Some time ago, I wrote about 小红书 (Xiǎo hóng shū) and its adaptability. Today, it's time to talk about DeepSeek, another startup emerging from China.

Artificial intelligence is everywhere—here and in China (literally). From virtual assistants that understand whatever they want to algorithms that decide which cat videos you should watch. But every once in a while, something comes along that doesn’t just change the game but reinvents it. That’s where DeepSeek comes in: an AI that is not only cheaper, more efficient, and more accessible but also leaving Western tech giants looking like they just woke up with a serious hangover.

While OpenAI and Google strut around like the rich kids of Beverly Hills, showing off their closed models and billion-dollar budgets, DeepSeek showed up with the street vendor mentality: cheaper, more efficient, and without all the unnecessary hype.

To put things into perspective, DeepSeek’s training cost was 95% lower than GPT-4’s. Yes, you read that right: NINETY-FIVE PERCENT LESS. OpenAI spent like they were buying a stadium in Dubai, while DeepSeek worked wonders on a budget worthy of a low-budget sitcom.

And when it comes to usage costs, the difference is so absurd it feels like a typo:

• OpenAI charges $60 per million tokens.
• DeepSeek says, “Relax, buddy,” and only asks for $0.28.

It’s like OpenAI charging you for a VIP ticket to a McCartney concert, while DeepSeek lets you watch the same show at a bar for the price of a beer.

DeepSeek uses something called Mixture-of-Experts (MoE), an architecture that activates only the necessary parts of the model at any given time. What does this mean in non-nerdy terms? Imagine you have a CSI-style detective team: instead of having everyone investigate a missing cat case, they just call in the feline expert.

This not only reduces costs but also makes the model faster and more efficient, while OpenAI continues to burn through computing power as if they had infinite energy (spoiler: they don’t).

While OpenAI and Google guard their models like the Holy Grail, DeepSeek has taken an open-source approach. It’s like someone handing you the keys to a Ferrari (or a BYD Yangwang U9, to be precise) and saying, “Do whatever you want, just don’t crash it.”

This move is not just a smart strategy—it’s a statement of principles. DeepSeek is democratizing AI. (China democratizing… see the irony?) They’re allowing startups, researchers, and small businesses to compete on a level playing field.

Now let’s talk about what really makes some people nervous (cough cough, the U.S.). DeepSeek is Chinese, and that alone is enough to make a few folks uneasy. The U.S. has been restricting China’s access to high-end chips from Nvidia and AMD, hoping that without them, China wouldn’t be able to advance in AI.

But… surprise: IT DIDN’T WORK.

DeepSeek didn’t go crying on Twitter (sorry, X). Instead, they did what anyone stretching their paycheck until the end of the month would do: they got creative. They used cheaper GPUs, optimized code, and achieved results that would make more than one CEO in California sweat.

DeepSeek literally does the same (or better) than OpenAI for 1/200th of the price. It’s like paying for a VIP cinema ticket only to realize the person next to you got in for free with their loyalty points.

While OpenAI engineers enjoy offices with ping-pong tables, unlimited snacks, and lunchtime massages, things in China are a bit different. DeepSeek’s developers probably (I'm saying probably) work under the  996 schedule, in offices without a decent coffee machine. But guess what? They delivered results.

Because innovation doesn’t always come from comfort. Sometimes, it comes from necessity, pressure, and the ability to do more with less. And that’s exactly what DeepSeek has done.

AI forces us to face an uncomfortable truth: we are not as indispensable as we thought.

DeepSeek is not just redefining what’s possible in technology—it’s challenging our ideas about power, creativity, and progress.

While the West is stuck debating whether AI will take our jobs (spoiler: yes, and it will leave us making TikToks to survive), China is already leading the race. DeepSeek is proof that you don’t need the best resources to lead—just the right mindset.

So, what do we do now? Simple: stop underestimating others, learn from DeepSeek, and accept that the future waits for no one.

Welcome to the future. No holograms or flying cars, but hey—cheap AI. It’s something.

PS: As expected, DeepSeek is overloaded. You might experience some lag, but they’ll probably fix it soon.