After DeepSeek’s explosive popularity, I received over 50 inquiries from readers, all with similar concerns: seeing these feedbacks, I couldn’t sit still. The first batch of individuals who truly leveraged the value of DeepSeek has already seized the industry’s dividends. Yes, the free DeepSeek has caught up with the paid OpenAI o1 model, and even surpassed it in some aspects. Just like the moment when the J-20’s maiden flight tore through the night sky in 2011, we finally don’t have to use AI on our knees.
Note: All techniques in this article come from real cases, and all prompt words have been repeatedly verified. The most important secret: discard your prompt templates. If you are still using various ‘professional prompt words’ and ‘templates,’ you are going in the wrong direction. DeepSeek doesn’t buy into this. Why? Because its core is a reasoning-based large model, not a command-based one. It’s like two interns: one is a bookworm who needs you to arrange task steps meticulously (command-based), and the other is a quick-witted person who can think for himself once you explain the objective (reasoning-based). Let me illustrate with a real case, a test conducted by one of our community members who operates in the new energy industry, preparing for negotiations with BYD suppliers. Traditional approach: Please act as a new energy industry analyst and analyze as follows: 1. Market size 2. Competitive landscape 3. Technology routes 4. Future trends. Requirement: Each part should be 800 words, citing authoritative data. .. Result: A dry report is obtained, which is obviously AI-generated. An obviously AI-generated report has no use except for being correct. Correct approach: I am going to negotiate with BYD’s suppliers next week, but I know nothing about power batteries. Help me explain in the most common language: 1. Where is their technological advantage 2. How much they might ask for 3. What professional terms can be used during negotiations. The focus is to make sure I understand and can act like an expert. Result: DeepSeek directly provides down-to-earth analysis, even including negotiation phrases. It even comes with phrases, which is the biggest difference: DeepSeek doesn’t require you to write ‘professional prompt words.’ What it needs is real scenarios and specific needs. Here’s a universal formula for you: I want xx, for xx use, hoping to achieve xx effect, but worried about xx issues. .. It’s like talking to a smart subordinate: Don’t say ‘Please write a weekly report according to the STAR method,’ but say: I want to write a weekly report, the boss needs to see it on Monday, hoping to focus on xxx, the focus is to make our department shine in front of the boss, outshine the neighboring R&D department, but worried about R&D questioning the detail of our product documentation… The most underestimated feature: Let it ‘speak human language. ‘ Many people complain that DeepSeek’s responses are too abstract, as if they are reading a heavenly book.For instance, you might not know that a simple prompt can completely transform an issue. This magical prompt is: ‘speak plainly.’ Yes, just those three words. My students didn’t believe it at first, but the results were… Original response: abstract, really damn abstract. After adding ‘speak plainly’: instantly grounded in plain language, right? Because DeepSeek is very sensitive to the phrase ‘speak plainly.
‘ Of course, sometimes these three words are not enough, and you can use this detailed version of the prompt, simply copy and paste it: [Please output in the following standard: 1. Use plain and straightforward language, avoiding abstract metaphors; 2. Use everyday scenario-based examples to illustrate; 3. Prioritize concrete nouns over abstract concepts; 4. Keep paragraphs concise (no more than 5 lines); 5. Technical expressions should be accompanied by popular explanations; 6. Avoid literary rhetoric; 7. Prioritize key information; 8. Break down complex content into points; 9. Maintain colloquialism without oversimplifying professional content; 10. Choose widely recognized vocabulary while ensuring information accuracy.] The most powerful skill: deep thinking. This is something I have to mention: a free domestic AI is making GPT-o1, which charges $200 a month, uneasy. Why? Because DeepSeek’s way of thinking is wiser than GPT-o1’s. Let me compare with a real case: OpenAI o1’s response: Image from @D&roi teacher. DeepSeek’s response: reasoning, not linear listing. This is the biggest difference: GPT-o1 linearly lists, like a high-level document tool; DeepSeek deeply thinks, like a thinking partner. The free DeepSeek directly makes the entire Silicon Valley AI company’s top talents work overtime, fortunately, they don’t have to celebrate the Spring Festival. But recently, I’ve noticed a phenomenon: due to the surge in users, DeepSeek has noticeably adjusted its response strategy: thinking time from 20 seconds to 5 seconds, the depth of the answer has significantly decreased, and the ability to reflect is limited. Computational tension, only 5 seconds is understandable as a temporary measure, after all, computational power is money-burning. But for us users, how to continue to stimulate its deep thinking ability? I have organized three core prompts, for the sake of showing off, called the deep thinking trio: Please engage in critical thinking at least 10 rounds during your analysis process, be detailed. Please consider your answers from the opposite side at least 10 rounds during your analysis process, be detailed. Please review your answers at least 10 rounds during your analysis process, be detailed. In this way, deep thinking will recover from 5 seconds to about 20 seconds.Italicized sections can be freely replaced with forms you are proficient in, or combined and overlapped, but the core is reflection. Deep thinking has turned into the most powerful 20-second style converter. Yesterday, I used DeepSeek to write a Chinese赋, praising Wang Xing’s affectionate, courageous, and intelligent girlfriend. Wang Xing is the actor who was recently sold to Northern Myanmar. I read it myself and found it particularly satisfying with its references and parallel prose; no one can match it.
After posting it in the group, it directly revealed three language teachers… A powerful tool that can awe others. Some friends even wrote text that professional scriptwriters praised as mature and flavorful. In fact, this is DeepSeek’s fourth secret weapon: the style converter. The prompt is simple; even dictation will do: imitate the style of xxx and write a piece of xxx in the xxx genre. However, there are two limitations to note: on one hand, it is more suitable for imitating classic writers because there is ample training data and distinct style characteristics. On the other hand, do not expect a 100% reproduction; an 80% similarity is already impressive, and the key is to capture the essence. I believe DeepSeek must have trained on a vast amount of literary materials, especially classical Chinese literature. For instance, Lu Xun and Mo Yan are both vividly imitated. To imitate Lu Xun’s style, remember the universal formula: I want to xx, to give xx to use, hoping to achieve xx effect, but worried about xx problem. .. For example: I want to write a piece praising Wang Xing’s girlfriend’s intelligence, courage, and dedication, to show off on Xiaohongshu, hoping to focus on imitating Wang Bo’s style, with the emphasis on my own literary flair, to gain a billion likes on Xiaohongshu, but worried that others won’t understand if it’s too obscure… Similar style migration can also be applied to contemporary writers, such as imitating Mr. Liu Run’s New Year’s speech: the imitation prompt can be freely noted, and one should first provide the original content (as detailed as possible, generally not less than 8000 words), and then directly request imitation. Of course, for better results, it’s best to use the universal formula. Usage restrictions: What situations should not use it? Having said so many advantages, we must also talk about its limitations. DeepSeek should not be used in the following scenarios: 1. Long text writing articles exceeding 4000 words are prone to logical breaks, it is suggested to use Claude200k. Because DeepSeek defaults to 64k, which is not enough for long texts. Long texts are indeed insufficient at present. 2. Sensitive content After all, as a domestic AI, the built-in audit standards are outrageous. Whenever such a reply is triggered, it means the audit has been activated, and often you don’t know which sentence has triggered the audit.How can we address this situation? Since DeepSeek employs a post-moderation system, there are three potential solutions:
1. Click on ‘edit’ in your question and resubmit it several times; eventually, one submission will bypass the review process. Click on ‘edit’ in the question. 2. While generating a response, rapidly click the ‘copy answer’ button. This ensures that when the review is triggered, your clipboard contains the content of the previous answer. Click on the ‘answer’ button. 3. For a permanent solution, switch to the top three AI platforms (GPT, Claude, Gemini). 3. Personalized style writing is not elaborated here, as it is a reasoning model suitable for problem-solving and imitation. However, it is challenging to ensure the desired style through precise control. This is not a flaw of DeepSeek but rather a characteristic. I will discuss this in more detail in my next article. How will it change our AI era? After several days of intensive testing, I am increasingly convinced that DeepSeek represents the future direction of AI—more human-like and thoughtful. You no longer need to learn its language; it is learning to understand yours. What does this mean? The barrier to using AI is rapidly decreasing. In the future, we won’t need to memorize prompt templates, learn specific commands, or study various parameters. Simply, state what you want, describe the specific scenario, and speak naturally. Because DeepSeek has eliminated the most anti-human aspect of AI since 2023—making humans learn machine language (prompts). In fact, the original intention of Andrew Ng’s prompt engineering course was good, much like the key turning point in the evolution of mobile phones: Nokia era: Read the manual, learn combination keys, flash the phone. iPhone era: Even a three-year-old can play with it. My investment circle friends put it more directly: In 2024, influencers who still teach people to write prompts are simply taxing intelligence. In summary, why bother learning prompts anymore! This is also why I insist on publishing this article before the New Year. DeepSeek represents a new paradigm for AI usage, bidding farewell to the old and welcoming the new! Disclosure: DeepSeek has not paid me a single cent, but I am determined to praise it highly.
