Moments

After enabling the system proxy in the VPN, local test domain names (such as *.test) suddenly can't be accessed?

Wrong move: Add a DIRECT (direct connection) rule in the proxy software.

This is useless! Because once traffic enters the proxy software, it will use public DNS to look up this local domain name. If it can't find the IP, it will immediately report an error and block it.

Correct approach: In the operating system's "Network Settings -> Proxy Bypass List (Bypass)", add *.test (or configure the Bypass bypass list in the proxy software).

Core principle:

Traffic is first received by the operating system.

Bypass (bypass): Traffic never enters the proxy software; the operating system handles it itself and uses local DNS to access directly.

DIRECT (direct connection): Traffic has already entered the proxy software; the proxy software attempts a direct connection, but usually fails at DNS resolution.

I've encountered this problem before and have always been able to solve it, but today is the first time I've fully understood it.

“It feels great to manually participate in coding, to not let AI write code for me.”

I just tried the new global Cursor rules. Thinking about functionality myself, focusing on code organization and quality, the feeling of full engagement is back.

In order to be able to participate in the process of code organization and writing, I added a few global rules to cursor:

1. Help me set up the global cursor rules, add one: just provide the plan, steps, or implementation code for each step, do not write the code for me.

2. Do not include Co-authored-by in commit messages.

3. Another rule: write commit msg for changes, do not commit. i will commit manually.

Today I added this site to bing and yandex.

Next, we can observe and familiarize ourselves with the features and characteristics of these search platforms.

As the website content increases, the Nextjs SSG build time also gets longer.

The previous consideration was: pre-render pages, use less SSR, to make pages load faster. Recently when deploying, I found that the waiting time has become longer, mainly stuck at the Nextjs build step.

I asked the AI, and based on the problem and website scenario, the AI helped me analyze it. After seeing the AI's response and thoughts, I removed SSG and only use SSR. This goes back to the traditional service routes like PHP, Rails, etc. Server-side rendering + caching feels much cleaner. The speed of deployment and page access is fine.

Through practice, I can feel: technology application should be used as problems arise and are solved; there is no standard answer.

The development experience of React Native is quite good.

Although many people are now using AI for vibe coding, after a period of intensive practice, my view is actually the opposite: as engineers, we still need to maintain technical learning, understanding, and mastery.

When I'm not too busy, I quietly read technical books. In the process, I feel my comprehension increase. Areas I hadn't noticed or wasn't solid on get supplemented.

Step-by-step tutorials make it easy to focus and immerse yourself. Writing tutorials also requires a strong ability to break down and organize knowledge.

function Echo() {
  const params = useParams();
  const [searchParams] = useSearchParams();
  return <h1>{params.msg || searchParams.get("msg")}</h1>;
}

const param = "From Param";
const query = new URLSearchParams({ msg: "From Query" });

export default function App() {
  return (
    <section>
      <p>
      </p>
      <Link to={'echo/${param}'}>Echo param</Link>
       <p>
       </p>
    </section>
    <Link to={'echo?${query.toString()}'}>Echo query</Link>
  );
}

Rybbit - an open-source and privacy-focused Google Analytics alternative, with a more intuitive and easy-to-use experience, fully 10x better.

Let Codex help me self-host this service on my RN VPS, and also integrate it with this current website. This kind of workload was unimaginable before Codex. It's not that it's difficult or complex, it's that installing software and configuring the environment is too troublesome. Now it's automatically done in a few minutes. I just help out by adding one domain A record.

fumadocs.dev is a beautiful and flexible documentation framework for any React.js project.

A few days ago, I tried translating a PDF ebook using Ollama with a local Qwen 3.5:4b model. Translation speed wasn't an issue, but the final layout turned out poorly. My workflow was to first convert the PDF to Markdown, translate it, and then generate a new PDF. However, the bilingual vertical layout (Chinese-English) in the resulting PDF was hard to read and very uncomfortable.

I looked into this problem and realized PDF text layout is indeed difficult to handle. Different books have different content and require different presentation styles. Optimizing and adjusting the layout for each book costs too much effort.

Finally, I found that HTML is the most suitable choice—styling is easy and results in a clean layout, and my goal is simply to translate and read comfortably. With this approach, I'll continue to improve it later and use it to help me read English books.