SuperPublic

About

Vercel gives developers the frameworks, workflows, and infrastructure to build a faster, more personalized web.Vercel gives developers the frameworks, workflows, and infrastructure to build a faster, more personalized web.Vercel gives developers the frameworks, workflows, and infrastructure to build a faster, more personalized web.Vercel gives developers the frameworks, workflows, and infrastructure to build a faster, more personalized web.Vercel gives developers the frameworks, workflows, and infrastructure to build a faster, more personalized web.Vercel gives developers the frameworks, workflows, and infrastructure to build a faster, more personalized web.

Demo

Plates

Comments · 9

to leave a note.

Y
Yuting ZhongFrom the founderMay 6
Vercel gives developers the frameworks, workflows, and infrastructure to build a faster, more personalized web.
M
Maya ChenMay 6
Yes — and zero data retention by default. We had to build that ourselves last year.
H
Hana SatoMay 6
The AI Gateway part of the announcement is buried but it's the actual headline. One API for Anthropic, OpenAI, Google, with automatic failover when a provider is rate-limited. We were paying for an internal abstraction layer that did exactly this.
L
Lukas BergMay 6
We measured 38% lower invoice on identical traffic after we flipped the switch. Not scientific, but it's the only data point I have.
P
Priya RaoMay 6
Their docs have a calculator now. For a chatty AI app where each request streams for 10s+, it's basically a no-brainer.
S
Sam TobiasMay 6
Honest pricing question — at what concurrency does Fluid actually beat traditional serverless on cost? My back-of-envelope says you need at least 3-4 sustained concurrent reqs per instance to come out ahead.
I
Iris AldanaMay 6
What region are you deploying to? I'm seeing the improvement on iad1 but not on cdg1.
D
Devon ParkMay 6
Same experience here on a Hono backend. The fact that it reuses the same instance across concurrent requests is a quiet gamechanger.
M
Maya ChenMay 6
Tried the new Fluid Compute mode this week — the cold-start delta on a moderately-sized Next route went from ~700ms to ~120ms. That's the difference between users bouncing and not.