During this year’s Spring Festival, an “AI rising star” from Hangzhou quietly emerged, named DeepSeek. It struck like a sudden lightning bolt, illuminating the global AI night sky and bringing a mysterious “Eastern power” to the open-source community.
As DeepSeek gained popularity, more and more people began to use this AI tool. So how can we make the most of it and unleash its powerful capabilities?
It starts with getting to know and understanding it—it’s like a knowledgeable friend who sometimes unconsciously “fills in the gaps”. It may confidently fabricate non-existent data, link unrelated information, or generate seemingly reasonable but actually incorrect content. Therefore, when using it, we must remain vigilant, avoiding misplaced trust in AI’s “nonsense”.
Additionally, we mustcontinuously learn “how to communicate with AI”. After all, even a powerful AI like DeepSeek can only provide vague answers if we cannot clearly express what we want.
Today, we have compiled a quick guide to DeepSeek for everyone, hoping to help everyone make good use of this powerful AI tool.
This image is generated under the guidance of DeepSeek
How does DeepSeek actually “think”?
If we were to describe DeepSeek with a metaphor, it’s like a very knowledgeable friend who has read a vast number of books and can instantly establish connections between various pieces of knowledge in their mind, answering all your questions (of course, excluding illegal matters). This is the working mechanism of modern large language models, and the core of this ability is the Transformer architecture pioneered by the Google team in 2017.
The most remarkable feature of the Transformer is its attention mechanism. For example, when you read a book, an ordinary person needs to read it from start to finish, while the Transformer is like a “super reader,” able to quickly find the most critical information in the text and understand the relationships between them.
In its growth process, DeepSeek resembles a knowledge-hungry top student, “reading” the vast knowledge accumulated by humanity—from the dry Wikipedia to beautiful literary works, from cutting-edge academic papers to professional technical documents. By continuously predicting missing words in sentences, it gradually masters the rules of language. This process is similar to how infants learn their mother tongue through listening and speaking; the large model learns to listen to questions and predict and generate the content we want.
However, DeepSeek’s “thinking” method is also unique; it is like a master of impromptu speeches, calculating every word precisely to ensure the content is coherent and creative. But due to this spontaneity, its answers can vary slightly each time, sometimes being brilliant and other times being vague or even frequently making mistakes. This is the phenomenon of ” AI hallucination “.
Before DeepSeek, foreign models like ChatGPT and Claude had already shone brightly, while domestic models like Wenxin Yiyan, Tongyi Qianwen, and Kimi also showcased their brilliance. So why did DeepSeek stand out and amaze the world after this Spring Festival?
The answer can be summarized in three keywords: hard power, cost-effectiveness, and open-source sharing.
Strong Capability: An AI “Tech Guru”
What sets DeepSeek apart is its “reasoning brain”. Compared to traditional instruction-based models, DeepSeek has particularly enhanced its reasoning capabilities in model design. Utilizing advanced techniques such as reinforcement learning, it is more like a scholarly thinker who not only understands what you say but also knows how to think deeply and innovate.
Currently, DeepSeek R1 has become the “strongest king” among open-source large models, closely chasing OpenAI’s top performers, and is the only one in the open-source camp capable of competing with closed-source giants.
Scroll up and down to see more, DeepSeek R1 compared to other mainstream models, image source: provided by the author
Affordable Price: The “Cost-Performance King” of AI
DeepSeek not only excels in technology but also impresses with its cost control. By optimizing training methods, they have reduced the training cost of the V3 model to $5.576 million—this figure is even lower than the annual salary of many executives at large model companies.
This high efficiency directly leads to ultra-low prices; now, inputting one million tokens costs only 4 yuan, and output only 16 yuan, leading some to jokingly call it the “Pinduoduo of large models”!
Image source: provided by the author
Open-Source Sharing: The “Wulin Alliance Leader” of AI
While other large models set up numerous barriers due to commercial interests, DeepSeek chose to fully open its code and allow free commercial use. It’s like a martial arts master openly sharing their lifelong skills, enabling the entire community to enhance their skills.
This philosophy of openness and sharing resonates globally. As a popular saying goes online, “The magical reality of 2025: we bought a model from a nonprofit organization (OpenAI) for $200 a month but received free open-source AI from a quant institution.”
How to Truly Make Good Use of DeepSeek?
The key to dealing with AI lies in “communication”. AI is like a genius with blindfolded eyes; it knows a lot but cannot directly perceive the world. Only through our descriptions can it understand what reality looks like.
It’s like trying to describe the sound of waves to someone who has never seen the ocean. If you say “splashing sound”, the other person might think it sounds like rain; if you say “like countless massive silks shaking in the air,” the imagery is entirely different. Similarly, communicating with AI requires such “descriptive ability”.
When communicating with AI, prompts are your “expressive ability”. Just like learning a new language, this ability needs continuous practice to improve. Even a powerful AI like DeepSeek can only provide vague answers if we cannot clearly express what we want.
Another crucial point is the “hallucination” problem of AI. It’s like a knowledgeable friend who sometimes unconsciously “fills in the gaps”. It may confidently fabricate non-existent data, link unrelated information, or generate seemingly reasonable but actually incorrect content, so we must be the “sword bearers” against AI hallucinations, remaining vigilant at all times.
In other words, AI is a powerful tool, but the effectiveness of the tool depends on the user’s ability. Through continuous practice and vigilance, we can truly unlock the value of AI while avoiding falling into the trap of “hallucinations”.
So, how can we make good use of this “AI tool”? Here are some tips.
1
Regular Question Template
In the past, when conversing with instruction-based models like ChatGPT, we often used complex prompt templates. However, DeepSeek is different; it’s more like a thinking partner, and simple, clear expressions yield better results.
I highly recommend this simple yet super practical “four-step questioning method”: Background + Task + Requirements + Supplement. For example:
[Background] I am the owner of a newly opened coffee shop;
[Task] I need a promotional plan for the opening;
[Requirements] Within a budget of 10,000 yuan, primarily targeting college students;
[Supplement] Our shop is located in a university town and focuses on specialty coffee.
The brilliance of this method lies in:
Background information helps AI understand your situation;
The specific task clarifies your needs;
Limitations specify key requirements;
Supplementary information adds important details.
This way of questioning not only helps DeepSeek better understand your needs but also avoids misunderstandings or off-topic responses. This is like talking to a smart friend; once you clarify the situation, they can naturally provide more targeted suggestions.
Remember, when conversing with DeepSeek, there’s no need for fancy techniques; keeping it simple and clear is key. Its reasoning ability is strong; as long as you clearly state your needs, it can understand your intentions and provide surprising answers.
Practical demonstration
First, let’s look at a wrong example:
Question: Write a promotional plan for the opening of a shop.
Output:
Scroll up and down to see more, image source: provided by the author
Now let’s look at a correct example:
Question: I am the owner of a newly opened coffee shop and need a promotional plan for the opening. The budget is within 10,000 yuan, primarily targeting college students. Our shop is located in a university town and focuses on specialty coffee.
Output:
Scroll up and down to see more, image source: provided by the author
2
Decomposing Complex Tasks
Currently, due to context and model capability issues, AI often “takes shortcuts” when handling complex tasks. In such cases, we need to help AI break down tasks.
Next, let’s take writing an article on “The Future Development of Remote Work” as an example to guide AI step by step to complete high-quality writing.
Step One: Have AI Write an Outline
Example: “I want to write an article about ‘The Future Development of Remote Work’; please help me list a detailed outline that includes an introduction, 3-4 main points, and a conclusion. Each main point should contain 2-3 specific subsections.”
Step Two: Have AI Elaborate on the First Part of the Content Based on the Outline
Example: “Based on the above outline, please help me elaborate on the introduction section. It should include the following points: Introduction to the current state of remote work, why this topic is important now, and the main directions the article will discuss. Please use vivid examples and data to support your points, and the language should be smooth and natural.”
Step Three: Have AI Continue to Output the Following Parts of the Content Based on the Outline
Example: “Please continue to complete the first main point of the outline. Ensure the content naturally connects with the introduction and includes specific case studies. Each subsection should be around 300-400 words, and please pay attention to the transitions between paragraphs. If data is involved, please clearly indicate the source.”
Step Four: Have AI Check for Errors in the Article Content and Verify Information
Example: “Please carefully read the entire article, focusing on:
1) Whether the data and facts mentioned in the article are accurate;
2) Whether the logical relationships between sections are coherent;
3) Whether there are contradictory viewpoints;
4) Whether the use of technical terms is accurate. If issues are found, please point them out specifically and provide modification suggestions.”
Step Five: Have AI Read the Generated Article and Optimize It
Example: “Please optimize this article from the following aspects:
1) Increase the readability of the article by using more vivid analogies and examples;
2) Optimize the transitions between paragraphs to make the article smoother;
3) Check and improve the language expression of the article to avoid repetition and verbosity;
4) Appropriately add some new insights or viewpoints to make the article deeper. Please modify it section by section and explain the reasons for each modification.”
3
Continuous Questioning Technique
Don’t expect AI to complete what you want in one go; when using AI, adjust your mindset and treat it as a knowledgeable but inexperienced intern. It has a broad knowledge base but needs your guidance to deliver satisfactory work.
For example, the first version of the coffee shop opening plan written by DeepSeek may not meet our needs. We need to optimize the content through continued questioning until it meets our requirements.
Using the coffee shop opening plan as an example, the first version provided by DeepSeek did not highlight the target audience of “college students” well enough. At this point, don’t rush to dismiss it; instead, you can continue the conversation like this:
The overall framework of the plan is good, but I think it can be more aligned with the college student group. Can you optimize it based on the following aspects:
– How to design the interior space based on college students’ learning and social needs?
– Considering students’ purchasing power, how to design a more attractive pricing strategy?
– Can you design marketing activities around campus events like finals and club activities?
This process is like brainstorming; each round of dialogue helps refine the plan. The key is to:
1. Be patient and not expect instant results
2. Provide clear optimization directions
3. Improve step by step, focusing on one aspect at a time
4. Summarize and consolidate promptly to ensure the direction is correct
4
Skillfully Using Different AIs
Each AI is like an assistant with a distinct personality; understanding their strengths allows them to shine in the most suitable roles.
For instance, DeepSeek R1 is like a meticulous strategic advisor. Give it a task, and it will think deeply from various angles, not only completing your requested content but also proactively discovering and supplementing details you might have overlooked. Especially when planning, it always provides surprisingly complete ideas.
However, when it comes to literary creation, DeepSeek can be like an overly enthusiastic writer, overly fond of embellishments, which sometimes detracts from the overall effect of the article. In such cases, the relatively restrained Claude can come in handy, as its writing is more concise and elegant, making it more suitable for tasks requiring literary flair.
Knowing who is best suited for what allows the team to maximize efficiency. Similarly, when using AI, selecting the most suitable “assistant” based on the nature of the task often leads to better results.
|
Name |
Uses |
Characteristics |
|
ChatGPT |
Copywriting, idea organization, knowledge Q&A, emotional companionship |
Text is somewhat stiff, with no significant weaknesses, but lacks standout features |
|
Claude |
Long-form writing, code generation, idea organization |
The best AI for generating Chinese content, with minimal AI flavor |
|
DeepSeek |
Long-form writing, code generation, mathematical calculations, idea organization |
Very suitable for idea organization and planning, with beautiful Chinese language, excels in poetry, but has a relatively high incidence of AI hallucinations |
|
Doubao AI |
Multimodal analysis, daily office assistance, emotional companionship |
Comprehensive functionality, suitable for daily work use, also great for emotional companionship |
|
Recraft |
Image generation |
Professional AI image generation tool with high image quality |
|
Midjourney |
Image generation |
High-quality images with strong playability |
|
Whisk |
Image generation |
Google’s image tool can generate different images based on characters and scenes, with good image quality |
|
Keli AI |
Image and video generation |
One of the most advanced video generation tools available, with high-quality video generation |
|
Hailuo AI |
Voice and video generation |
Generates high-quality voice quickly and is free |
Table: Designed by the author
Planning and Production
Author: Tian Wei, AI tool researcher
Review: Yu Naigong, head of the Robotics Engineering program at Beijing University of Technology, director of the Robotics Research Center at the Beijing Academy of Artificial Intelligence, doctoral supervisor
Planning: Lin Lin
Editor: Lin Lin
Proofreading: Xu Lai
Related Recommendations
1. “Nezha 2″’s groundhog is on fire! Do groundhogs really make an “ah~” sound?
2. This “code” from doctors must be understood! Otherwise, you might suffer a big loss…
3. What seemingly normal designs in restaurants are actually draining your wallet?
4. We challenged the physics competition questions using the recently popular DeepSeek, and the result was…
5. Five seemingly money-saving habits that actually increase cancer risk!
The cover image and images in the article are from copyright stock.
Reprinting may trigger copyright disputes, and for original images and text, please reply “reprint” in the backend.
Light Up “Looking at It”
Let’s Grow Knowledge Together!
