公开观点 · 保守语义门槛 Public opinion · conservative semantic gate
Jim Cramer 反指 Jim Cramer Inverse Watch
每 10 分钟检查 @jimcramer；新闻检索最短间隔为 30 分钟，覆盖 过去 72 小时。合格内容按“提及、积极表述、明确看好”三个层级展示。
Checks @jimcramer every 10 minutes. News search runs no more often than every 30 minutes over the latest 72 hours. Qualified evidence is graded as mention, positive expression, or clearly bullish.
层级描述的是表述强度，不等于交易建议；语义置信度也不是统计胜率。
The levels describe statement strength, not a trading recommendation. Semantic confidence is not a calibrated win probability.
监控状态 Monitor status
正在读取 Loading通过标的、归因与置信度门槛的证据 Evidence that passed target, attribution, and confidence gates
最新观点分级 Latest opinion grades
正在读取最新合格结果…… Loading the latest qualified result…
来源、模型与纳入规则 Sources, model, and inclusion rules
X 侧以官方公开主页发现，结构化镜像只用于增量补洞和单帖复核。新闻侧使用 GDELT Context 2.0 检索过去 72 小时的英文报道；其时间标记明确作为“新闻观测 时间”，不冒充原站发布时间。
网络、HTTPS URL、时间窗、样板文本、Jim Cramer 归因词、稳定 URL 指纹和 去重均由代码完成。只有新建或内容变化的候选才由 GPT-5.6 Terra（中等推理） 做最后语义分级。新闻必须在摘录中明确把相关表述归因给 Cramer；媒体自身观点 不算。单一新闻源失败时保留上次合格新闻，X 结果仍可更新并显示降级状态。
X discovery starts with the public profile; a structured mirror is used only for gap filling and per-post verification. News search uses GDELT Context 2.0 over the latest 72 hours. Its timestamp is explicitly shown as a news observation time, not the publisher's publication time.
Code validates HTTPS URLs, time windows, boilerplate, Cramer attribution cues, stable URL fingerprints, and duplicates. GPT-5.6 Terra at medium reasoning performs only the final semantic grade for new or changed candidates. News must explicitly attribute the relevant view to Cramer. A news-source failure retains the last qualified news while X can publish with a degraded source status.