WebStreams are a nice, ergonomic standard for streaming data. And they're fast too... right? Turns out: not always. On servers, we found they're up to 10x slower than Node.js streams. So we Ralph Wiggum'd WebStreams for the server, powered by Node internals ↓ https://t.co/YwjsvJozNH
Vercel Uses AI to Rewrite WebStreams for a 14.6x Server-Side Performance Gain
· Updated
Vercel used AI to autonomously produce a new WebStream implementation optimized for server-side Node.js. The rewrite removes browser-specific overhead baked into the original spec port, delivering up to 14.6x faster performance with 1,100 of 1,116 Web Platform Tests passing. The implementation was written autonomously - AI planned, wrote, and validated the code without human-driven iteration.
This benefits virtually all server-side WebStream use cases, including React Server Components and Next.js streaming workloads. The upstream path is: npm package → Node.js merge → Node.js release → Vercel fluid runtime update. Once merged into Node.js, the gain extends to self-hosted deployments too, not just Vercel infrastructure.
If your apps use Node.js streaming, this performance gain lands without any code changes on your end. Vercel users get it automatically via fluid runtime; self-hosted Node.js deployments benefit once the upstream PR merges. If streaming latency matters to your stack, it's worth tracking.
Vercel
@vercel
22retweets
View on X





