使用者：基于爬取的真实世界使用情况的上下文感知技术栈推荐
在构建新产品时，我不断遇到同一个问题：选择工具所花费的时间比实际构建产品的时间还要长。“十大工具”列表大多是广告驱动的，而大型现有工具往往在缺乏上下文的情况下被推荐。
因此，我创建了UsedBy。
在技术层面，我们会爬取公共来源，以提取真实世界中的工具使用情况和技术栈组合。在此基础上，我们利用GPT添加上下文理解，例如项目类型、团队规模和使用场景，而不是将工具视为平面的类别。
我们的目标不是对工具进行排名，而是帮助人们理解哪些工具在实际中是一起使用的，以及原因是什么。我们的推荐是基于上下文的，而不是单纯的流行度。
我们的盈利模式故意保持轻量化：广告是根据用户的兴趣信号（如点赞和技术栈互动）展示的，而不是基于竞价或赞助排名。
欢迎提问关于爬虫、数据模型或我们如何处理上下文和推荐的相关问题。反馈也非常欢迎。
查看原文
When building new products, I kept running into the same problem: choosing tools took longer than building the actual thing. “Top 10” lists were mostly ad-driven, and LLMs tended to recommend large incumbents without much context.<p>So I built UsedBy.<p>Under the hood, we crawl public sources to extract real-world tool usage and stack combinations. On top of that, we use GPT to add contextual understanding, such as project type, team size, and use case, instead of treating tools as flat categories.<p>The goal is not ranking tools, but helping people understand which tools are actually used together in practice and why. Recommendations are context-aware rather than popularity-based.<p>Monetization is intentionally lightweight: ads are shown based on user interest signals like likes and stack interactions, not bidding or sponsored rankings.<p>Happy to answer questions about the crawlers, data model, or how we handle context and recommendations. Feedback very welcome.