Kai

💻 大厂程序员・🚀 独立开发者・🎨 设计爱好者

AI 工具实战分享・AI 编程经验交流・AI+ 设计实践

芸術と技術の狭間に立って

[Kai @KQ89778230]

@libukai ai 时代想清楚做什么和为什么非常重要

[Kai @KQ89778230]

@yeshenggezi @leo114119 这 tm 有点离谱了，是 plus 嘛 还是 pro

[Kai @KQ89778230]

@rachell331 @GeorgetownLaw 是的 hhh，毕竟是马老板的平台，AI 和科技类话题基本都很火，受众里爱折腾的理工男也居多

祝你在 X 玩得开心！

[Kai @KQ89778230]

@leo114119 可以，让子弹飞一会儿

[Kai @KQ89778230]

@opencode support landed before most people had even read the launch post, day one model integration is table stakes for coding tools now

[Kai @KQ89778230]

@mylifcc classic openai

[Kai @KQ89778230]

@shneural motion design is a fun one because there is no benchmark for it, you just watch the output for five seconds and you know how good it is!

[Kai @KQ89778230]

@cognition same score at 81% less cost is the actual headline. a decent chunk of that benchmark gap was just willingness to burn tokens

[Kai @KQ89778230]

@thsottiaux people spent two days doing token per second forensics and building quota conspiracy theories and the answer was just capacity, classic

[Kai @KQ89778230]

@AndyL5cc 日常写代码 6.1 Sol 够用了，便宜量大。Opus 5.5 留给苦难任务

[Kai @KQ89778230]

@ArtificialAnlys the 7 day replacement is the real story. one point under astra for a quarter of the cost per task, and cache reads halved to 10 cents.

[Kai @KQ89778230]

@sashikantsingh_ internal benchmarks have been crushing the competition since 2025, the independent charts are where it gets interesting

[Kai @KQ89778230]

@theo a model doing way better in the tool devs actually use than in the benchmark everyone cites is peak benchmark comedy

[Kai @KQ89778230]

@MaryCruzxfwf same，求互关🙏

[Kai @KQ89778230]

@tuxi2036 一起加油，求互关

[Kai @KQ89778230]

@mingAIgo 同刚开蓝 V，求互关～

[Kai @KQ89778230]

@rangerswin129 same，求互关

[Kai @KQ89778230]

@Kongingking same，互关一下？