how it fells when you know how to use coding tools like claude code

awesome github project that can help to get better with claude code github.com/luongnv89/c...

I think once your app has charts, you need to add visual regression tests; otherwise, it's hard to test.

You can ask Claude to list available skills. You'll also notice it references your custom slash commands as skills. The Anthropic team changed this to make commands easier to auto-discover. The downside is that if your project now has many commands with descriptions, it may overwhelm your context.

will publish a blog post later that explains exactly the difference between claude md prompts skills and subagents . Now time to go to the gym and work on my humanoid body

Who do I need to follow when it comes to Cloud code? I want to get better at not writing any code and only using Cloud code to build apps.

can recommend 8/10. Since I am a huge fan of detective stories (detective conan Sherlock holmes)

MCPS is a good example of how quickly best practices in the agent world evolve. Probably even skills will become obsolete in a few months, and we'll have something better.

good video www.youtube.com/watch?v=olv...

claude code now finally has lsp support. so no need for me to switch to opencode

in my humble opinion, this is the best testing strategy for Vue projects. I have written tests for Vue projects for over 7 years, and all my experiences led up to this pattern. alexop.dev/posts/vue3_...

In a few years, we won't be writing code anymore. All the roles (UX, UI, BA, dev, Qa) will merge into one. Teams will shrink and mostly delegate work to AI agents. The only way to prepare for this is to become a generalist. You need to be able to build your own software with AI coding tools.

If you feel stuck with your career as a developer, watch this.https://www.youtube.com/watch?v=IMHneaMO-dg

When using AI agents capable of creating subagents, like Claude Code, you can prompt them to devise a plan to accelerate progress.

who needs figma when you have claude code

Every company should use something like Code Rabbit. It's better at code reviews than most devs. It will also read your style guide and can learn things

You can also stack slash commands in Claude code. First will push my current changes and create good commit messages. Second will create a PR with a good description.

I wonder why mcps got so much hype from all influencers but not skills

if claude code will do that at the end of a session that you have a good setup

must watch for anyone that is using claude code www.youtube.com/watch?v=X2c...

I copied screens from my workout tracking app to Google Gemini and asked it to generate an image of how the design system would look.

I don't think that any company needs a separate QA role. It should be the responsibility of devs to test the code that they write.

I will give a talk at the Vue.js Talks Online Meetup. I will share everything I learned while building three projects using only Claude Code for Vue.

I believe this is the difference with Opus 4.5. These models are now smart enough to fetch and read documentation when they run out of their own training data.

This was an annoying fix. It took Claude Code a while to realize that every time it ran lint, the linter automatically made changes.

This is how you do code reviews in the age of genAi

It's crazy how different the mindset of USA software companies and Germans is...

Google Gemini is a good designer. I gave it the current design of my workout app that is only optimized for mobile at the moment. And I asked it to convert it to desktop.

The Netflix-Warner deal is crazy. I hope they will still release the movies in cinemas.

damn opus 4.5 is funny and has personality . I wanted to understand if it was using a subagent or how the parallel update did work for a eslint fix.

When you see that someone used your own blog post as a resource while you were working on a task, haha.

Who wants to start a podcast with me next year? We will just complain about Scrum all day.

when I am the lead of a project

This is the difference between testing with JSDOM and a real browser: I think using JSDOM or Happy DOM for testing should now be considered an anti-pattern.

It's crazy how fast I can work with claude code. Big refactorings only take a few minutes.

this is how it feels when you use claude code www.youtube.com/watch?v=JKo...

the power of subagents

Use the standard Node.js "imports" field instead of TypeScript aliases:

This is how you can do a code review of the quality of your Vue components in Claude Code. You just need to define a subagent and tell it what a good Vue component is made of.

using tdd with claude code is so nice. I dont even have to test if the functiolity works if claude code is doing tdd

I wonder if this will be the future of software teams, where most roles are just AI agents and one person is managing them all.

One trick to encourage Claude to use the skills that are available is a hook that, at the start of a session, will give him a hint about which skills are available.

With the Google Notebook LLM, you can convert any text blog post into a good video. the quality is crazy good. Probably in the future when it comes to learning people will generate their own learning material on the fly with the help of ai.

Best ESLint rule for any project is to not allow `else` or `else if` and always prefer early return or ternary operator.

With this custom hook, I forbid Claude code to use else or if-else. I am a huge fan of early returns or ternary operators.

I think it will only be a matter of time until Cursor and Copilot also add hooks and skills from cloud code into their toolset.

did I already said that skills are awesome . I basically converted most of my blog posts into skills and now claude code writes code like me.

you vs the guy not to worry about

Wrote a blog post on how you can customize your cloud code experience by enforcing a specific workflow like TDD with skills and subagents.

I think in the future, the need for manual QA will not be so big in projects. AI agents are quite good at doing that. You can just give them a well-crafted ticket, and they can use the Google Cloud Platform server to do manual QA themselves.