Met the most wonderful person in the world 🤍

`tsnapi` - a snapshot testing utility for the public API for library maintainers. Snapshots for both JavaScript and TypeScript declarations. It would help prevent unintended breaking changes on public API signatures.

Thanks @sxzz.dev for the initial idea and implementation.

github.com/antfu/tsnapi

After years of working in silos, in 2024 @netlify.com, @cloudflare.social, OpenNext, Google @firebase.com, and the @nextjs.org team at @vercel.com built bridges.

We started discussing how to truly solve for full portability of Next.js. And we kept at it.

Today marks a culmination of this work.

It has been a very long while since I was posting on social media. I kinda want to get back into it and share more, but I feel like I've lost the momentum. I'll try to get it back on track.

GitHub - afonsojramos/slidev-decks: CLI companion for managing multiple Slidev presentations CLI companion for managing multiple Slidev presentations - afonsojramos/slidev-decks

Every Slidev power user ends up writing a custom picker script for their talks repo. @antfu.me has one. I had one. So I packaged it properly.

slidev-decks: TUI deck picker, fuzzy matching, scaffold wizard, build --all, pm detection, slidev CLI pass-through.

github.com/afonsojramos/slidev-decks

Thanks for sharing! I am glad people are using and enjoying it! I am definitely up for including resources in the readme; feel free to PR! Thanks!

test(reporters): add outputFile tests by AriPerkkio · Pull Request #978 · vitest-dev/vitest Adds missing test cases for outputFile configuration. I'm working on another bug fix and thought let's merge these ones first.

This day 4 years ago I sent my first PR to @vitest.dev. Back then it was on v0.7.4 with 80K weekly downloads. After one year and +70 more PRs, @antfu.me and @erus.dev invited me to join the team.
Today we are at v4.1.0 with 35M weekly downloads. Maybe in 2030 we'll have Vitest 8 with 140M downloads.

simply, more OSS projects should DO THIS.

OSS is a long playing game, if not forever. but we are people behind those projects, and we need good rest and balance to keep things sustainable. I am super happy to see this move and looking forward to see how it could change how OSS works for all of us.

I am trying both Cursor and Claude Code

GitHub - antfu/skills: Anthony Fu's curated collection of agent skills. Anthony Fu's curated collection of agent skills. Contribute to antfu/skills development by creating an account on GitHub.

late to the party. I have finally been convinced by multiple awesome developers to give agents another try.

this is my first premature contribution:
github.com/antfu/skills

feat: add official skills by antfu · Pull Request #2447 · slidevjs/slidev Add generated skills from the docs with Claude Code To give it a try: npx add-skill slidevjs/slidev

👀

if people no longer read the code, API is no longer targeting humans, is there still a need for "good API design"? is it still worth the effort to figure out what would be the best for users, instead of the best of ai to understand/use? I don't know.

for a little bit of context, I build tools for humans for better dx or productivity, and I consider it a creative work, so I am not worried at all about "being replaced by ai"

what I am concerned is the tools' target users. if people no longer care about user-experience but more on "ai-experience"

🫂

do people still care?

is it still relevant?

License This project is not licensed under an open-source license and is the intellectual property of Tailwind Labs Inc. The source is available only as an educational resource and to accept fixes for minor mistakes.

TIL that the tailwinds docs are not open source. It is the first time I see a popular OSS project with this kind of non-license approach for their docs. Are there other examples out there? Maybe I missed them too assuming that docs followed the same license as their code.

Oxlint Integration Plan · Issue #767 · antfu/eslint-config I have received several requests asking to support oxlint in this config. Creating this issue as the epic issue to track the progress and have a single place for discussion. The most important thin...

Check this issue
github.com/antfu/eslint...

Yes, for a while longer, I am sorry about that:
bsky.app/profile/antf...

😇❓

I am trying, but everytime they do it with different account 😇

Don't even know if I should build some LLM-powered tools to fight back magic with magic, or if I should just quit altogether.

Agents create the illusion that you can accomplish complicated work with simple prompts, and people become less appreciative of the hard work behind the scenes.

I feel bad that I have to enable this for @shiki.style, as I am getting AI-bot-accounts creating the same issues again and again, and sending PRs trying to "fix" the same thing. I don't know if that repo has been somehow labeled as "AI playground" or something, but I can't keep bearing with it.

I certainly don't want to act rudely or terribly to any human contributors and make the environment worse, but I also don't want to waste my time being nice and supportive to AI bots.

I feel the dilemma is, that I feel responsible for being friendly and welcoming to make the experience nicer for newcomers to OSS, who may not yet know the manner well.

But with AI messing this up, it's really hard to tell if it's a human behind I should take care or just AI doing terrible work.

a GitHub PR titled `fix(adobe): handle provider configuration objects properly` with the following text: 🐛 Bug Fix Potentially fixes Adobe font-face declaration generation regression introduced in v0.12.1. Closes #736 ⚠️ Note: This fix was developed with AI assistance This PR was blindly developed by AI, without me having deep any knowledge of @nuxt/fonts internals. While it's (claimed by Claude to) been tested and appears (again, according to Claude) to work correctly, a thorough review by @danielroe or someone familiar with the fontless/unifont architecture would be appreciated.

I actually love this disclaimer

I am burning out on this...

Multiple PRs created by AI-bot accounts are trying to solve the same issue that has not yet even been identified, with verbose plain-text PR descriptions.

I don't even know if I should bother to reply...

Any ways to stop this? 😇

npm downloads chart of pacakge @antfu/eslint-config, generated by https://npm.chart.dev/@antfu/eslint-config

Didn't even think that my personal ESLint config deserves so many downloads, 1M/mo 🫨. Thank you all!

That's an interesting approach. To me personally, I don't think I would restrict anyone from using any tools they think would be helpful, including agents or anything, as long as they do a self-review. More like I want to see opinions from people, rather than the LLM's results as-is.

I want to think that teaching OSS etiquette as part of regular education would help devs empathize with maintainers and understand when a PR is actually helpful. GitHub could guide new users too. But spam will likely worsen significantly. Platforms need to find ways to remain usable in the long term