Ask HN:为什么 Claude 在回答关于 Claude Code 的问题时表现如此糟糕?

1作者: nico5 个月前
每次我在网页界面上向 Claude 提问关于 Claude Code 的问题时,它给出的回答都非常无用。<p>一两周前,我问它关于我遇到的 329 错误,它却指向一个不存在的状态页面。当它最终找到了正确的状态页面,我让它阅读,它却拒绝了。<p>就在刚才,我问它是否有 Claude Code 与 GitHub 的集成。Claude 给出了一个通用的“我没有具体信息”的回答,并建议我查看官方文档。<p>相比之下,当我向 ChatGPT 提出完全相同的问题时,它进行了快速的网页搜索,搜索了多个网站,并给出了详细的答案,其中包含指向 Anthropic 官方关于其 GitHub 集成的文档的直接链接(甚至还提到了考虑到他们最近发布的新速率限制公告的自动化考虑事项)。<p>这感觉像是一个可以解决的问题。Anthropic 难道不能简单地使用 RAG(检索增强生成)来处理他们自己的文档,从而更好地回答关于他们自己产品的问题吗?当他们的竞争对手提供的信息比 Anthropic 自己的 AI 提供的关于 Anthropic 工具的信息更好时,这令人沮丧。<p>这似乎是一个基本的的产品集成问题。<p>其他人也遇到过这种情况吗?
查看原文
Every time I&#x27;ve asked Claude (on the web interface) questions about Claude Code, it gives me really unhelpful responses<p>A week or two ago I asked about 329 errors I was getting, and it pointed me to a non-existent status page. When it finally found the right status page, I asked it to read it and it refused<p>Just now, I asked if there was a Claude Code integration with GitHub. Claude gave me a generic &quot;I don&#x27;t have specific information&quot; response and suggested checking official docs<p>In contrast, when I asked ChatGPT the exact same question, it did a quick web search across multiple sites and gave me a detailed answer with direct links to Anthropic&#x27;s own docs about their official GitHub integration (and even a comment about considerations of automation given their recent new rate limits announcement)<p>This feels like a solvable problem. Couldn’t Anthropic just RAG their own docs to provide better answers about their own products? It&#x27;s frustrating when their competitor gives better information about Anthropic&#x27;s tools than Anthropic&#x27;s own AI does<p>Seems like a basic product integration issue<p>Anyone else run into this?