随机比特
1.2K posts

随机比特
@XShude
腾讯高级开发 @WeChat|AI 工程、Agent 与研发效率实战|分享技术变革中真正不会过期的东西|Tencent/WeChat engineer writing about AI systems & developer workflows



We removed ~80% of the Claude Code system prompt for our newest models, this is what we've learned about writing system prompts, skills and Claude.MDs for them. x.com/i/article/2080…






一次令人失望的支付争议经历。 我支付 Claude AI 订阅,被扣款 200 美元。账号当天被封,没有享受到服务。 Anthropic 只退了 100 美元,剩余 100 美元一直没有退回。 我按照 Plasma 的要求: 联系商家 提供邮件证据 等待超过 30 天 接受 30 美元争议手续费 当我以为终于可以进入正式 dispute / chargeback 流程时,Plasma 却突然关闭了我的账户。 没有解释原因。 一个消费者已经按照流程提供所有证据,却在申请争议退款前被关闭账户,这样的处理方式合理吗? @PlasmaOneApp 希望你们给用户一个解释。 A payment provider should not be able to avoid handling a customer dispute by simply closing the customer’s account. I followed the required process in good faith and provided all requested evidence. After weeks of waiting, I was told the dispute could proceed — then my account was terminated. This is extremely frustrating and raises serious concerns about customer dispute handling. I hope @PlasmaOneApp @Plasma can provide a transparent explanation.



一次令人失望的支付争议经历。 我支付 Claude AI 订阅,被扣款 200 美元。账号当天被封,没有享受到服务。 Anthropic 只退了 100 美元,剩余 100 美元一直没有退回。 我按照 Plasma 的要求: 联系商家 提供邮件证据 等待超过 30 天 接受 30 美元争议手续费 当我以为终于可以进入正式 dispute / chargeback 流程时,Plasma 却突然关闭了我的账户。 没有解释原因。 一个消费者已经按照流程提供所有证据,却在申请争议退款前被关闭账户,这样的处理方式合理吗? @PlasmaOneApp 希望你们给用户一个解释。 A payment provider should not be able to avoid handling a customer dispute by simply closing the customer’s account. I followed the required process in good faith and provided all requested evidence. After weeks of waiting, I was told the dispute could proceed — then my account was terminated. This is extremely frustrating and raises serious concerns about customer dispute handling. I hope @PlasmaOneApp @Plasma can provide a transparent explanation.








🧵 1/7 I gave a Flash model a written spec for a 3D physics game. It broke the task into steps, wrote the code, used the tools, ran the project, and eventually produced a billiards game I could actually play. @AntLingAGI Ling-3.0-flash has 124B total parameters, but activates just 5.1B per token. That low activation count didn’t limit it to a simple one-shot demo. It kept a multi-stage 3D build moving from spec to playable result. This test made one thing clear: stop using your biggest model for every step of an agent workflow. The project may be complex. Most of the execution steps inside it don’t need a heavyweight planner to rethink everything from scratch. Ling-3.0-flash is built to execute. Here’s the full hands-on video: 👇













几个月前看到这样app还欣赏一下。现在在看到觉得好无聊,ai agent要么用codex要么Claude code就够了。真有特定场景的定制需要就用codex或claude写一个,几乎没啥成本了,我的工作流中已经跑了好几个运行良好的agent几乎都是Claude个把小时搓出来的。没有嘲讽的意思,纯属个人观点。













