We've also been working on improving how we execute JS today for webpack loaders and babel plugins. The folks from github.com/utooland/utoo have contributed a thread-based implementation for Node.js workers that will replace our current system that uses a pool of child processes.
@mischnic.bsky.social is working on a solution for collecting global per-route information that should allow us to enable more CSS-in-JS solutions. I can connect you with him or our manager if you want to chat about it.
Understood.
Sadly it'll likely get slightly larger in 16.2 because we enabled LTO, and we'd rather have a faster Turbopack than a smaller Turbopack. github.com/vercel/next....
But... we might replace SWC's WASM runtime for plugins and get some of that space back soon! github.com/vercel/next....
Turbopack is a dev-only dependency, so it shouldn't impact your deployed application's size.
Yeah, it's mostly the Turbopack binary. We haven't done a ton of work to optimize this, as we're prioritizing memory and CPU usage optimizations instead, but we have merged some changes that have helped, e.g. switching away from serde.
github.com/vercel/next....
Finally, here are a few more pictures of the prints! I think they came out great!
Here's the useworkflows.dev logo. It's designed to print in multiple glued-together parts. It can print flat on the bed without supports and you can easily pick different colors for the base and the logo.
www.printables.com/model/163794...
Here's the Next.js logo. It's inspired by the Next.js Conf 2024 website. It comes in a small size and a gigantic 1ft (298mm) tall version!
I printed the large version over the course of a few days. It uses nearly a full 1kg roll of filament.
www.printables.com/model/163778...
I've created easy-to-print 3D models of the blocky Next.js Logo from the Next.js Conf 2024 website, and the "𝚞𝚜𝚎 𝚠𝚘𝚛𝚔𝚏𝚕𝚘𝚠" logo, with detailed instructions!
Next 15 is already pretty behind when it comes to Turbopack. We shipped a file system cache for dev in Next 16.1 that helps a lot. nextjs.org/blog/next-16...
Without agents, I likely wouldn't have made the effort to report the bug.
This worked so well because the task was easy but tedious (delete code) and trivially verifiable (compile, run, check process exit code).
Some of the best LLM uses don't involve code generation.
I found a Rust miscompilation, but the repro was inside Turbopack (huge, many crates).
It would've taken hours to minimize by hand, but I used an agent to prune code/dependencies until the repro was just 34 lines.
github.com/rust-lang/ru...
Great write-up from @sentry.io! They now support Turbopack, and in the process they were able to simplify their implementation and reduce their build time overhead! blog.sentry.io/turbopack-su...
Wow, this is really cool! Many apps spend a lot of time doing static site generation in Next.js, and need incremental solutions if we want to solve that.
I'm very interested in seeing all the places you end up using picante.
I shipped a blog post!
Variance has got to be one of the least intuitive aspects of most type systems.
Keep an eye on where the server is located, and where in the world you and your friends are. Network latency is important for games like Minecraft.
There's a whole section on this in The Little Book of Rust Macros. Your solution will form an AST that's inefficient to process. lukaswirth.dev/tlborm/decl-...
We've been dogfooding and iterating on file system caching for over a year. We didn't want to release it until it met our own high quality bar. We're excited to announce that it's now stable and on-by-default for 𝚗𝚎𝚡𝚝 𝚍𝚎𝚟.
Glad to hear it. The offer still stands to dive into the traces!
There is critical vulnerability in React Server Components disclosed as CVE-2025-55182 that impacts React 19 and frameworks that use it.
A fix has been published in React versions 19.0.1, 19.1.2, and 19.2.1. We recommend upgrading immediately.
react.dev/blog/2025/12...
I'm happy to dive into traces with you over DMs. Often there's some minor things that can be tweaked to improve performance.
nextjs.org/docs/app/api...
No, we've discussed it, but:
- We think the filesystem cache is a better all-around solution.
- If you have a large app, which routes should we compile?
- We're keeping an eye on memory usage, and with our current architecture, this would likely waste more memory than most users would want.
The filesystem cache is available in beta on Next 16.0, and we're hoping to ship it as stable and on-by-default very soon (we've been working on it for over a year!). nextjs.org/blog/next-16...
Regarding compilation upon navigation: It should be better in Next 16, but we are working on a couple things to address this:
- A file-system cache that persists across restarts of `next dev`.
- An alternate bundling mode that shares more work across pages, at the cost of a slower first page load.
@runofthemillgeek.com Happy to help dig in, especially with traces, if that document doesn't solve it for you. My DMs are open.
Happy Halloween!
Local microfrontends support is a big deal!
You can see small spikes in October months around Next.js Conf 👀
I'm excited about the new release, and I'm looking forward to seeing the community tomorrow at Next.js Conf!
Build time cut in half after upgrading.
• next@15.4 - 1m 33s cold
• next@16-beta - 45s cold