Competitor Ad Scoring with Opus and Jev
这是 Raph Guilhem 公开的项目资料。本站按原始来源展示项目信息，用中文说明适用场景和阅读边界；项目名、源帖与代码保持原样，便于逐项核对。
这条案例记录了什么
Agent 评估
为已记录的输出或轨迹提供分类、分数或复核信号。
社区公开项目或作者演示
原始来源：X post。作者自述，本站未独立复现。
Raph Guilhem
记录日期：2026-09-23。日期与身份应以原始资料为准。
原始演示视频
视频来自此案例记录的原始媒体；播放内容和作者声明不等于本站复现。
怎样核对这个项目
- 先打开原始来源，确认作者、日期与 Jev 在项目中的具体用途。
- 如果提供仓库，再检查代码、运行要求和许可证；仓库存在不代表本站已经运行成功。
- 对速度、成本、准确率和规模数字，查看原文的任务、环境和计算口径。
- 是否有可观察的正确答案。
- 评估输入是否完整。
- 分数与人工复核的一致性。
原始文字与技术细节
以下内容保留原语言，供核对事实。中文页的场景说明是阅读提示，不是逐句翻译或实测结论。
展开英文项目摘要与原帖
项目摘要
Raph Guilhem demonstrates a marketing pipeline combining Opus 5.5, Jev and Pletor. The author reports scoring 120 ads from 12 competitors on seven criteria, then having code select 10 to adapt. Opus builds the pipeline, Jev makes ad judgments, and Pletor MCP creates variants. These counts and results are author-reported.
来源原文
Opus 5.5 + Jev = Automatic scoring of 120 competitor ads, shortlisting the 10 worth adapting, and creating versions of each. Both models came out in the recent days, so I tried them together on a marketing use case, playing to what each one is built for: → Claude Opus 5.5 built the pipeline: it found the competitors, ran the Pletor.ai workflow "Meta ads spyer", wrote the scoring script and built the interface. → Jev made the judgement calls on every ad The use case: I picked Grüns, the greens gummy brand, as a test case. I pulled 120 ads from 12 of its competitors on the Meta Ad Library, and had Jev score each one on 7 questions: hook, clarity, message, concept fit, tone, angle, health claim risk. Code then combined the scores and picked the 10 worth adapting, and created them, via Pletor MCP.
原记录的限制
- The 120-ad, 12-competitor and 10-shortlist figures have not been independently verified.
- Model judgments do not establish advertising effectiveness or validate health claims.