coding wins once the task, acceptance criteria, and release owner are known.