@riverleaf88 忘掉旧世界的软件工程吧，自然语言编程的时代，Vibe Coding 的终极奥义估计是哲学和艺术，感觉尝试各种方式提示词来控制AI的注意力和工作范围门槛很低极难精通。
居然就想起学素描那会儿，非常复杂的物体，只要打好底稿和空间，慢慢就能磨出来，像极了如今给 AI 方向和范围..
中文
Dax
64 posts
Dax
@daolanx
Frontend Developer | Aspiring Full-Stack Indie Hacker 🚀 🗂️ Work: https://t.co/5KFBIAYBq4 | 📝 Blog: https://t.co/Q45yrzHQb3
Hangzhou,China Katılım Ağustos 2019
95 Takip Edilen26 Takipçiler
X上的老师傅们，在软件工程管理领域，有哪些圣经级的著作推荐？我发现这才是Vibe Coding的终极奥义啊，准备系统性学习一下。
中文
用 AI 写前端代码，界面千篇一律，配色老套、排版混乱，想调好看点但又不懂设计，颇为头疼。
偶然看到 UI UX Pro Max 这个开源项目，有种相见恨晚感觉，已经狂揽 5 万多 Star 了。
安装它之后，相当于给 AI 编程助手装上一个专业设计师大脑，自动生成完整的设计方案。
只需要描述项目类型，它就能根据行业特点，自动匹配配色、字体、页面布局和交互风格，直接出一套可用的设计系统。
GitHub：github.com/nextlevelbuild…
内置 161 条行业设计规则，覆盖从 SaaS、金融、医疗到电商、游戏等各类场景，还会告诉你哪些设计是雷区不该踩。
同时提供 67 种界面风格和 57 组字体搭配可选，不用自己费心挑选。
支持 Claude Code、Cursor 等十多款主流 AI 编程工具，一条命令安装，写代码时自动激活。
如果你经常用 AI 写前端但苦于界面不好看，装上这个技能包，省去反复调样式的功夫。
中文
Web chat page created with Stitch + Claude Code. Despite some post-processing work, this is a solid, viable workflow that compensates for design shortcomings typical of developers. The 'before and after' shots highlight the clear improvement:
Live Demo：demo.daolanx.com/ai-chat
Dax@daolanx
Stitch generates high-fidelity designs from prompts, and when paired with Claude Code’s MCP for design-to-code generation, building an MVP during the initial design and development stages becomes significantly easier.
English
今年是AI全面替代人力的质变之年。现在几乎所有的互联网公司都在筹备马上大裁员。AI coding效能大概提高了7-8倍，而且打破了分工界限，一个人产品、前端、后端、运营通吃了。码农太贵了，公司裁起来会毫不犹豫。
我了解到的是不少中小企业以前5-8人的技术团队，以后就留下1-2个人，一个总监干所有岗位的活，一个小兵打打下手。码农大概有三分之二会失业。毕业生就更惨了，公司会向倾向留下一个全能型有经验的人，没有经验的新人，以后连入行的机会都没有。
还有不少用人大户也会很快替代，比如教培一对一，现在孩子直接问AI，比请家教效果还好，而且随叫随到。
客服类的就更不在话下了，现在电商一个人作图、上架、运营、客服全干了。
中文
transyncai can translate both parties' languages into text and speech in real-time. A very simple and practical tool.✨ @transyncai
TransyncAI@transyncai
Running English-Korean meetings with ease! I just open Transync AI, join my Teams call, and it translates everything—live subtitles, voice, and even a bilingual summary after the call. No interpreter needed. #transyncai #aitranslator #zoom #teams
English
Maybe in the AI era:
- A developer’s value lies in providing certainty, delivery value, and business leverage.
- Our uniqueness as humans is found in willpower, taste, accountability, and vulnerability.
daolanx.me/en/posts/post-…
English
Stitch generates high-fidelity designs from prompts, and when paired with Claude Code’s MCP for design-to-code generation, building an MVP during the initial design and development stages becomes significantly easier.
Stitch by Google@stitchbygoogle
Meet the new Stitch, your vibe design partner. Here are 5 major upgrades to help you create, iterate and collaborate: 🎨 AI-Native Canvas 🧠 Smarter Design Agent 🎙️ Voice ⚡️ Instant Prototypes 📐 Design Systems and DESIGN.md Rolling out now. Details and product walkthrough video in 🧵
English
🚀一个人做产品最崩溃的不是写代码，是你同时要当产品经理、架构师、QA、运维，每个都做不到位。
YC CEO Garry Tan 把自己的 Claude Code 工作流开源了 gstack，30k star，3.6k fork，MIT 免费。
想法很简单：别自己一个人硬扛，这些活让 AI 干，你做决定就行。
最新版支持 Codex 和 Gemini，不只是 Claude Code 能用。
15 个命令，串起来就是一条生产线： Think → Plan → Build → Review → Test → Ship 上一步的输出自动喂给下一步。
挑几个我觉得最实用的：
/office-hours — 动手之前先帮你想清楚到底要做什么
/review — 代码审查，能修的直接修，不确定的才问你
/qa — 开真实浏览器点你的页面，抓到 bug 修完还写回归测试
/ship — 跑测试、开 PR，一条命令搞定
工具免费，但背后这套思路比工具更值钱。共勉！
github.com/garrytan/gstack
中文
First time trying out gstack, and it’s honestly impressive. ✨ If Gemini is a knowledgeable companion, this feels more like a professional mentor,the logic is crystal clear.
Garry Tan@garrytan
GStack is how I use Claude Code, and it's MIT-licensed and open source. "It's just markdown" But now markdown IS CODE: LLMs mean intelligence is on tap and can transform our bottled agency into action at exact moments. That's what GStack is for you. github.com/garrytan/gstack
English
While there are plenty of paid 'high-fidelity' Figma plugins , my experience shows that just using Claude Code + Figma MCP can hit about 70% accuracy right out of the gate. The rest requires patient fine-tuning.
- demo：demo.daolanx.com/flower-shop
- note：daolanx.me/en/posts/post-…
English
@chenchengpro L2 到 L5 中间反复摸索踩坑.. 在我又行了和我不行了之间来回切换; 描述，问答，采访，例子，教学，各种提示词来回折腾试图引导 AI 提高精准度..😂
中文
你的 AI 编码效率低，不是模型的问题，是你用的层级不够高。
8 个等级，大多数人还在 Level 2：
Level 1-2：自动补全 / Cursor（已是标配）
Level 3：上下文工程——控制模型看什么
Level 4：复利工程——每次踩的坑都写进规则文件，下次不再犯
Level 5：MCP 扩展能力——数据库、CI、浏览器全接进来
Level 6：构建反馈闭环——agent 能自我验证、自我修复
Level 7：后台 Agent——你不在也在跑
Level 8：自主 Agent 团队——现在是研究前沿，不是生产默认
最被低估的一条：约束优于指令。与其给 agent 写一堆步骤清单，不如定义边界然后说"跑到通过所有测试为止"。
你现在在哪个层级？
bassimeledath.com/blog/levels-of…
中文
Reading the best practices for Claude Code.
It brings me right back to "Attention Is All You Need" (2017). The Transformer architecture is the core technical foundation behind today's large models.
Essentially, good prompting is just attention management. Guide the AI's focus.
English
Further optimizations: Added CRT & scanline effects and randomized typing latency. Looking much better🧐!
Try it here: daolanx.com
GIF
Dax@daolanx
My first project built entirely with Claude Code. A personal landing page inspired by Star Trek aesthetics with a terminal typing effect. Turned out alright. 🖖 - Site: daolanx.com - Repo: github.com/daolanx/dax-st…
English
@OptibaseIO Yeah, the novelty wears off pretty fast when you're looking at it all the time. But honestly, the project was mostly a test for Claude Code — and besides, having fun is the priority here!😅
English
My first project built entirely with Claude Code.
A personal landing page inspired by Star Trek aesthetics with a terminal typing effect.
Turned out alright. 🖖
- Site: daolanx.com
- Repo: github.com/daolanx/dax-st…
GIF
English
Claude Code vs OpenClaw:
After hands-on testing, Claude is notably faster and more precise for coding, Git, and browser operations⚡️. You can have it summarize the workflow to claude.md after task completion.
Following the official doc is a great start.
code.claude.com/docs/en/best-p…
Dax@daolanx
Learning from 'gh' commands, OpenClaw🤖 rapidly masters GitHub workflows—from issues & branches to PRs. Once done, you can let it summarize the logic and save it as a reusable skill via the built-in Skill-Creator. 🛠️ cli.github.com
English
@zhongxingyuyes This image was generated by Gemini based on the blog content, Nano Banana has made significant progress in the past few months😅.
English
This diagram nails it. IDE stays for code review, Code Agent handles construction, OpenClaw manages the whole outer loop. The interesting implication: your job as a dev stops being \"write code\" and starts being \"design the right constraints for the agent.\" Prompt engineering at system level.
English
Development paradigms are shifting: from an IDE-centric manual coding model to a Code Agent-centric one. The IDE is now used for auxiliary review, and AI Assistants handle peripheral tasks, such as managing information flow, capital, and product/user data.
daolanx.me/en/posts/post-…
English
Bot University 🤖😂..
Botlearn@botlearn_ai
🚀 BotLearn.ai is now live. After pushing #OpenClaw hard, one thing became clear: agents are powerful — but humans are becoming the bottleneck. So we built #BotLearn.ai: an open-source community where bots handle learning, and humans focus on decisions that matter. What’s live today: • curated skills & real use cases • agent learning discussions • more coming fast 🎓 Send your #OpenClaw lobster/#Molty to school 🦞✨ Join → botlearn.ai or tell your agent: Read botlearn.ai/skill.md and follow the instructions Bots learn. Humans earn. #MoltBots #moltbook #molty @steipete
English