I tried writing code with Playwright Test Agents, and the experience was quite good. They were able to determine how the page is supposed to behave and generate test scenarios. Although they didn't perfectly reflect how I wanted to test, I could use them as a base for the test cases.

I’ve migrated Naive UI components to Nuxt UI 4 in a project. I used the "plan" feature provided by Cursor 2 with the Nuxt UI MCP server, and it worked pretty well. When a document is well-written and good examples are provided, AI tends to figure out what to do.

I guess I gotta ditch Vercel. Let's move to Netlify.

I've been trying out the GitHub Copilot coding agent. It seems to handle easy refactoring tasks quite well. It's nice that I can work on different tasks while letting Copilot do its job. Let's see if it can handle more difficult tasks.

Zod 4 looks so cool. Performance improvement is always great, but what I like the most about it is that now you can easily define template literals :)

v4.zod.dev/v4

I didn't know that you can use XML to give LLM instructions. I should learn these things more, since understanding these techniques would improve the efficiency of my work

docs.anthropic.com/en/docs/buil...

I just learned that there is "allowImportingTsExtensions" option in TSConfig, which allows us to put the ".ts" extension for import statements. I like it because it can clarify where the types, functions, etc, are coming from.

www.typescriptlang.org/tsconfig/#al...

What I like about writing tests with Playwright is that I can also ensure that the application cares about accessibility. Writing tests becomes difficult if the application doesn't correctly support accessibility.

Sometimes, we need to use a specific version for an NPM package to avoid a known issue. As such, I want to leave a comment package.json, but it's impossible since it's a JSON file at the end of the day. I wonder if there's good practice to achieve the same thing.

I've been using shallowRef by default for a while unless I need deep reactivity. IMO, we should always be aware that ref watches change deeply, so this PR makes sense to me.

I'm reading Engineering Management for the Rest of Us by Sarah Drasner. Although I'm not a manager, I have gained new insights into leading an engineering team. It's not about being smart; it's about gaining trust and sharing values 🙂

Seems like the latest Cursor supports multiple rule files defined in .cursor/rules directory. So, I guess I should use this rather than .cursorrules file?

www.cursor.com/changelog

I've been waiting for Temporal since I was born.

developer.mozilla.org/en-US/blog/j...

This is amazing!
TypeScript can finally narrow types correctly when the return type needs to be decided based on the code path.
youtu.be/lUl6gtytG64?...

I agree with the conclusion. GitHub Copilot is improving, but Cursor is still better. I love that Cursor gives me suggestions even before I ask, and all I have to do is just press the tab keys multiple times.
www.builder.io/blog/cursor-...

Recently, I've been building e2e tests with Playwright, Drizzle, and Cursor. The experience is so good since I can quickly generate the Drizzle schema, which makes it easy to set up the test data. Plus, I can write test code easily since Cursor almost always figures out what I want to test next.