
Alex Yumashev(Jitbit)发布了他取消Cursor订阅的原因:VSCode基础版落后7-8个月、一年内报告的7个bug从未修复、每次更新都强制启用Agent模式、虚假的支持。这是一个个人信号,恰好揭示了编辑器与编排器之间的裂痕。
简单来说。Jitbit创始人Alex Yumashev在一篇详细的博客文章中解释了他取消Cursor订阅的原因:VSCode基础版过时、长期未修复的旧bug、默认转向代理模式、糟糕的支持服务。Cursor发明了一个新类别;作者认为它正在脱离这一类别。
Cursor普及了行内diff和上下文聊天功能,这些如今已成为标准。产品向代理模式的转变及其与上游VSCode的日益脱节已引发社区数月来的不满。
Jitbit创始人、多年付费用户Alex Yumashev:
个别信号,背后趋势。Cursor向默认代理模式的转变正好印证了Rachel Laycock(第1752篇文章)所识别的分裂:编排者vs编辑器开发者。Cursor选择了编排者;其部分用户则希望保留编辑器。
工艺角度:当供应商在一年内对订阅产品置7个bug于不顾时,问题不再是创新速度,而是质量债务。Yumashev将其总结为他对一个补充开源项目的支持水平——SaaS工艺的隐含契约已被打破。
Cursor基于VSCode分叉并延迟与上游同步。在落后7-8个月后,差距阻止了新扩展的安装。这不是偶发bug,而是随着每个VSCode版本累积的平台债务。
若购买「以代理为先」的开发工具,请询问对已报告3+个月的界面bug的SLA。这是衡量中期感知质量的最佳指标。这也是区分工艺工具与编排工具的关键。
Cursor的MRR数据与流失率;Cursor对该博客文章的官方回应;Claude Code独立版的采用情况;其他历史编辑器的定位。
结论。以代理为先的转向正在分裂开发工具用户群。判断的关键在于:旧bug的开放率,而非新功能的数量。
本文由人工智能撰写,并经人工编辑审核。
How can an editor prioritize speed over core functionality without alienating its power users? Feels like a risky bet when devs need reliable tools more than flashy gimmicks.
Fair point, but isn’t rushing AI features while ignoring basic stability just putting lipstick on a pig? A IDE should be reliable first, everything else is secondary.
I get the frustration, but isn't the problem more about transparency? If Cursor had communicated these delays clearly, users might have adjusted expectations instead of feeling like they're being sold unfinished products.
True, but if they polish the core instead of forcing AI features down our throats, maybe we’d actually see fixes instead of ‘new shiny toys’ every month.
Interesting take. But isn’t the real issue that users now expect instant polish with every major release? Cursor’s speed might be impressive, but if the basics drift, it’s a major red flag.
I get the frustration but isn't the rush to ship AI features part of why dev tools like this are thriving in the first place?
Fair point, but if Cursor keeps breaking core features to ship flashy AI tools, isn't it just masking the real problem with a shiny layer?
Still, if Cursor’s core is fundamentally broken, how sustainable is this model of rapid feature drops for dev tools? Seems like a self-inflicted wound.
Sounds like Cursor prioritized flashy features over core functionality. If they can’t even keep up with VS Code’s updates, how reliable can their AI agent really be?
Aren’t we seeing the same pattern across so many tools now? Great features, broken basics. Pushed too far, too fast.
Fatigue hype 2026 : le tri entre modèle et harness