43/ 50
＋ It presents the noted facts in concise one-line bullets.
－ It entirely omits the new landing-page design review.
Draft a weekly status report
Converts scratch notes into an outcome-first report, without inventing progress the notes do not show.
| Category | Office work › Reports |
|---|---|
| Tags | DraftingReformattingOffice workerTemplate |
Write a weekly report from my notes. Structure: 1. **Completed this week** — outcomes, not activity. "Finished X" rather than "worked on X". 2. **In progress** — for each, the current stage and what remains. 3. Next week. 4. Issues and support needed. *If there are none, write "none". Do not manufacture one.* Rules: - Bullet form. One line per item. - **Do not infer an outcome the notes do not state.** - Use only the completion figures I gave you. Where there is none, describe the stage instead of estimating a percentage. - A manager reads this in five seconds. **Keep section 1 to five items** — group them if there are more. - No adjectives about effort. The reader judges that.
After pasting, fill in the fields at the bottom (My notes · Reader)
What this prompt does
Weekly reports fail by listing activity instead of outcomes. This rewrites to completion state and refuses to estimate progress figures you did not supply.
Model comparison
ChatGPT is the most accurate and concise but omits the landing review. Claude preserves uncertainty but adds excess commentary, while Gemini is scannable yet makes unsupported inferences.
38/ 50
＋ It avoids treating the landing review as completed.
－ The follow-up questions exceed the requested format and length.
36/ 50
＋ Its concise sections make the report easy to scan.
－ It invents landing completion and remaining analysis work.
| Criterion | ChatGPT | Claude | Gemini | Leader |
|---|---|---|---|---|
| Instruction following | 8 | 7 | 6 | ChatGPT +14% |
| Accuracy | 9 | 9 | 5 | Tie |
| Specificity | 8 | 9 | 7 | Claude +13% |
| Structure | 9 | 7 | 9 | Tie |
| Right length | 9 | 6 | 9 | Tie |
Scored 1–10 by gpt-5.6-sol with model names hidden (2026-09-24). 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.
My notes: 월: 신규 랜딩 시안 검토. 화: 광고 소재 3종 제작, 2종 반려. 수: 데이터 대시보드 오류 확인 요청. 목: 소재 수정 완료, 집행 시작. 금: 초기 성과 확인 중. 대시보드는 아직 답이 없음. Reader: 팀장 (숫자보다 막힌 지점을 본다)
1. 완료한 일
- 광고 소재 3종 제작 완료, 이 중 2종 반려.
- 반려된 소재 수정 완료 및 집행 시작.
2. 진행 중
- 광고 초기 성과 확인 중.
- 데이터 대시보드 오류 확인 요청 후 답변 대기 중.
3. 다음 주
- 광고 초기 성과 확인 지속.
- 대시보드 오류 확인 요청에 대한 답변 확인.
4. 이슈 및 지원 필요
- 대시보드 오류 확인 요청에 아직 답변이 없음. 담당 부서 확인 필요.
More in this category
| AI Productivity Artifact Generator | |
| AI Workflow Automation Specialist | |
| Comprehensive Image Analysis Report | |
| Corporate Intel Report | |
| Developer Daily Report Generator |