Han Su

CMS
NYU SH

[Han Su @suhan1996]

We provide private social media at https://t.co/47tW7ZnvLD. Now looking for users with more than 10k twitter followers who want to build a private social sphere--we can help you set up your personal site for free if you don't have one;)

[Han Su @suhan1996]

Context is the new product

[Han Su @suhan1996]

I’m building that with rust (minimal size and CPU consumption, runs 24/7 in the background unless stopped manually, works better than granola) and we now have a team of 2. We need a designer to join the founding team if you are interested. Send an email to [email protected] and I’ll send you a beta version. VCs seeing this are also welcomed to join the beta.

[Han Su @suhan1996]

@Saccc_c 欢迎用https://t.co/c8mdLhvfaq上线项目，可以有public channel讨论idea等功能

[Han Su @suhan1996]

If you want to share your vibe coded agent to your friends but do not have a server or domain. Use https://t.co/47tW7ZnvLD

#vibecoding

[Han Su @suhan1996]

@boardyai building Privoce, a video+AI co-work space with identity-driven rooms and shared AI widgets. count me in for Boardy Pro

[Han Su @suhan1996]

Is there a real-time AI speech translation API that can let participants all speak their mother tongue in a zoom call and I will hear the translated version as if everyone is speaking in my mother tongue? Basically an API to realize the AirPods live translation feature). #AI

[Han Su @suhan1996]

@bhalligan Anyone interested in M&A our rust written chat (supporting matrix)/social video platform?

[Han Su @suhan1996]

The future: Vibe coding + VPS = anyone can build & launch AI-native apps, bypassing platforms.

Social login (Google, Apple, WeChat) unifies data & payments.

Mobile browsers rise again, dodging app stores, as vibe coding revives the web in the AI era.

[张平 @pingzhang632]

切星球视频没完没了了！🤣🤣🤣，这一版没那么漂亮，但似乎更真实一些。

[Elon Musk @elonmusk]

Legal immigration to America is ridiculously slow & difficult, even for super talented people. Needs to be fixed.

[Junyi Zhu @JunyiZhuMIT]

I am joining University of Michigan, EECS department as an Assistant Professor starting Jan 2025, leading the Sensing, Intelligence, and eXperience (SIX) Lab. We are recruiting PhD students (fall 2025), master & undergraduate researchers!
More information: https://t.co/ONZMa0afdY

[Huaishu Peng @huaishup]

📢 Thinking about pursuing a PhD in technical HCI? Our group at UMD CS is recruiting! We work on novel/sustainable fabrication systems, XR haptics, and AI for accessibility, publishing at #CHI and #UIST. Learn more at https://t.co/TYo3ZiZZGB & feel free to reach out /DM!

[Zining Zhang @_ziningzhang_]

🤩What an incredible experience! #UIST2024 Thank you to all my collaborators @huaishup @umd_emd @ZYYan20H @jsli_21 for your support, and to everyone who attended my talk and demo. It was a pleasure having you all! 🥰

[Zining Zhang @_ziningzhang_]

Excited to present and demo my first first-author HCI paper at #UIST2024! 🎉 "JetUnit: Rendering Diverse Force Feedback in Virtual Reality using Water Jets" showcases how a compact, self-contained device delivers a wide range of force intensities & frequencies. 🌊Come try it out!

[Han Su @suhan1996]

@waylybaye 我最近用AI合成自己的声音学习，类似听到自己再说流利的英文，大脑会自动感知到哪里的肌肉不太对，这是一个很好的方式，欢迎私信讨论。

[Han Su @suhan1996]

VideoLLM...interesting. Should be more like an editing based on captions job rather than a video generation job, this videodb seems like a good start.

[Anup Gosavi @IamAnupGosavi]

LLMs/AI applications are great with text, but with video they can't support simple requests like "iPhone 15 event highlights in 10 minutes"

Why? Because video infra hasn't evolved since 2001. Introducing VideoDB, a new kind of LLM Infra for videos. See https://t.co/OApnSIyH2S

[AI Will @FinanceYF5]

1/

Vercel v0 的新代码截图太疯狂了。

有人刚刚截取了定价表的屏幕截图，然后使用人工智能在几秒钟内将其转换为代码！

它甚至输出代码并允许您修改任何特定部分。https://t.co/ZnGzeOOikN

[Barret李靖 @Barret_China]

self-operating-computer，https://t.co/0L3DqgNGhI，这个项目演示了如何让 GPT-4V 来控制自己的电脑，你需要做的就是告诉它完成一个怎样的任务，例如，打开 Google Docs 写一篇文章，然后发布并分享给同事。

它的 Prompt 写的比较简单，定义了一个可以与机器交互的 DSL，主要包含三种动作：Click/Type/Search，分别对应 mouse_click/keyboard_type/mac_search 几个封装好的系统函数。

每次程序执行动作时，都会携带任务目标、上一步执行结果以及当前屏幕截图作为上下文，然后将信息传递给 ChatGPT，并让它给出��一步操作指示。

这个项目中 Prompt 的定制化程度偏高，说明完成复杂的工作还比较有挑战，如果参考之前提到的 Mutil-Agent 方案，https://t.co/n1S8YdwlWn，做出来效果应该会更好一些。

[Geoffrey Hinton @geoffreyhinton]

Yann LeCun thinks the risk of AI taking over is miniscule. This means he puts a big weight on his own opinion and a miniscule weight on the opinions of many other equally qualified experts.