Pooya Parsa
@pi0.io
Building opensource stuff ◦ @unjs.io ◦ @nitro.build ◦ @nuxt.com
🌱 github.com/pi0 ❌ x.com/_pi0_
🌱 github.com/pi0 ❌ x.com/_pi0_
September 16, 2026 at 11:13 PM
I remember when yarn was born. It was cool and fast and fancy. then it got more complex up to a point that users couldn't keep up.
Now pnpm is the star. It is getting more complex requirements to be setup "as intended" and honestly i am worried.
sigh... just random thoughts.
Now pnpm is the star. It is getting more complex requirements to be setup "as intended" and honestly i am worried.
sigh... just random thoughts.
September 9, 2026 at 8:55 AM
I remember when yarn was born. It was cool and fast and fancy. then it got more complex up to a point that users couldn't keep up.
Now pnpm is the star. It is getting more complex requirements to be setup "as intended" and honestly i am worried.
sigh... just random thoughts.
Now pnpm is the star. It is getting more complex requirements to be setup "as intended" and honestly i am worried.
sigh... just random thoughts.
📦 jup 0.5.6
Now respects the pnpm version set in pnpm-lock.yaml with devEngines ranges like pnpm@^12
Also fixes IPC channels for Bun and Deno.
Getting closer to mass adoption 🤞
Now respects the pnpm version set in pnpm-lock.yaml with devEngines ranges like pnpm@^12
Also fixes IPC channels for Bun and Deno.
Getting closer to mass adoption 🤞
September 9, 2026 at 8:47 AM
📦 jup 0.5.6
Now respects the pnpm version set in pnpm-lock.yaml with devEngines ranges like pnpm@^12
Also fixes IPC channels for Bun and Deno.
Getting closer to mass adoption 🤞
Now respects the pnpm version set in pnpm-lock.yaml with devEngines ranges like pnpm@^12
Also fixes IPC channels for Bun and Deno.
Getting closer to mass adoption 🤞
There’s a Persian saying:
"The potter drinks water from a broken jug."
I use changelogen for literally all releases i make, yet it has had bugs over a year, but I’ve never found the time to properly fix them.
Spending part of this evening on it!
"The potter drinks water from a broken jug."
I use changelogen for literally all releases i make, yet it has had bugs over a year, but I’ve never found the time to properly fix them.
Spending part of this evening on it!
September 4, 2026 at 4:27 PM
There’s a Persian saying:
"The potter drinks water from a broken jug."
I use changelogen for literally all releases i make, yet it has had bugs over a year, but I’ve never found the time to properly fix them.
Spending part of this evening on it!
"The potter drinks water from a broken jug."
I use changelogen for literally all releases i make, yet it has had bugs over a year, but I’ve never found the time to properly fix them.
Spending part of this evening on it!
🚀 Nitro v3.0.260903 is here!
🧭 Smarter, Safer and Faster Route Rules
💾 Powerful Caching Utils
🗄️ New DB Connectors and ORM Integrations
🔭 Native Vercel & Cloudflare Tracing
▲ Vercel Deploys with Immutable Assets
📦 Smart prompt to install deps
🔒 Strong Security
🧭 Smarter, Safer and Faster Route Rules
💾 Powerful Caching Utils
🗄️ New DB Connectors and ORM Integrations
🔭 Native Vercel & Cloudflare Tracing
▲ Vercel Deploys with Immutable Assets
📦 Smart prompt to install deps
🔒 Strong Security
September 3, 2026 at 11:16 PM
🚀 Nitro v3.0.260903 is here!
🧭 Smarter, Safer and Faster Route Rules
💾 Powerful Caching Utils
🗄️ New DB Connectors and ORM Integrations
🔭 Native Vercel & Cloudflare Tracing
▲ Vercel Deploys with Immutable Assets
📦 Smart prompt to install deps
🔒 Strong Security
🧭 Smarter, Safer and Faster Route Rules
💾 Powerful Caching Utils
🗄️ New DB Connectors and ORM Integrations
🔭 Native Vercel & Cloudflare Tracing
▲ Vercel Deploys with Immutable Assets
📦 Smart prompt to install deps
🔒 Strong Security
I tried SIEVE a long time ago, handwritten in rou3. It wasn't that impressive, but now, for H3's route rules engine, it is effective!
September 3, 2026 at 10:28 AM
I tried SIEVE a long time ago, handwritten in rou3. It wasn't that impressive, but now, for H3's route rules engine, it is effective!
New @unjs.io drop 🍕 jup:
It provides shims for package managers and runtimes and automatically selects the right one from package.json.
It is a full Corepack rewrite, but smaller, faster, and nicer.
It supports ^ ranges, pnpm 12, Bun, Deno, Node.js, Aube, Nub.
Give it a try!
It provides shims for package managers and runtimes and automatically selects the right one from package.json.
It is a full Corepack rewrite, but smaller, faster, and nicer.
It supports ^ ranges, pnpm 12, Bun, Deno, Node.js, Aube, Nub.
Give it a try!
September 3, 2026 at 8:25 AM
New @unjs.io drop 🍕 jup:
It provides shims for package managers and runtimes and automatically selects the right one from package.json.
It is a full Corepack rewrite, but smaller, faster, and nicer.
It supports ^ ranges, pnpm 12, Bun, Deno, Node.js, Aube, Nub.
Give it a try!
It provides shims for package managers and runtimes and automatically selects the right one from package.json.
It is a full Corepack rewrite, but smaller, faster, and nicer.
It supports ^ ranges, pnpm 12, Bun, Deno, Node.js, Aube, Nub.
Give it a try!
🎊 srvx has reached v1.0 🎊
Two years ago, I went on an ambitious quest.
My dream was to make export default { fetch } a universal consistent format for servers.
srvx has been adopted and battle-tested by Nitro, H3, Elysia, TanStack, and many others projects 🖤
Two years ago, I went on an ambitious quest.
My dream was to make export default { fetch } a universal consistent format for servers.
srvx has been adopted and battle-tested by Nitro, H3, Elysia, TanStack, and many others projects 🖤
September 2, 2026 at 8:29 AM
🎊 srvx has reached v1.0 🎊
Two years ago, I went on an ambitious quest.
My dream was to make export default { fetch } a universal consistent format for servers.
srvx has been adopted and battle-tested by Nitro, H3, Elysia, TanStack, and many others projects 🖤
Two years ago, I went on an ambitious quest.
My dream was to make export default { fetch } a universal consistent format for servers.
srvx has been adopted and battle-tested by Nitro, H3, Elysia, TanStack, and many others projects 🖤
🏎️ perf tip of the day:
If your ESM package uses "type": "module" and ships .js files, Node.js must find, read, and parse package.json to determine their format.
Shipping .mjs avoids all this runtime cost.
That’s why UnJS toolchain (obuild) has always shipped .mjs.
If your ESM package uses "type": "module" and ships .js files, Node.js must find, read, and parse package.json to determine their format.
Shipping .mjs avoids all this runtime cost.
That’s why UnJS toolchain (obuild) has always shipped .mjs.
August 28, 2026 at 10:25 AM
🏎️ perf tip of the day:
If your ESM package uses "type": "module" and ships .js files, Node.js must find, read, and parse package.json to determine their format.
Shipping .mjs avoids all this runtime cost.
That’s why UnJS toolchain (obuild) has always shipped .mjs.
If your ESM package uses "type": "module" and ships .js files, Node.js must find, read, and parse package.json to determine their format.
Shipping .mjs avoids all this runtime cost.
That’s why UnJS toolchain (obuild) has always shipped .mjs.
Best vacation ever!
I’ve been building lots of things lately, and I should start releasing them soon...
The final steps always take A LOT longer than the building itself.
Releasing at least one cool project today 😁
I’ve been building lots of things lately, and I should start releasing them soon...
The final steps always take A LOT longer than the building itself.
Releasing at least one cool project today 😁
August 28, 2026 at 10:13 AM
Best vacation ever!
I’ve been building lots of things lately, and I should start releasing them soon...
The final steps always take A LOT longer than the building itself.
Releasing at least one cool project today 😁
I’ve been building lots of things lately, and I should start releasing them soon...
The final steps always take A LOT longer than the building itself.
Releasing at least one cool project today 😁
On a road trip in beautiful Tuscany, steering a swarm of agents from my phone to build a new UnJS lib.
What a time to be alive!
What a time to be alive!
August 22, 2026 at 9:06 AM
On a road trip in beautiful Tuscany, steering a swarm of agents from my phone to build a new UnJS lib.
What a time to be alive!
What a time to be alive!
⛁ unjs/db0 0.4:
- Neon serverless posgress driver (@atila.io)
- Exposed database .capabilities (@onmax.me)
- Built-in tracing channels (@awad.dev)
- Drizzle v1.rc support
- Prisma adapter (Inesh Bose)
- No more peer dependencies. Nitro will smartly prompt to install them
- Neon serverless posgress driver (@atila.io)
- Exposed database .capabilities (@onmax.me)
- Built-in tracing channels (@awad.dev)
- Drizzle v1.rc support
- Prisma adapter (Inesh Bose)
- No more peer dependencies. Nitro will smartly prompt to install them
Release v0.4.0 · unjs/db0
compare changes
🚀 Enhancements
Expose connector name (#205)
Add neon for serverless postgres (#191)
Add database capabilities (#207)
Add prisma adapter (#210)
Add tracing channels support (#193)
d...
github.com
August 20, 2026 at 10:05 PM
h3 rules also has a built-in compiler that generates an optimized entry, ready to bundle!
August 19, 2026 at 10:58 AM
h3 rules also has a built-in compiler that generates an optimized entry, ready to bundle!
H3 v2 rc.27:
- Route Rules Engine, heavily audited
- Session idleTimeout for sliding expiration
- JSON-RPC improvements (mcp is comming!)
- 18 bugs and security hardenings found by deepsec (pi+opus)
- Route Rules Engine, heavily audited
- Session idleTimeout for sliding expiration
- JSON-RPC improvements (mcp is comming!)
- 18 bugs and security hardenings found by deepsec (pi+opus)
August 19, 2026 at 9:31 AM
H3 v2 rc.27:
- Route Rules Engine, heavily audited
- Session idleTimeout for sliding expiration
- JSON-RPC improvements (mcp is comming!)
- 18 bugs and security hardenings found by deepsec (pi+opus)
- Route Rules Engine, heavily audited
- Session idleTimeout for sliding expiration
- JSON-RPC improvements (mcp is comming!)
- 18 bugs and security hardenings found by deepsec (pi+opus)
Randomly saw H3 Core (that Nitro bundles) is 3,140 bytes mingz, what a picidence 😆
August 19, 2026 at 8:46 AM
Randomly saw H3 Core (that Nitro bundles) is 3,140 bytes mingz, what a picidence 😆
🚀 unjs/ocache v0.3 is out! It was a big milestone.
Faster. Safer. Binary Blob Storage, Streamed Cache Fills, Composable storage, and more! 🚀
(soon available in H3 and Nitro, can be used directly anywhere with anything even with Hono!)
Faster. Safer. Binary Blob Storage, Streamed Cache Fills, Composable storage, and more! 🚀
(soon available in H3 and Nitro, can be used directly anywhere with anything even with Hono!)
August 18, 2026 at 3:14 PM
🚀 unjs/ocache v0.3 is out! It was a big milestone.
Faster. Safer. Binary Blob Storage, Streamed Cache Fills, Composable storage, and more! 🚀
(soon available in H3 and Nitro, can be used directly anywhere with anything even with Hono!)
Faster. Safer. Binary Blob Storage, Streamed Cache Fills, Composable storage, and more! 🚀
(soon available in H3 and Nitro, can be used directly anywhere with anything even with Hono!)
TIL Uint8Array.fromBase64 exists in browsers!
August 18, 2026 at 9:31 AM
TIL Uint8Array.fromBase64 exists in browsers!
unjs/ohash 2.0.12
- Serializer up to 8% faster (thanks to @zsilbi.hu)
- Diff ~3x faster for field changes
- Locale-dependent issues resolved
@zsilbi.hu has also been working on a couple more performance boosts coming in the next major version 🔥.
- Serializer up to 8% faster (thanks to @zsilbi.hu)
- Diff ~3x faster for field changes
- Locale-dependent issues resolved
@zsilbi.hu has also been working on a couple more performance boosts coming in the next major version 🔥.
August 14, 2026 at 12:34 PM
unjs/ohash 2.0.12
- Serializer up to 8% faster (thanks to @zsilbi.hu)
- Diff ~3x faster for field changes
- Locale-dependent issues resolved
@zsilbi.hu has also been working on a couple more performance boosts coming in the next major version 🔥.
- Serializer up to 8% faster (thanks to @zsilbi.hu)
- Diff ~3x faster for field changes
- Locale-dependent issues resolved
@zsilbi.hu has also been working on a couple more performance boosts coming in the next major version 🔥.
md4x 0.27 is fully ported from C to Zig.
Faster, and more memory-efficient.
It also has a new /standalone portable build.
A fast, spec-compliant Markdown parser, healer (for AI streams), and renderer (HTML, Text, CLI, and JSON) + a full YAML parser in a ~114 kB gz bundle.
Faster, and more memory-efficient.
It also has a new /standalone portable build.
A fast, spec-compliant Markdown parser, healer (for AI streams), and renderer (HTML, Text, CLI, and JSON) + a full YAML parser in a ~114 kB gz bundle.
August 14, 2026 at 9:14 AM
md4x 0.27 is fully ported from C to Zig.
Faster, and more memory-efficient.
It also has a new /standalone portable build.
A fast, spec-compliant Markdown parser, healer (for AI streams), and renderer (HTML, Text, CLI, and JSON) + a full YAML parser in a ~114 kB gz bundle.
Faster, and more memory-efficient.
It also has a new /standalone portable build.
A fast, spec-compliant Markdown parser, healer (for AI streams), and renderer (HTML, Text, CLI, and JSON) + a full YAML parser in a ~114 kB gz bundle.
Horrible finding: npm downloads ALL OPTIONAL PEER DEPS of packages just to resolve their metadata, then throws all away 🗑️
Fortunately, pnpm and friends don’t do this. But since many users rely on npm, we need to find work around in UnJS/Nitro packages...
Fortunately, pnpm and friends don’t do this. But since many users rely on npm, we need to find work around in UnJS/Nitro packages...
August 14, 2026 at 7:09 AM
Horrible finding: npm downloads ALL OPTIONAL PEER DEPS of packages just to resolve their metadata, then throws all away 🗑️
Fortunately, pnpm and friends don’t do this. But since many users rely on npm, we need to find work around in UnJS/Nitro packages...
Fortunately, pnpm and friends don’t do this. But since many users rely on npm, we need to find work around in UnJS/Nitro packages...
TIL: fetch spec has a fixed list of HTTP verbs (DELETE/GET/HEAD/OPTIONS/POST/PUT) to be normalized
...StAnDaRdS
...StAnDaRdS
August 11, 2026 at 10:29 AM
TIL: fetch spec has a fixed list of HTTP verbs (DELETE/GET/HEAD/OPTIONS/POST/PUT) to be normalized
...StAnDaRdS
...StAnDaRdS
Experimenting with a new library audit workflow.
A script bundles the core logic (treeshaken) into a temporary directory, then spawns multiple agents to audit it. This removes a lot of bias from comments, docs, etc.
Already really helped in H3.
A script bundles the core logic (treeshaken) into a temporary directory, then spawns multiple agents to audit it. This removes a lot of bias from comments, docs, etc.
Already really helped in H3.
August 11, 2026 at 10:04 AM
Experimenting with a new library audit workflow.
A script bundles the core logic (treeshaken) into a temporary directory, then spawns multiple agents to audit it. This removes a lot of bias from comments, docs, etc.
Already really helped in H3.
A script bundles the core logic (treeshaken) into a temporary directory, then spawns multiple agents to audit it. This removes a lot of bias from comments, docs, etc.
Already really helped in H3.
Recently I’ve noticed Claude using strange keywords like “seam” and “chrome.”
It’s getting worse. It writes great code and understands its own language well, but they’re becoming barely human-readable.
I hope it’s just a glitch, but it looks suspiciously like vendor lock-in.
It’s getting worse. It writes great code and understands its own language well, but they’re becoming barely human-readable.
I hope it’s just a glitch, but it looks suspiciously like vendor lock-in.
August 7, 2026 at 8:46 PM
Recently I’ve noticed Claude using strange keywords like “seam” and “chrome.”
It’s getting worse. It writes great code and understands its own language well, but they’re becoming barely human-readable.
I hope it’s just a glitch, but it looks suspiciously like vendor lock-in.
It’s getting worse. It writes great code and understands its own language well, but they’re becoming barely human-readable.
I hope it’s just a glitch, but it looks suspiciously like vendor lock-in.