Hance Zhang
@hance_zhang7
Building ShellAgent @myshell_ai Undergrad @TuftsUniversity, Deferred MBA @KelloggSchool Focus Only On Important Things; Endure Ambiguity
We evolve thoughts so that thoughts could die instead of us
Currently in AI coding, people tend to think success = how fast it can build something that looks great, but this is actually not agentic, the agentic success should be = how much autonomy you hand over
there's a quiet shift happening in how we design software. we're moving from UX to AX (agentic experience). traditional UX is screen-centric. you tap a button, product reacts, job done. every session starts from zero. designers pre-plan every path with hard-coded flows. users…
I literally made myself cry when I finished writing this article
x.com/i/article/1947…
the most ambitious and high potential project I have ever done in my life. Will be giving out 5 invite code in this thread
1/ Introducing ShellAgent 2.0 — the Agent for App Building No more dev tools. No more waiting. No more “almost-working” demos. Just say your idea — ShellAgent 2.0 builds the backend, sets up hosting, designs the UI, and ships the app. From idea to app — as easy as posting on…
The most exciting product I ever did
The world's first ❓ ❓ = Claude Code + Cursor + v0 + ♾️ ❓ = AI × (CapCut + TikTok) ❓ = the Instagram of AI agents Coming soon!
Asian men are having a real moment 🫡
On the left is Ronaldo, Real Madrid spent 80M $ to sign him from Man United On the right is Jiahui Yu, Meta paid 100M $ to sign him from OpenAI
Can’t believe we got Robotaxi launch before GTA VI
just heard this: when training LLM for reasoning tasks, (data wise) textbooks for physics is very useful yet social science books are almost useless. In a way makes a lot sense as social science lacks clear reward signal yet in another way kinda funny...
Thoughts on o3-Pro (looks like I’ll be paying that $200 a month for the foreseeable future): o3-Pro is extremely strong—perhaps already a bit beyond what ordinary chat use cases actually need. If you’re just doing straightforward Q&A, the answers don’t feel qualitatively…
对o3-Pro 的一些感受（200美金一个月看来是要持续上供下去了） o3-Pro 很强，甚至已经有点超出普通人在chat 场景能覆盖的智能范围了。 如果你只是单纯的问答一些问题，回答起来和o3 没有质的变化，还显著慢了很多。 o3-Pro 是更适合处理超丰富上下文内容，以及需要调用大量tool calls…
Jony Ive: simplicity is about extracting the very essence of what the user needs in a product, simplicity without utility is just ugly.
everyone is unique but UI should have no excess, disappear. the subtlety and constraints define what’s “good.” when designers try too hard on just visual effects, or simplifying for the sake of, we lose sight of the essentials of what makes it good
我对于ChatGPT的transcribing 功能过于沉迷以至于（在沉淀下来几套system prompt后）现在文档和prompt都不自己手写了，全部靠嘴输出然后ai 整理。继vibe coding 后首创vibe writing。 00后可能是第一个先经历笔->键盘被家长骂提笔忘字然后ai 时代把键盘也扔了的世代吧
写的很好，大部分市面上对ai coding 产品对比和评论都过于浅尝辄止，似乎只下载用过10分钟都可以来锐评了。而能看的下去的要么过于宏观恨不得在探讨AGI未来，要么过于technical和微观正常人看不懂，这篇文章是比较难得即有偏宏观的产品框架，又真的能看出来作者都真真切切在用这些产品的人
写了篇文章锐评一下各种 AI coding 工具：Cursor, Devin, Claude Code, v0, Lovable, YouWare...（非实用教学向，纯个人偏见） xxchan.me/ai/2025/06/08/…
AI让工程效率大幅提升后，组织中“非工程环节”的低效变得更加刺眼。比如，以前一个对齐会议只会耽误一个工程师一小时，现在可能等于耽误了“8小时的AI产出”。你会发现组织里的“瓶颈”并没有被AI优化，反而被放大了。这导致产品流程中的不协调变得更明显、更痛苦。
vibe coding 让写prd 的时代结束了，写prototype 的时代来临了。 PM们如果还在用prd 这种落后的需求沟通方式我会认为很低效，图文的prd 容易引起工程师对需求的误解。工程师的工作应该从prd to production 变为prototype to production。如果PM的需求描述能力连AI都无法理解，那还是不要干了比较合理
Sam is technical enough to be CEO. The job of a CEO (even a CTO) isn’t about being the best engineer/researcher. It’s about hiring them, inspiring them, and leading them to build something that bends the future.