Day 8 building VibePaper in public.
We shipped Velic, a personal agent you name and @ in a chat thread.
My favorite trick: send it a reference clip and it breaks it down shot by shot. Then you ask for one of those shots as a still.
Learning from a video used to take an hour of scrubbing. Now it's one message.
我悟了 personal agent是未来
云电脑重要性不只是执行任务
如果有一个简单的任务
我可以抬手问豆包 得到答案
那如果下次有其他的任务
那同类的基础模型也能回答
下一次可能选的就是kimi
所以基础模型永远在卷能力
但如果是personal agent
我有很长的任务（甚至过夜）让bot跑
我会预估一个差不多该结束的时间
打开看看
这一看 就是留存 就是产品依赖
然后 开始下一个任务
这就是微信、qq的逻辑
你永远在等待ta的回复
I don’t think AI infrastructure will kill SaaS. Haven’t you wondered why some people pay the same for ChatGPT and Claude as we do, but they make good money and we don’t?
Today we launch VibePaper 2.0.
Three linked modes: Canvas, Design, Work.
Canvas: Node-driven image/video creation
Design: Layer & layout editing
Work: Task workflows with Paper Agent
Shared context. Switch between manual and AI work.
VibePaper 2.0: Bring ideas to life.
Just finished watching Hisefield’s interview.
A key insight for AI products: flashy visuals don’t win.
In the AI era, the real value is solving user problems — and helping users make money.
An open letter to AI comic & short‑drama creators.
Many AI tools shine in demos yet break in real production.
We respect builders like TypeNow & Lip‑TV. At VibePaper, we prioritize reliable workflows over flashy one‑click tricks.
We are live, no waitlist.
#BuildInPublic
