Blog Update After Half a Year — Practices and Reflections

1/2

UN Village

Baekhyun·City Lights

0:003:56

Flag Review

Let's start by checking how many of the flags I set when building the site last time have been completed.

Even though I set these flags at the start, I didn't strictly follow them later. Thankfully, the end result still met my expectations. If time permits, I'll gather some audio clips and finish that remaining item.

I also made a development tracker to log feature requests — I saw ZaiHuaOvO's tracker and thought it was great — but I didn't keep updating it either. Felt like too much of a hassle. Went back to ignoring it. Time to bring it back!

The server has also been switched from CentOS to Ubuntu, and I've stopped deploying with Docker.

Iterating with Just My Mouth

With the AI explosion over the past half year, my focus has gradually shifted from traditional programming to AI-assisted programming. Many of this blog iteration's features were completed entirely through AI — SEO, style optimization, login mechanisms, SSL certificate requests, and more — and they all turned out quite well. Searching keywords on Google actually brings up my blog now, and browser audit scores are pretty decent too.

So my blog has essentially become a testing platform for my wild ideas — I try to implement whatever weird ideas I have through AI.

Now tools are all supporting voice input, so in the future it really will just be a matter of talking...

Random Musings

Mindset Shift

Although AI can now do a great many things, it's still not as invincible as people imagine. I still find GPT-generated UIs unattractive, and plenty of people complain that even GPT's own product interfaces look ugly. Of course, Western component library aesthetics also differ from Chinese ones — maybe that's part of the reason too.

You still need to look, learn, and borrow more. Prompts are written by humans, and many things ultimately have to come from a human mind. What makes great artists great is perhaps their ability to create things the human mind couldn't conceive — just like how we imagine ghosts and monsters, we can only piece together fragments from horror or disaster films we've watched. Without those references, our brains can't even conjure what a demon looks like.

Of course, if you really can't imagine something, letting AI free-associate is also a good option.

Right now I have the highest hopes for DeepSeek. Anthropic still isn't very friendly toward us — they even see us as their biggest rival, because we're the ones most likely to surpass them. While domestic models still lag behind foreign ones, I believe homegrown AI will rise and give us affordable, practical tools.

Learning Pace

Initially, I dove headfirst into Nuxt and Nest docs, furiously reading documentation and trying to learn systematically. To this day, I still haven't systematically finished reading all the docs. Back then I had a clear goal: level up my technical skills. But as AI exploded, my mindset gradually shifted, and I've been stuck in a fog of confusion ever since.

I don't know if it's the AI, the times we live in, or my own state of mind — but I've become unable to settle down and do one thing at a time. This feeling is getting more and more pronounced. I can't calmly immerse myself anymore — can't read a document or a book cover to cover. I rely too much on AI, and I wonder if that's a bad thing or a good thing. Many people have discussed this topic; AI might genuinely be eating away at my patience.

Since January, I've gradually stopped digging into code details. From a learning perspective, I think the focus should shift to project architecture — considering engineering problems and thinking in terms of architecture: how to make AI generate project-compliant code more easily, faster, and with fewer tokens. That's what I want to experiment with and practice next.

Video Creation

During the peak of the AI wave, I made an opencode tutorial. I happened to catch the wave of hype and got a decent number of views — total income was roughly 170 yuan. I have to say, relying purely on video income these days is really low, though I genuinely didn't expect to make any real money from it. Still, it feels like not enough to cover the effort — I worked late into the night recording, often until 2 or 3 AM, and had to show up at work the next day. But at the time, I was full of energy and drive.

Later, I stopped uploading videos. The feedback was mixed, and I certainly wasn't immune to negative comments. But ultimately, the reason I stopped was a change in mindset — I felt that teaching others how to use a tool in excessive detail was just too tedious. New alternatives emerge so fast, with endless concepts and tools, and every company is trying to make it simpler for users to operate AI. It suddenly felt like spending hours explaining a single tool wasn't very meaningful anymore, so I stopped.

Beyond that, through these videos I also built a small community and gradually got to know many friends. To me, that has been truly tremendous positive feedback. When the chance arises, I'll continue sharing useful things and hope to grow the community further — hopefully it works out.

I used to wonder why people chase whatever is trending — why pursue the hype? Later I gradually understood: hype is money. Thinking back, from my school days making videos I was always trying to ride trends, and I'm still doing it now. Along the way I've talked with many people, learned a lot, and seen through the true nature of some chat groups. I'm going to steer back toward the money-making path — because earning money truly matters.

Of course, if making money isn't the goal, I actually think the smaller the community, the better.

New Flags

Final Words

I can't force out anything more to write — I'll just write whenever inspiration strikes from now on.