Dear frontend and design engineers. I present to you floating-ellipsis. It's a package for detecting and customizing text truncation, implementing middle ellipses, and improving general UX.
It's headless and customizable. You can use it with @reactjs, @shadcn, or any framework.
@DanteAndroid1 Not sure what you mean by Cloudflare. But deictic is platform-, framework-, and agent-agnostic. It can be used in most modern web projects.
Introducing πππππππ
Alt-click on the page. Leave tasks, annotations, and recordings for the agent to fix.
Type /deictic in any agent. It watches and addresses the tasks you add. Uses Vercel Board internally.
Introducing ππππππ πππππ, a secure message board for people & agents.
β’ Give agents a shared place for tasks
β’ Build swarms and software factories
β’ Set permissions for humans and agents
Board is experimental. Tell us what you think. https://t.co/jAAkcf1oPt
It can be added to most web projects as a single bundler plugin. Supports Vite, Next, Bun, Webpack, and more.
See the project at https://t.co/IZlHN9PWAu
CSS tip: Use a scroll() timeline to fade the edges of a scroll container only where there is more to scroll.
The fade above grows as you scroll away from the start, and the one below shrinks as you reach the end. A list that doesn't scroll gets no fade.
Harness tip: Use tool search to load tool definitions only when the agent needs them.
This keeps MCP servers from filling the context before the first message. The agent sees the tool names, searches, and gets only the definitions it asked for.
CSS tip: Use the animation-range property to choose which part of an element the scroll-driven animation timeline considers.
"entry" is when it enters, "contain" is when it's fully in view, "exit" is when it leaves, and "cover" is when any part is visible.
Harness tip: Truncate long tool outputs and save the full output to a file. Give the agent the truncated output and the full file path.
This keeps a long log or a big JSON response out of the context. The agent can still search the file for the part it needs.
Introducing HappyCAD: the AI teammate that designs hardware 10Γ faster, every mechanism proven.
It uses the best LLMs like Fable 5.2 and GPT-6 Astra to design hardware.
Comment HAPPYCAD to get an invitation code for https://t.co/9fQH0mZmwG.
@minke3338611 Yeah, normally I would spell those out one by one myself. But I just roasted the agent, and it found a good design by itself. I am questioning why it didn't do it in the first place though.
CSS tip: Use the view() function for animation-timeline to run a scroll-driven animation based on where an element is in its scroll container.
The animation starts as the element enters the view and finishes as it leaves.
@_chenglou's Handwired compiles programs into plain neural networks. That gave me an idea. If a UI is a neural network, the techniques used to verify neural networks can check it.
I first built a checker for Handwired itself to test the idea. Once that worked, I translated React Nativeβs layout engine (Yoga) into the same kind of circuit, and wrote checks like: does any element overflow its parent or overlap its neighbour, at any screen size and any text length?
It finds layout bugs a tester would likely miss, including a quirk in Yoga itself. Tests try a few sizes. But the checker covers every size and text length in a range. That means when a layout passes, itβs a proof of correctness, not just a sample. Iβm very excited about what this opens up for UI testing.
CSS tip: Use the scroll() function to run a CSS animation based on the progress of a scrollbar. This is called a scroll-driven animation.
Instead of the progress of time, how far the view has scrolled becomes the timeline of the animation.
Harness tip: Use Code Mode to allow agents to call tools as code.
This allows agents to make multiple calls in a single turn, operate on data, and keep the data out of the context, thus saving costs.