English - 蜡笔小新《大人帝国的反击》Nostalgia, and Why I Almost Built DeepSeek

English translate of the chinese blog

So this is a blog in Chinese. I haven’t written in Chinese for a long time, so I just wanted to try writing again.

It’s probably been three or four years since I last wrote like this. Recently I’ve been using Chinese less and less, so I feel like I should put in a bit more effort.

So this time I want to talk about an old anime movie I watched before—Crayon Shin-chan: The Adult Empire Strikes Back. I also linked a YouTube video below (a YouTube video), you can check it out—it’s a pretty detailed explanation, kind of like a quick promo. In terms of popularity, if you’ve seen subtitles before, you probably know it.

If we’re just talking about this movie, I personally think it’s one of the best animated films I watched as a kid. The Crayon Shin-chan series releases a movie almost every year, but I think I stopped watching around 2016 because I got too busy and basically didn’t have time anymore.

I thought about it again recently because I started casually watching Crayon Shin-chan again. It works well as something to watch while eating—pretty relaxing. I’m not watching super seriously, but some scenes still hit.

My basic impression of The Adult Empire Strikes Back is that all the adults kind of “leave,” and it becomes like a paradise for kids.

But more specifically, there are actually a lot of emotional scenes later on. The one that left the biggest impression on me is Hiroshi’s life. There’s this part where he’s watching his father, riding a bicycle along a road through rice fields. Then it transitions—from him watching his father, to Shin-chan watching his father—on what feels like the same road, but the background isn’t just simple rice fields anymore; it’s more modern, more developed.

You get what I mean—it feels like looking at your past self from the present.

When I first watched it, I was probably in elementary school. I remember feeling touched, but I didn’t really think too deeply about it. Watching it now feels kind of similar, except I understand a bit more.

At the time, I didn’t really have a strong “father” connection to the movie. It felt a bit distant to me since I didn’t grow up with that exact kind of family experience. But I could still tell the emotion was there, and it felt good.

I guess I kind of envied Shin-chan a bit—having a home, a car, a stable life, not having to run around everywhere. That kind of thing actually feels pretty nice.

As for the plot, it’s still pretty typical Crayon Shin-chan style, but the main thing is that every character has their own personality, and the movie uses those personalities to drive the chase scenes—showing how they escape through those interactions.

Even though the earlier movies were also pretty good, I feel like this one is the best. At least to me, every character feels distinct.

The villains, on the other hand, didn’t leave as strong of an impression. Looking back now, I can barely remember their background or motivation. But overall, this is still one of the best entries in the series for me.

Looking at the overall story, the later part is basically about fighting over the future. The theme is kind of about returning to the past, nostalgia. To me, it feels like a past vs future thing.

A lot of people feel like the old days were better, and the future feels like a downgrade, like it’s not as good as before.

It’s a pretty normal theme—past vs future. But here it’s framed as the 20th century vs the 21st century, which made it feel interesting when I first watched it.

This setup also reminds me a bit of Doraemon—that kind of imagination about the future—but the direction is different. Past and future both have their own good sides. The future just has more innovation.

Like AI now, and all the new infrastructure, new technology, new innovation—these things are gradually reshaping the systems built in the past.

I feel like this kind of change is inevitable. There might be some chaos in between, but eventually things settle back into order.

Oh… I think I’ve drifted off-topic. I was supposed to be writing about Crayon Shin-chan, but somehow I ended up talking about tech.

I’ll just treat this as digging a few holes. I’ve already dug a few anyway. Writing like this in a blog kind of leaves things open. I don’t even know what “digging a hole” is in English—is it a plot hole? But a plot hole is more like a mistake by the author. Mine feels more like something I plan to come back and fill later.

I remember back then, when my family wanted me to get a winter job, I was thinking: if these models are trained on the entire internet, why not use them to generate the training data I need?

Most of my exposure was during the GPT-3 beta test era. Later I also tried writing some of my own machine learning algorithms, mostly for flow detection.

After that I started thinking—could I build my own custom AI model? Like one with a consistent tone or personality, instead of relying on system prompts every time.

Back then, “synthetic data” wasn’t really a common term, or at least I didn’t know it. I just had this vague idea that maybe I could use generated data to train my own model.

So in my sophomore year, I started exploring this idea and tried a bit, but then I got busy with a winter job and never really finished anything.

I also built something before. I used Flutter to create a cross-platform social app. The original idea was to put a bunch of AI models inside it—like each user could have group chats, and most participants would be AI models. I just wanted to see how AIs would talk to each other.

At the same time, I thought maybe I could collect some synthetic monitoring data from those interactions and use it for training later. But that was more of a later idea.

In the end, I spent most of my time on the UI—probably about a month. I’m okay with Flutter, but if you ask me to write UI in Kotlin Multiplatform or Android SDK, it would probably take me 3–4x longer.

Now AI can help with UI, but I still feel like I should build things myself.

After finishing the UI, I ran into tons of bugs. I tested it across iPhone, Windows, macOS, Linux, even iPad—and there were issues everywhere. In the end, I only really finished the UI, and nothing else.

Feels like a hole I dug myself.

Looking back now, a lot of the ideas I had back then are being built by others. Like DeepSeek, they also talk about using synthetic data to train models, or using larger providers like ChatGPT to assist training.

Then when I saw products like Motbook, I had this weird feeling—like, “wait, this is kind of similar to what I was thinking back then.” It’s a bit strange, but also just how things go.

Honestly, building something like this properly is pretty hard. I really respect the people who actually manage to do it.

There are also things like coding agents and similar AI stuff—I might look into those later. Right now I feel like I should focus more on learning, and not rely too much on these tools.

It’s kind of like—if you rely too much on tools, your growth gets limited by what those tools can do. That’s how it feels to me. The “hole” you dig should probably be bigger than the tool itself… okay, I’m rambling again.

Anyway, I don’t even know what I’m talking about anymore at this point.

Maybe next time I’ll write a mixed Chinese-English blog, or just write in English entirely.