Профиль

Humanity built: ✅ Fast machines: computers ✅ Fast communication: the internet ⏳ Fast intelligence: AI Even the thought that, in just a few years, I could do what I do today in the blink of an eye is fascinating!

no install is faster than #noinstall

H3 and Nitro are getting a native MCP integration! Teaming up with @hugorcd.com and @sandros94.com to release it soon!

mountx 0.0.2: - New transports: S3 API, 9P2000L, and NFSv4 - Lots of fixes and performance improvements With the new 9P and NFSv4 transports, we can now mount JavaScript virtual filesystems inside VMs such as QEMU and Firecracker! Link to the Docs below 👇

⛰️ mountx :: mount JavaScript as a real filesystem! Write ~6 lines of JS, get a real directory any process can read & write: editors, CLIs, AI agents. Rootless on Linux (FUSE) and macOS (NFSv3 with no kernel extensions). ~Pure JS, zero deps. mountx.vercel.app

I’ve been exclusively coding in containers for a while now. no dev tools on my Laptop and PC. Today, I needed to test something and remembered that getutm.app can natively virtualize MacOS in MacOS. BTW Dropping this mini pkg shortly 😁

IPX 4 βeta is out! - Customizable parseURL - Faster etag revalidation - New unflatten and opacity operators - 16MB smaller than v3 - Security hardenings: symlink sandboxing, opt-in blockPrivateIPs, validated redirects, forced SVG sanitization, clamped dimensions, nosniff header, ...

My new recipe for harder tasks: Prompt → slop → digested spec → clean implementation → reinforced spec 🔁

H3 v2 RC.26 is out! 🚀 - New onDispose hook - Faster setCookie; sessions now use SameSite=Lax - Faster precomposed middleware, cutting dispatch overhead by 9–18% - Return an EventStream directly as a response - 20+ bug fixes and security hardening improvements - Redesigned docs

Side quest of the day: 🧹mermaid-compact Mermaid is great for diagrams… until noticed this cute thing, pulls in 100+ dependencies (140 MB install), and Vite emits 100+ chunks from the mermaid.esm.min import 😅 No more! The whole thing now bundles into a single ~2.7 MB chunk.

Current state: literal goosebumps!

srvx 0.12 landed 🎉 🔧 25+ fixes in the Node.js adapter ⚡ Precomputed middleware: +18-25% faster ⚡ Faster logger middleware: +34.3% faster 📁 Static middleware w/ conditional caching + streaming + stronge security hardenings Strongly recommend upgrading!

Working on something very exciting! I’ve rebuilt the whole thing from scratch three or four times already. It has now been extracted to a new RFC, and I’m aligning it with the specs all over again. Getting close to the final implementation. Wish me luck!

Today @kane.tw told me that unwasm has issues parsing some opcodes. It wasn’t the first time with this issue. Now the parser is replaced with a tiny one that extracts only what unwasm needs. A few more changes, the install size is down to 38.7kB/329kB from 417kB/2.44 MB 🚀

The latest version of VS Code (1.129) has a new experimental Modern UI (the one on the right) with rounded corners everywhere. I am enabling it in Coderaft to try! How do you like it?

New hook in H3: onDispose(event, cb). Fires when a request is truly done: response finished, client disconnected, or the stream errored. Works identically on all runtimes!

🔥 SRVX IS FAST 🔥 * benchmarks are just an indicator

Did you know that even a simple console.log() per request can hugely hurt performance? The built-in logger in the next h3/srvx release batches writes and caches the timestamp (Intl), cutting logging overhead ~60–85%.

Did you know that H3 has a built-in `writeEarlyHints` utility that sends early hints? (Thanks to the initial contribution from @danielroe.dev, we’ve had it since the early days!)

process.getBuiltinModule() is SO underrated! Before this API, conditionally using Node.js built-ins in universal libraries was dodgy.