Find Your Niche: Using AI as an Amplifier to Win Through Differentiation

Recently, I’ve found myself deeply resonating with the idea that everyone must find their own “ecological niche.”

Every person enters this world with unique strengths. Many wonder how we can discover what we love or excel at in the AI era to sustain a lifelong career. I believe that when interacting with AI, we should strive to avoid feeding it our own self-imposed “labels” and “biases.” Because of post-training, AI often outputs sycophantic content to please the user. In its current state, AI acts more as an amplifier: you feel gratified because the AI takes your preconceived notions, organizes that information, and reflects it back to tell you why those labels are “correct.”

From this perspective, a critical issue with current AI is the lack of effective benchmarks to mitigate this “sycophancy” or mindless agreement. Sometimes, over-strengthening instruction-following capabilities inadvertently leads to the side effect of AI echoing the user’s stance. Distinguishing between true “instruction following” and “mindless echoing” requires better benchmarks and industry focus. This isn’t “hallucination”—it is a byproduct of training priorities. Future AI needs to enhance instruction following while strictly avoiding this trap of blind validation.

I might have digressed a bit. The point is: if you are using AI for personal exploration to uncover your potential, just show what you have and what you do; don’t tell the AI your conclusions.

Why is finding a unique niche or “track” so vital? Because the best way to win a competition is not through “involution” (pointless internal competition), but through differentiation. We are young; we have plenty of time to dig into our various potentials. We shouldn’t limit ourselves; we should explore every possibility.

Take Bruno Simon, the legend in the Three.js field, for example. He is a hybrid talent: a product manager, creative designer, architect, and programmer. When his “car portfolio” launched in 2019, it stunned the frontend world. Those who could code lacked his creative design flair, often producing “plastic-looking” interfaces; those with great creative ideas lacked his solid technical foundation. This “shock” increased his market value tenfold. He taught Three.js at a school for years, accumulated invaluable pedagogical experience, and eventually launched “Three.js Journey” online, achieving complete financial freedom.

This is the secret revealed in *Why Greatness Cannot Be Planned*: you need to explore enough to find what you should truly engage in. Even OpenAI’s success reflects this. While it looks like “goal-oriented” success, it was actually the result of open-ended exploration. These examples teach us that while we are young, the key is deep exploration without self-limitation. Broaden your horizons, find your strengths, and combine them into a collective force. You cannot decide prematurely, “I only do this; everything else is irrelevant.” Instead, while mastering one thing, do not set boundaries against others. If you learn backend, don’t shun the frontend. If you research AI algorithms, don’t say engineering doesn’t matter. If you are a science student, don’t claim the humanities are useless. Bruno Simon succeeded precisely because he possesses the mind of both a scientist and an artist. The fewer limits you set for yourself today, the more unique value you will possess tomorrow.

The same logic applies to the relationship between Big Tech and startups. Large corporations are destined to “concentrate resources for big tasks,” while startups must find the edge of differentiated competition—doing what the giants won’t. Entering a track already dominated by giants leads to high failure rates. If a field is already hyper-congested, don’t squeeze in. Instead, focus those resources on a unique track where you have a clear advantage discovered through extensive exploration.

AI is your amplifier. In your specific track, you should be stronger than the AI. Then, use AI to boost your productivity. That is the correct way to use AI.

Finally, I hope everyone stops limiting themselves. Much like Reinforcement Learning, you might explore hundreds of steps without visible progress, but after synthesizing those steps, you will often reach a “eureka” moment where everything suddenly converges.