89 comments • Comments Link

• 现在很流行做一个能屏蔽 AI 内容的 Hacker News 版本，各类项目尝试用关键词过滤、基于 BERT 的分类器或人工策展等方法来应对 AI 帖子泛滥的问题。

• 内容过滤的技术手段各式各样——从基于浏览器的 uBlock Origin 自定义规则，到扫描 GitHub 提交记录和仓库文件以识别 AI 作者身份的复杂自动化后台系统。

• 创建 Hacker News 的专门视图在精神上被看作是一种"hacker-newsery"行为，即便这会把个人偏见带进信息流，也反映出社区想掌控信息消费的愿望。

• 这些工具的目的并不全是排斥 AI 技术，更多是为了避开企业营销、对大型 AI 实验室的盲目吹捧以及重复出现的影响者内容——这些东西目前占据了首页。

• 用户希望有更一体化的解决方案，比如兼容浏览器扩展或移动应用，而不是只能依赖那些镜像现有内容的分离网站。

• 有人觉得用 AI 自身去过滤"AI 内容"既讽刺又有趣，但当过滤器漏掉很多 AI 相关条目时，也会让人感到沮丧。

• 像 HckrNews 这样的社区替代品因为采用简单按时间排列的"reverse river"布局而受到好评，为基于算法推荐或高度策展的动态信息流提供了一种更可预测的替代选择。

• 在如何界定"AI 内容"（从纯粹的企业广告到真正的技术试验）以及如何在不漏掉有趣技术进展的前提下呈现高质量讨论方面，存在重大分歧。

• 对于常刷 Hacker News 的高级用户来说，键盘操作和界面响应速度至关重要。有报道指出，很多自定义界面难以达到原站点的性能和流畅度。

• 一些人把这些工具看作缓解社区疲劳的临时出口，类似过去面对新闻过载时的反应；另一些人则建议转到 Lobste.rs 等平台，寻求不同的策展体验。

这场讨论反映出社区对首页上泛滥的 AI 内容普遍感到厌倦，尤其对重复出现的企业营销和影响者驱动的帖子感到沮丧。尽管从简单的关键词过滤到复杂的机器学习分类器都有尝试，但大家普遍希望在信息流的策展上能有更多用户掌控权。总的来说，"去 AI 化"信息流被看作是技术社区的一种自然（尽管带点讽刺）的反应，但是否能在不牺牲真正创新讨论的前提下精确界定并隔离"AI 内容"，仍存在较大分歧。 • Creating a version of Hacker News that filters out AI-related content is a popular pursuit, with various projects attempting to address the saturation of AI posts via keyword filtering, BERT-based classifiers, or manual curation.

• Technical approaches for content filtering range from simple browser-based uBlock Origin custom filters to sophisticated, automated backend systems that scan GitHub commit logs and repository files for AI authorship.

• Building specialized views for Hacker News is considered inherently "hacker-newsery" in spirit, even if it introduces personal bias into the feed, as it reflects the community's desire for agency over their information consumption.

• The motivation for these tools is often less about a total rejection of AI technology and more about avoiding the constant influx of corporate marketing, "glazing" of major AI labs, and repetitive influencer content that dominates the current front page.

• Users have requested more integrated solutions, such as browser extensions or mobile app compatibility, rather than relying on separate websites that mirror existing content.

• The use of AI itself to power the filtering of "AI content" is viewed by some as an ironic or humorous necessity, though others find it frustrating when filters fail to catch all instances of AI-related stories.

• Existing community alternatives like HckrNews are praised for their simple, chronological "reverse river" layouts, which offer a predictable alternative to algorithmic or heavily curated feeds.

• There is significant debate over where to draw the line between "AI content"—which ranges from pure corporate advertising to genuine technical experimentation—and how to effectively surface high-quality discussions without missing interesting technical developments.

• Keyboard navigation and UI responsiveness are critical for power users who frequently browse Hacker News, with reports that custom interfaces often struggle to match the performance and fluidity of the original site.

• Some participants argue that these tools serve as a temporary relief valve for community fatigue, similar to historical periods of intense news overload, while others suggest switching to platforms like Lobste.rs for a different curation experience.

The discussion reflects a shared community fatigue regarding the current dominance of AI-related content on the front page, specifically highlighting frustration with repetitive corporate marketing and influencer-driven posts. While diverse methods for filtering this content exist—ranging from simple keyword lists to complex machine learning classifiers—the consensus suggests a desire for greater user control over feed curation. Ultimately, the effort to "de-AI" the feed is viewed as a natural, albeit ironic, expression of the community's technical culture, though users remain divided on the feasibility of perfectly defining and isolating "AI content" without sacrificing genuinely innovative technical discourse.