All activity
Jai Prakash Singhleft a comment
I built a React component that renders streaming LLM text on instead of the DOM, eliminating layout reflows entirely. The problem: every token from an LLM stream triggers a DOM mutation → style recalculation → layout reflow → paint → composite cycle. At 50+ tokens/second, this causes visible scrollbar jitter and dropped frames. The solution: measure text in a Web Worker (via Intl.Segmenter +...

ZeroJitterStream LLM tokens without layout jitter. Canvas + Web Worker
Jai Prakash Singhleft a comment
Hey Product Hunt! 👋 I'm Jai. I built StreamMD because I noticed every AI chat app has the same hidden performance bug: they re-parse the entire markdown string on every single token. After 500 tokens, you're re-parsing a 2,000+ character string 500 times. The total characters processed grows QUADRATICALLY. At 100 tok/s, that's insane. StreamMD's incremental parser only processes the NEW...

StreamMDStreaming MD for LLMs. 300x fewer chars parsed per token
Every AI chat app re-parses the entire markdown on each token—O(n²) slowdown. After 500 tokens, that’s 500 full re-renders. StreamMD fixes this with incremental block parsing: only new text is processed, completed blocks are memoized (no re-render), and only the active block updates. Result: ~300x less work. Built-in syntax highlighting (15 langs, ~3kB, zero deps). Drop-in:

StreamMDStreaming MD for LLMs. 300x fewer chars parsed per token
Every AI chat app has the same bug — scrollbar jitter when streaming tokens.
ZeroJitter is a React component that renders streaming LLM text on instead of the DOM. The result:
→ 0 layout reflows (DOM causes 1 per token)
→ 60fps locked (DOM drops to 45-58fps)
→ <1ms layout time (DOM: 0.3-2ms)
→ Zero dependencies
→ Full accessibility (aria-live mirror)
→ International text (CJK, BiDi, Thai, emoji)
Text measurement happens in a Web Worker. The main thread just paints pixels.

ZeroJitterStream LLM tokens without layout jitter. Canvas + Web Worker
Jai Prakash Singhleft a comment
Hi Product Hunt! 👋 I’m Jai, the maker of Teleport. I got incredibly frustrated with the modern file-sharing workflow: waiting 10 minutes to upload a 5GB video to Google Drive, just to send a link to a client so they can wait another 10 minutes to download it. I wanted to build something that felt like magic. Teleport skips the server entirely. It uses WebRTC Data Channels to establish a secure,...

TeleportZero-cloud, hardware-accelerated P2P file transfer.
Stop waiting for cloud uploads. Teleport is a lightning-fast, end-to-end encrypted file dropper built on WebRTC. No file size limits, no server storage, and a buttery-smooth 60fps spatial UI. Drop a file, share the 6-digit code, and beam it directly to your peer.

TeleportZero-cloud, hardware-accelerated P2P file transfer.
