There is a question that has always made me feel a little strange: many people regard AI as a speed-increasing tool. It used to take half an hour to write an email, now it takes five minutes. It used to take a day to adjust a piece of code, but now it takes two hours. That's certainly good, but is it ten times better? no. It's just the same thing done faster. Ten times better is another thing. Where does this theory come from? The term 10x comes from Peter Thiel’s book Zero to One. His original words were: For new startups to survive, the solutions they provide must be ten times better than existing solutions. Why ten times, not three times? Because switching has costs—user time, learning curve, migration friction. Three times better is not enough for people to be willing to bear these costs. Only a gap of orders of magnitude can break the inertia. Later, Dan Sullivan pushed this idea a step further in "10x Is Easier Than 2x" and came to a more counter-intuitive conclusion: pursuing ten times is easier to achieve than pursuing two times. The logic is this: Double thinking is based on the existing path. You will list all the things you are doing now and think about how to improve each one. This can be distracting. And when the goal is 10x, most current paths immediately seem inadequate—you’re forced to give up 80% and focus on the 20% that can actually sustain 10x. The more aggressive the goal, the easier it is to focus. The first thing Steve Jobs did after returning to Apple was to cut the product line--cutting more than 70 products down to
- The iPod wasn't the first MP3 player, but its ease of use, combined with the iTunes ecosystem, made it impossible for competitors to compare. The iPhone is not a "phone with more buttons" and directly replaces the camera, map and music player. This is refactoring, not optimization. So does Tesla. Musk did not try to "increase the battery life of electric vehicles by 10%." He first asked "how to make electric vehicles drive better than fuel vehicles" and then pushed the technical path in turn. From acceleration performance to OTA upgrades to super charging networks – every dimension is redefining what a car is. SpaceX’s example is even more extreme. The aerospace industry has been optimizing rocket performance for decades. Musk asked another question: Why can't rockets be reused like airplanes? With vertical recovery, launch costs are reduced by an order of magnitude. What these three cases have in common is that they all involve abandoning a familiar direction, accepting a goal that initially seems impossible, and then working backward from that goal to a completely new path. The Boundary of AI Capabilities in 2025 Before talking about how AI can help you achieve ten times, you must first be honest about what it can do now. The ability of large models Gemini
- 5 Pro supports the context of 1 million tokens (about 30,000 lines of code), the Deep Think mode has reached the top level in the 2025 mathematics competition, and the Computer Use function can directly operate the browser and mobile interface. Claude Sonnet 4.5 achieved
- 2% on the SWE-bench software engineering benchmark, was proven to last over 30 hours on complex multi-step tasks, and scored
- 4% on the OSWorld Computer Operations Benchmark. Qwen3-235B's performance is comparable to the top model under 22B activation parameters. For the first time, it supports dynamic switching of fast thinking and slow thinking in a single model, and supports 119 languages. Kimi k2 is specially designed for Agent and natively supports tool calling and multi-step reasoning. The API cost is about 1/10 of Claude
- What do these numbers say? AI is no longer just a "conversation tool" - it is moving in the direction of "task performer". The context window is large enough, meaning it can handle the entire project at once; Computer Use means that it can operate any software designed for humans without waiting for an API interface; working autonomously for a long time means that it can complete complex tasks without anyone watching. The actual boundaries of AI products But it needs to be made clear what it can’t do well. AI is strong in structured tasks and repetitive tasks. It can generate qualified code, draft contracts, handle standardized processes, and analyze data. But where in-depth industry judgment, complex negotiations, and creative breakthroughs are required, it is still only an auxiliary. It can write code, but it cannot define the strategic direction of the product; it can organize interview records, but it cannot judge the potential of candidates; it can generate marketing copy, but it does not understand what the brand represents in the user's mind. This boundary is important because it determines what is the "80%" that can be handed over to AI and what is the "20%" that you need to keep. AI + is ten times better, how to implement it for individuals? An independent designer used to spend more than 70% of his time on customer communication, project management, and file organization, and less than 30% of his time was actually spent on design. He used AI to reallocate this time: AI handles routine customer inquiries, automatically generates project documents, and organizes design materials. Design time increased to 80%. But the more critical change is that he began to use AI to generate first drafts, quickly iterate plans, and automatically adapt to different sizes. As a result, he was able to take on projects that required a small team to complete by himself, and the order volume and unit price increased by 3-5 times. Moreover, by using AI to handle standardized work, he has more time to study cutting-edge trends and communicate in depth with clients. The uniqueness of his design works is actually higher than before. There is something worth noting here: AI did not let him "do it faster", let him "do different things". Use AI + tenfold thinking to think about individual growth. The formula is roughly as follows: Clarify your core values → Use AI for non-core but necessary things → Invest the freed time in the depth of core capabilities → Form professional barriers that are difficult for others to copy. Everyone can become a top expert in a certain niche. The traditional scaling logic for organizations is linear: serving 10 times customers requires 10 times customer service; developing 10 times functions requires 10 times engineers. AI makes this logic fail. A SaaS company has a customer service team of 50 people, each of whom handles about 30 issues per day. After deploying AI customer service, 90% of common problems will be handled by AI and the remaining 10% will be handled manually. Then what? Instead of laying off employees, they redefined the jobs of those 50 people. One group turned to in-depth user success management - proactively helping corporate customers optimize usage and increase renewal rates; the other group transformed into a product feedback team to analyze the real needs behind user problems and promote product iteration. Six months later: User satisfaction from 75% to 92%, renewal rate from 80% to 95%, NPS from 30 to
- The number of people has not increased, but the service quality and business value have jumped. This case illustrates a very valuable direction: "How does AI allow organizations to redefine the value of each person?" The real question is, before using AI, there is a test method to judge whether you are pursuing ten times or two times: see whether what you do after using AI is fundamentally different from before. If you use AI to write the original email from half an hour to five minutes, it’s still the same email, that’s twice as much. If you use AI to automate the standard parts of customer communication, and then use the time saved to provide in-depth services that you couldn't do before, it starts to get closer to ten times. The key question is "If this thing can be completely automated, what should I do with the freed time to achieve a qualitative leap?" The answer to this question is often harder to find than the AI tools themselves. It requires you to truly figure out what your core value is - "What can I do that others can't do?" The AI of 2025 will be powerful enough to handle more repetitive tasks than any tool. But this is just leverage, and the leverage itself does not determine what you can lever up. What determines the outcome is the direction you decide to pry. The ten times better theory is more possible in the AI era—but only if AI is used in the right direction. The criterion is simple: if after using AI, your life becomes simpler and more focused, and you are doing more of the things that only you can do - then you are going tenfold. If it's just busier, it's just faster, then maybe it's just running anxiety faster.
