DengRuoAn

@Umesh__digital Linux leans towards low-level server environments, whereas Mac caters more to general developers, designers, and the like.

If Linux is so powerful.. Why do most devs still use MacOS?

@jkudish I agree with you—for example, the unlimited-access experience for a certain number of hours.

I wish codex had a truly "unlimited" plan - i'd pay a good amount for it and not have to worry about usage.

@DimpleAroro23s sleep

Just bought Claude Pro. What's the fastest path to $10M?

@patloeber It's very fast, but it tends to forget the previous part of the conversation.

thank you for >1K replies, i tried to reply to as many as possible! we captured all of this and are reviewing it. 24h after launch is early, but this gives us clear signals on where to refine. thank you! tldr: we need to keep pushing :)

@mikenevermiss Don't mess around—Claude flags this kind of behavior and will ban your account.

A 22-year-old college student in shenzhen reportedly made $14,700 in one month with a 900-yuan redmi phone. his parents still think he’s spending all his time preparing for exams. here’s what the phone does. every 30 minutes, it opens six chinese news apps, scans the latest headlines, takes screenshots, and sends them to claude for analysis. claude compares that information with polymarket odds and suggests potential betting opportunities. the entire workflow runs without writing a single line of code. it’s powered by phonedriver, an open-source tool that lets ai see your phone’s screen and operate it like a human. you simply describe the task in plain english, and it does the rest. the edge comes from timing. news often appears on chinese platforms like weibo and toutiao 20 to 50 minutes before english speaking traders fully react on polymarket. that small window can create an opportunity before the market catches up. some of those apps block web scrapers, but phonedriver doesn’t scrape data. it simply looks at what’s on the screen, just like a person would, then taps and swipes automatically.

@_mohansolo Thank you very much for the remake; I look forward to seeing even better models from you.

Gemini 3.6 Flash is out on Antigravity! We just reset everyone's weekly quotas, enjoy building.

@VersunPan 所以 是哪家呢，避坑一下

讲个搞笑的事，这周接了某个大厂的商单，要我用它家的模型做个东西吹一下，但由于模型太垃圾，我 vibe 了3天，连滚带爬的改，也做不出啥好东西了。 最后，我拿 kimi k3 做了个，伪装成它家模型，昧着良心吹了😆

@wangli0001 色香味 10分 你给多少分

新疆の名物拌面、特に辣皮子拌面（ラーピーツーバンミエン）はぜひ食べてみてください。 唐辛子が入っていますが、意外と辛くなくて美味しく食べられます。

@jukan05 学方言 才是yyds

今天是我学中文的第一天。有什么学习建议吗？

@KimiDevs @AMD Intel has been abandoned.

Agent workloads bring long contexts, bursty traffic, and frequent tool calls, creating new demands for AI serving. Together with @AMD , we rebuilt the stack for Kimi K2.6 on AMD Instinct™ MI355X, with scheduler-aware multi-tier KV caching. Up to 3.2× smaller p99 TTFT and 7.7% higher total-token throughput, with no accuracy tax. Happy to see more Kimi running on AMD chips! Tech blog: amd.com/en/developer/r…

@haharung135 味真足

나를 살찌워서 잡아먹으려는지 음식이 계속 나옴 중국 좋다..

@CodeWithTamara We've made it!

🚨 DO NOT FOLLOW 🚨 I’m only 4 followers away from 500, and I promised to release Stacklight at 500, the app that alerts you when something changes in your dev stack. There's a tiny problem. It’s not finished yet 😂 Please respect my procrastination😂

@moriyama197071 日语呢 教教我吧

今日の

@SpoxCHN_MaoNing 我爱我的国家，是多么的美丽，和平，科技

In east China’s Anhui, the water glimmers like polished emeralds, crystal clear. Bamboo rafts drift gently, as if floating on air.

@pkpkmnstr93 幽默的人

中国に来たばかりの頃、夫の友人夫婦が集まる場で、中国語で自己紹介してみてよ〜って言われて、 緊張した私、 「我是小日本」 って言い間違えてその場が一瞬凍りついた事ある。 今でも会ったら必ずその話で笑われる。

@Meyane224 生日快乐🎂🎊🎈

今天是我的生日哦！

@SypuWindows 欢迎体验 自动驾驶

楽しい

@guansi 我身边也有这样的同事，但是他没有发现一个问题。聪明模型帮我们节省了很多时间？

我身边有个同事，技术能力很强，AI Coding 也用得特别多。 不过他有个特点，就是不太舍得花钱。 从免费的 Kimi 开始，一路买各种套餐。Kimi Coding Plan、Trae、OpenCode、阿里百炼……基本都试了一圈。最后得出的结论特别朴素：哪个便宜用哪个。 便宜确实是便宜。 但是便宜也有代价，很多最新的模型都没有，像 Kimi K3、GLM 5.2、Claude、GPT Codex 这些，他基本都体验不到。 他属于那种特别会过日子的工程师。 模型不给力，那就拼命榨模型。 前几天他写了一个 Skill，想根据一套 C++ 代码自动反推出详细设计文档。整个 Skill 十几 K，几乎全是提示词。 我拿内网模型跑了一遍。 五分钟结束。 吐出来一份二十 K 左右的“详细设计”。 我看完以后，总觉得哪里不对。 不是提示词的问题，而是感觉整个活儿都压在一个 Agent 身上了。 于是我几乎没改提示词，就是把流程拆了一下。让一个 Agent 负责规划，再分几个 Sub Agent 去写不同章节，最后回来汇总、补充、检查。 还是那个内网模型。 最后硬是跑出来一份一百五十 K 的详细设计。 说实话，质量还是一般，离 Claude 和 GPT Codex 那种效果还有距离。 但是我突然发现一个挺有意思的现象。 同样都在用 Agent。 他研究的是怎么把一段 Prompt 写得越来越长。 我研究的是怎么让几个 Agent 少聊天、多干活。 到底谁对，我现在也不敢下结论。 没准再过一年，模型聪明到一句话就全干完了，什么 Skill、什么提示词工程、什么工作流，统统都成了历史包袱。 不过至少今天，尤其是在内网这些能力没那么强的模型上，我还是觉得，Workflow 比 Prompt 更值钱一点。 以前写程序，总有人喜欢把三千行代码塞进一个函数里，也总有人喜欢拆成几十个小函数。 现在回头看，AI 好像也开始走这条路了。

yes yes yes

@FeiSvoice 用日语 形容好吃怎么说

生徒さんに 「中国ドラマに出てきそうなアフタヌーンティーを見てみたい」 と言われたので、38℃だけど 小程序で予約して故宮へ🇨🇳 故宫下午茶のデザートが可愛すぎて、 食べるのがもったいない🤭