If you've used Vite, you know tree-shaking. It's fast, it's smart, and it runs on every single build — from scratch — every time. That's the part nobody talks about.
Today I want to introduce something we've been quietly building inside Ionify: packSlimming. It's our architectural answer to what tree-shaking actually costs at scale.
The problem with tree-shaking
Tree-shaking is a per-build operation. Every time your build tool runs, it re-analyzes your entire dependency. That analysis disappears when the process exits. Next run? Start over.
packSlimming: analysis that lives
packSlimming is backed by Ionify's persistent graph and CAS layer. Here's what that means concretely:
From your very first cold start, packSlimming analyzes your deps and stores the result. Every subsequent run — whether it's a dev server restart, a production build, or a CI run on the same branch — reuses that analysis. Your deps optimization is live from day one and never repeated unless something actually changes.
Ionify isn't trying to be the fastest version of what already exists. It's a different model entirely — a build engine that accumulates intelligence instead of discarding it.
pnpm create ionify@latest
⭐ github.com/ionifyjs/ionify
👉 ionify.cloud

Top comments (0)