second: devs tend to fixate on repetitiveness as a sign of poor code quality. but the cure is often worse than the disease, because if unchecked it leads to the thing that, in my experience, really _is_ a sign of poor code quality: indirection. when you're writing code, indirection is your nemesis.

bsky.app/profile/fili...

Test repository: github.com/filipsobol/b...

Looks like Rspack 1.5.0-rc.0 improves build speed by another 2.52%, bringing the total improvement to 18.19% compared to 10 months ago.

Rspack: 1.2.0-alpha.0 => 1.4.11
esbuid: 0.24.0 => 0.25.9
Parcel: 2.13.2 => 2.15.4
Rolldown: 0.15.1 => 1.0.0-beta.34
Rollup: 4.27.4 => 4.48.1
Vite: 6.0.3 => 7.1.3
Webpack: 5.96.1 => 5.101.3

Here are the charts with absolute numbers👇

Rollup / @vite.dev also saw improvements:

Bundle size improved by 1.48% (-31 KiB).
Build speed improved in Vite by 1.5%.

Solid but incremental progress.

Meanwhile, @rspack.dev made huge gains in both areas.

Build speed improved by 15.7%.
Bundle size also improved by 0.91%.

For developers, upgrading is basically a win-win.

But that gain came at a cost.

Build speed with Rolldown + oxc dropped by 9.5%.

It moves further away from esbuild (the speed leader), but still remains much faster than most other bundlers.

The big story: @rolldown.rs + built-in oxc minifier.

After 10 months the bundle size improved by 18% (-463 KiB).

This result not only outperforms the recommended Rolldown + esbuild combo, but also beats more mature bundlers *in this specific scenario*.