This article starts from the unique advantages of DeepSeek and comprehensively shares the methods of using DeepSeek, including: the correct way to discard prompt templates, how to make it “speak human language”, deep thinking skills, the powerful style converter, and usage pitfalls.
The “DeepSeek Usage Guide” is as follows:
Note: All techniques in this article come from real cases, and all prompts have been repeatedly verified.
1. The Most Important Secret: Discard Your Prompt Templates
If you are still using various “professional prompts” and “templates”, you are going in the wrong direction.
DeepSeek does not operate this way.
Why?
Because its core is a reasoning-based large model, not an instruction-based large model.
It’s like two interns:
-
One nerd needs you to meticulously arrange the task steps. (Instruction-based) -
One clever one can think for themselves as long as you explain the purpose. (Reasoning-based)
Let me illustrate with a real case:
A colleague in our community tested it for industry analysis in the new energy sector, preparing for negotiations with BYD suppliers.
Traditional Approach:
Please act as a new energy industry analyst and analyze according to the following steps:
1. Market size
2. Competitive landscape
3. Technology route
4. Future trends
Requirements: 800 words for each part, citing authoritative data...Result: A dry report that looks like AI.
Correct Approach:
I will negotiate with BYD's suppliers next week, but I know nothing about power batteries. Help me explain in the simplest terms:
1. What are their technical advantages?
2. How much might they charge?
3. What professional terms can I use during negotiations?
The key is for me to understand and sound like an insider.Result: DeepSeek provides grounded analysis and also includes negotiation language.
This is the biggest difference:
DeepSeek does not require you to write “professional prompts”;
What it needs are real scenarios and specific needs.
Here’s a universal formula for you:
❝
I want xx, to use for xx, hoping to achieve xx effect, but worried about xx issue…
It’s like talking to a clever subordinate:
-
Don’t say “Please write the weekly report according to the STAR method”
-
But say:
❝
I want to write a weekly report, the boss wants to see it on Monday, hoping to focus on xxx, the key is to make our department look good in front of the boss, outshine the neighboring R&D department, but worry that the R&D will question our product documentation’s detail…
2. The Most Underestimated Function: Make It “Speak Human Language”
Many people complain that DeepSeek’s replies are too abstract, like reading a heavenly book.
But you may not know that just a simple prompt can completely change this issue.
The magical prompt is:
❝
Speak human language.
That’s right, just these three words.
My students didn’t believe it the first time they tried it, but the result…
Original response:
Response after adding “speak human language”:
Because DeepSeek is very sensitive to the phrase “speak human language”.
Of course, sometimes these three words are not enough, you can also use this detailed version of the prompt, just copy and paste:
【Please output according to the following specifications: 1. Use plain language and avoid abstract metaphors; 2. Use everyday situational examples to assist in explanation; 3. Prefer specific nouns over abstract concepts; 4. Keep paragraphs concise (no more than 5 lines); 5. Technical expressions must include plain explanations; 6. Avoid literary rhetoric; 7. Key information should be prioritized; 8. Complex content should be explained point by point; 9. Keep it conversational but not overly simplified; 10. Ensure information is accurate while prioritizing commonly understood vocabulary】3. The Most Powerful Skill: Deep Thinking
This is something I must say:
A free domestic AI is making the $200 monthly GPT-o1 uneasy.
Why?
Because DeepSeek’s thinking style is more intelligent than GPT-o1.
Let me compare with a real case:
GPT-o1’s response:
DeepSeek’s response:
This is the biggest difference:
-
GPT-o1 lists linearly, like a high-end document tool. -
DeepSeek engages in deep thinking, like a thinking partner.
The free DeepSeek has directly made all AI companies in Silicon Valley work overtime,
Fortunately, they don’t have to celebrate the Spring Festival.
But recently, I’ve noticed a phenomenon:
Due to a surge in users, DeepSeek has clearly adjusted its response strategy:
-
Thinking time reduced from 20 seconds to 5 seconds. -
Depth of answers has noticeably decreased. -
Reflective ability is limited.
This is an understandable temporary measure, after all, computing power costs money.
But for us users, how can we continue to stimulate its deep thinking ability?
I have organized three core prompts, to sound sophisticated, called Deep Thinking Trio:
❝
Please simultaneously conduct critical thinking for at least 10 rounds during your analysis, be thorough. Please simultaneously consider your answers from the opposite perspective for at least 10 rounds, be thorough. Please simultaneously review your answers for at least 10 rounds, be thorough.
This way, deep thinking can return to around 20 seconds.
The italicized parts can be freely replaced with forms you are good at, or combined,
but the core is reflection.
4. The Most Powerful Style Converter
Yesterday, I used DeepSeek to write a Hanfu.
I praised Wang Xing’s girlfriend for her loyalty, intelligence, and courage. Wang Xing is the actor who was recently sold to northern Myanmar.
This allusion and this parallelism are truly unmatched.
After posting it in the group, it directly attracted three language teachers…
Some friends even said the writing was so sophisticated that even professional screenwriters were impressed.
In fact, this is DeepSeek’s fourth secret weapon: Style Converter
The prompt is simple, you can even dictate it:
❝
Imitate xxx’s writing style to write a xx genre piece about xxxxx.
But note two limitations:
On one hand, it’s more suitable for imitating classic authors, as the training data is abundant and stylistic features are prominent.
On the other hand, don’t expect a 100% reproduction; an 80% similarity is already impressive, the key is to capture the essence.
I believe DeepSeek must have trained on a large amount of literary corpus, especially Chinese classical literature.
For example, Lu Xun and Mo Yan are portrayed vividly.
Remember that universal formula:
❝
I want xx, to use for xx, hoping to achieve xx effect, but worried about xx issue…
For example:
❝
I want to write a piece praising Wang Xing’s girlfriend for her intelligence and courage, to showcase on Xiaohongshu, hoping to focus on imitating Wang Bo’s style, the key is to make my writing impressive and gain a billion likes on Xiaohongshu, but worry that others might not understand it and find it too obscure…
Similar style transfers can also be applied to contemporary writers, for example, we can imitate Liu Run’s New Year’s speech:
Of course, for better results, it’s best to use the universal formula.
5. Usage Pitfalls: When Not to Use It
After discussing so many advantages, we must also talk about its limitations.
DeepSeek is not recommended for the following scenarios:
1. Long Text Writing
Articles over 4000 words may have logical breaks; it’s better to use Claude 200k.
Because DeepSeek’s default is 64k, it’s insufficient for long texts.
2. Sensitive Content
After all, it’s a domestic AI, and the built-in review standards are excessively strict.
Often, you won’t know which sentence triggers the review.
How to solve this issue?
Since DeepSeek has a post-review process, there are three solutions:
3. Personal Style Writing
This won’t be elaborated; it’s a reasoning model, suitable for solving problems and imitation.
But it’s difficult to ensure the desired writing style effect through precise control.
This isn’t really a flaw of DeepSeek; it can only be considered a characteristic.
Specifics will be discussed in my next article.
7. How Will It Change Our AI Era?
After several days of intensive testing, I am increasingly convinced:
DeepSeek represents the future direction of AI — more understanding of human language, more capable of thinking.
You don’t need to learn its language; it is learning to understand your language.
What does this mean?
The barrier to using AI is rapidly lowering.
In the future, we won’t need to memorize prompt templates, learn specific instructions, or study various parameters.
All you need to do is clearly state what you want, tell it the specific scenario, and speak human language.
Because DeepSeek has eliminated the most inhumane settings of AI over the past 23 years —
— making humans learn machine language (prompts).
This is like a key turning point in the history of mobile phone evolution:
-
Nokia Era: Read the manual, learn the key combinations, flash the device. -
iPhone Era: Even a three-year-old can play with it.
My friends in the investment circle put it more bluntly:
❝
In 2024, any big influencer teaching people to write prompts is just collecting IQ taxes.
In short, stop learning prompts!
So, this is why I insist on publishing this article before the New Year.
DeepSeek represents a new generation of AI usage paradigm, welcoming the new!!
Disclaimer: DeepSeek hasn’t given me a penny, but I just want to promote it.
Chapter 8: The J-20 Moment of Domestic AI
We have waited too long for this day.
When I first used GPT-4 on March 15, 2023, my hands trembled.
Partly because of its power.
On the other hand, I also knew:
We might have to chase this level of AI for ten years.
Until DeepSeek released R1, I still tried it with a skeptical mindset, without any hope,
but seeing its deep thinking gave responses that were comparable to, and even surpassed, the top three (GPT, Claude, Gemini).
My hands trembled again.
I know, the era has changed.
DeepSeek has shown me:
-
It understands the human affairs behind “speaking human language”. -
It comprehends the social rules behind the “impressive effect”. -
It can use the “Tengwang Pavilion Preface” to write about the absurdity of scams in northern Myanmar.
While Silicon Valley is still teaching users how to “tame AI”, DeepSeek is doing something greater:
Teaching AI to understand human thinking.
The key is that DeepSeek is also open-source, which is the first time open-source AI has overtly caught up with closed-source AI.
As J-20’s chief designer Yang Wei said:
❝
“We no longer chase; we are defining the new battlefield.”
This may be the best New Year’s gift:
In the AI race, we are standing on the same dawn as the world for the first time.
Postscript
As I write this, someone will surely scold me in the comments,
thinking that a mere DeepSeek is not worth elevating to such national sentiment.
I don’t care, I’m willful, I just want to elevate it.
Technology has no borders, but I have a homeland, I have my own people.
I just want to use my own things, who are you to care?
Today, I can finally say to the western AI giants:
You have what we will have. What you don’t have, we are creating.
PS
As I finished this article, a firecracker went off outside.
Actually, fireworks are prohibited within the sixth ring road in Beijing; who knows which unknown hero did it.
I suddenly remembered what OpenAI’s chief scientist Ilya Sutskever said:
❝
True emergence often happens beyond mainstream visibility.
At this moment, in an unquiet office building in Hangzhou, perhaps there are engineers debugging the next generation model.
And the DeepSeek in your and my hands is that already lit firecracker.
Listen, to the sound of the AI revolution.
Happy New year
