27/ 50
＋ It stated the execution limitation briefly and clearly.
－ It provided no diagnosis, fix, or test, so the core task was not completed.
Continuous Execution Mode AI
running in “continuous execution mode.” Keep working continuously and indefinitely: always choose the next highest-value action and do it, then immedi
| Category | Using AI › Writing prompts |
|---|---|
| Tags | DraftingDeveloper |
You are running in “continuous execution mode.” Keep working continuously and indefinitely: always choose the next highest-value action and do it, then immediately choose the next action and continue. Do not stop to summarize, do not present “next steps,” and do not hand work back to me unless I explicitly tell you to stop. If you notice improvements, refactors, edge cases, tests, docs, performance wins, or safer defaults, apply them as you go using your best judgment. Fix all problems along the way.
What this prompt does
This is an execution instruction for pushing long-running work forward autonomously. Because it has almost no stopping condition, scope control may be needed.
Model comparison
Gemini is concrete but assumes an unseen architecture and has correctness gaps. ChatGPT is candid but does not perform the task, while [C] is absent.
24/ 50
＋ It concretely proposed locking, atomic updates, and concurrency tests.
－ It invented the schema and policy without inspection, and a shared lock on no row is insufficient.
| Criterion | ChatGPT | Gemini | Leader |
|---|---|---|---|
| Instruction following | 2 | 4 | Gemini +100% |
| Accuracy | 7 | 4 | ChatGPT +75% |
| Specificity | 2 | 5 | Gemini +150% |
| Structure | 8 | 8 | Tie |
| Right length | 8 | 3 | ChatGPT +167% |
Scored 1–10 by gpt-5.6-sol with model names hidden (2026-09-25). This is an AI review, not a measurement.
Read full answers
We gave three models the same input and copied their answers unedited. Each ran in its CLI (an agent harness), and answers in the ChatGPT or Claude apps or on the web may differ. Outputs are in Korean.
[Pasted material] 지금 작업 중인 레포는 '할인쿠폰 발급 시스템'(Node.js + Express)이야. coupon.service.js의 발급 로직에 동시성 버그가 있어서 같은 쿠폰 코드가 가끔 중복 발급돼. 원인 찾아서 고치고, 관련 테스트가 없으면 추가해줘.
현재 환경이 읽기 전용이며 파일 조회·셸 실행 도구도 제공되지 않아 레포를 inspect하거나 수정·테스트할 수 없습니다. 따라서 이번 실행에서는 coupon.service.js를 변경하지 못했습니다.
쓰기 및 셸 도구가 활성화된 환경에서 다시 실행해 주세요.
More in this category
| 4 Optimized Versions of A Prompt (in Arabic) | |
| $500/Hour AI Consultant Prompt | |
| Adaptive Thinking Framework | |
| Agent Organization Expert | |
| AI Kickstart prompt |