01 DeepSeek is an Opportunity for Everyone
AI is transforming from “black technology” to “oxygen”—those who cannot use AI may not even have the right to breathe in the future.
The domestic large model DeepSeek, which has shocked Silicon Valley and Wall Street, deserves everyone’s close attention, because it can easily convert into high-quality productivity.
For ordinary people, ease comes in four ways:
First, registration and use are convenient, without the need to bypass restrictions like with GPT or Claude;
Second, it has a Chinese advantage that crushes foreign large models, with logical reasoning surpassing domestic counterparts;
Third, it is free, with absolutely no economic burden;
Fourth, it has strong comprehension abilities and high compatibility with prompts, allowing users to ask questions as they wish—finally, one can clear out the “strongest prompt guidelines” and “universal prompt formulas” from their bookmarks.
02 DeepSeek Disrupts AI Technology
Domestic AI large model DeepSeek has emerged, topping the free app download charts in China and the U.S. yesterday.
At the same time, DeepSeek’s ultra-low training costs have shocked global computing power, causing an epic sell-off in U.S. tech stocks that night.
DeepSeek’s ability to create such a storm is due to its many “heaven-defying operations”:
Cost Slayer: Performance comparable to GPT-4o, but took only 53 days to train at a cost of just $5.58 million, less than OpenAI‘s 1/20, with usage costs as low as $0.14 per million words, which is 1/18 of OpenAI’s, equivalent to the price of a 4 yuan ice cream cone from Mixue, allowing you to chat with AI through the entire series of “Empresses in the Palace” (76 episodes in total).
Sanction Buster: In the context of chip restrictions, using a stripped-down NVIDIA H800 chip, it optimizes algorithms to directly compete with the U.S. H100 clusters, forcing foreign media to admit that “sanctions have instead spurred Chinese innovation.”
These alone have caused Wall Street analysts to exclaim: AI companies are entering their “Sputnik moment”—the panic that arose in the U.S. society after the Soviet Union successfully launched the world’s first artificial satellite, Sputnik 1, in 1957.
Foreign media referred to it as a “mysterious Eastern power,” with Feng Ji, the creator of “Black Myth: Wukong,” saying, “This could be a technology achievement of national significance.”
03 User Experience
One particularly interesting and practical aspect of DeepSeek is that its thought process is transparent; before answering questions, it engages in deep thinking and displays this thought process. It has a sci-fi feel reminiscent of the Trisolarans, along with an endearing self-talk quality, and this analytical process helps the questioner learn frameworks and ideas.
For example, if you ask it to score a middle school entrance exam essay, it will consider the scoring criteria for that specific region, focusing on aspects such as topic understanding, content structure, language expression, and creativity, while also considering word count.
Surprisingly, it doesn’t just score; it also considers questions that the user hasn’t asked, such as besides knowing the score, the user might want to know the strengths and weaknesses, and suggestions for improvement.
Moreover, its humanized approach is remarkable; it uses encouraging language when making suggestions, helping students improve, and maintains a friendly tone…
Not only is it understanding, but the suggestions it provides in 10 seconds of thought are also particularly professional. These suggestions are content not found in the original text, proposed based on logical context, and are very relevant, with thoughtful explanations noted in parentheses.
After a few suggestions for revisions, it indeed provided friendly encouragement as it had thought.
Subsequently, after multiple requests for refinement, such as making the language more vivid and the expressions slightly more ornate, it consistently achieved ideal results. It could even extend beyond your questions and requirements, acting like a professional, patient teacher guiding you step by step.
The entire process of scoring, revising, and finalizing takes only a few minutes. If there are any revision requests in between, they can be made at any time, and based on its feedback conclusions, you can continuously suggest directions and requirements for corrections, which will get closer to perfection.
I also tried asking DeepSeek to solve math problems, and the content of its three seconds of thought spanned several pages, with a very clear and understandable problem-solving process.
While researching an unfamiliar industry, I asked DeepSeek how to quickly get started, and it provided a framework, explanations of core terminology, and recommended professional books. For overly obscure parts, if I said, “Please explain it in simpler terms,” it would respond with plain language and examples…
In summary, the clearer and more direct the question, the closer the answer will be to your requirements. The principle of communication is to treat it as an all-knowing and tireless teacher; if you don’t understand, just ask. After receiving an answer, honestly express your feelings and feedback, and it will infinitely approach your requirements.
There are many more possibilities to explore, and I believe you will immediately think of many new ways to use it. Go ahead and try it out! Search for ‘DeepSeek’ in the mobile app market or on your computer’s web page, and don’t forget to enable “Deep Thinking” and “Online Search”. Have fun!
