Original video in Chinese.
Key Takeaway
- AI clearly divides people into two groups: one group uses lagging AI (free ChatGPT-level tools) and thinks AI is a scam and a bubble; the other group uses advanced AI (Claude Code, multi-agent setups, heavy paid usage) and gets excited every day creating value with AI. The cognitive gap between the two sides is widening rapidly.
- There are two major mindset traps for those who have fallen behind: first, “It has nothing to do with me” (they feel AI’s development path has nothing to do with them and are unwilling to adapt proactively); second, “It’s not worth paying for” (they are unwilling to spend money on top-tier productivity tools). At its core, this is the result of a “consumer mindset.”
- You must complete a mindset shift: from consumer to builder, and actively go all in on advanced AI tools, systems, and practices. AI is an elimination round, and those who do not transform will be quickly left behind. The author and Karpathy strongly warn about this.
Because of the emergence of AI, there are now two kinds of people in the world. And the cognitive gap between the two sides is getting wider and wider.
One kind uses lagging AI. They think AI doesn’t work, that AI is a scam, that AI is a bubble.
The other kind uses advanced AI. They think AI is advancing too fast, they are very, very excited about it, and every day they are using AI to create new value.
This is Karpathy’s latest view. I really couldn’t agree more. In fact, earlier this year I also expressed a similar point of view:
Those who use AI and those who don’t are two different species. Those who use Doubao-level AI and those who use Claude Code-level AI are two different species. This year, multi-agent orchestration will screen out another wave.
I’m being a bit sharper, while Karpathy is more considerate of other people’s feelings. In his tweet, he said that the typical representative of the former group is someone who tried free ChatGPT last year and then felt like they understood AI, thinking that hallucinations were severe and the intelligence wasn’t high enough.
That group isn’t exactly wrong either. AI really was like that before, with plenty of problems. But today, the most advanced models (such as Opus 4.6) and the most advanced tools (such as Claude Code) are no longer at the level of the chatbots of those days. A week’s worth of work for a team can now be done by AI in half an hour, and at a much lower cost.
And the second group are the people who use agentic AI every day. They don’t just use it, they max out the quotas too. A $200 subscription may not even be enough; they’ll open several at the same time. That feeling of running alongside technology and alongside the times is just so satisfying!
So these two groups are both not wrong in their understanding of AI; they just picked the wrong version and are living in two different worlds. The former group has already fallen behind, and it’s very likely they will never catch up again. Why?
I think the most fundamental reason is their mindset: they have always held on to a “consumer mindset” and have never been willing to shift to a “builder mindset.”
On the issue of AI, a “consumer mindset” creates two traps.
First, the trap of “it has nothing to do with me.”
This wave of AI development broke through via programming. Programming ability is a necessity for AI; I explained this in a previous video.
However, what most people need is an AI that can help them write articles and make spreadsheets, not an AI that can program. So they think this has nothing to do with them, can’t be bothered to understand the logic behind it, and then naturally miss it.
The problem is that something at the level of AI has its own development path. If you always think of yourself as a consumer, then all you can do is wait—wait until AI’s path intersects with yours, wait until AI comes to find you.
If you want to enjoy the dividends of AI, you must shift to a “builder mindset” and proactively go looking for AI.
How do you go looking for it? I’ll give an example: a knowledge base.
More than half a year ago, I made a video suggesting using Cursor to operate an Obsidian repository. Later, after Claude Code matured, I suggested switching to Claude Code, and even building an AI operating system based on it. Now almost all the top AI-savvy creators are using this method too.
Before, organizing notes, adding tags, and building indexes would take several hours a week. Now I say one sentence to AI, and it’s all done in a few minutes.
So once your mindset shifts and your role shifts, you’ll discover that AI can actually already serve you—it’s just that the path is different from what you imagined.
Second, the trap of “it’s not worth paying for.”
Today, quite a lot of people are unwilling to pay for AI. Let alone the $200-a-month Max version, even a $20 subscription feels not worth it to them.
Why is it not worth it?
Feeling that it’s not worth it is, at its core, because you can’t create value with it. And the reason you can’t create value is that you have never regarded yourself as a builder.
A consumer doesn’t need to think about, and won’t think about, a question like “creating value.” A consumer just consumes.
So, “not worth paying for” is absolutely not an AI problem; it’s these people’s own problem.
Or think about it this way: AI is the most advanced productivity tool right now, right? After this most advanced productivity tool is placed in your hands, you still can’t create value with it in work and life, and you still don’t use it—how is that not a frightening, deeply worth-reflecting-on problem?
The two traps I just mentioned are traps I’ve seen many people already fall into. In this video, I just want to give everyone a reminder:
When you feel that AI has nothing to do with you, when you feel that spending money on AI isn’t worth it, you really need to be careful.
This AI transformation is an elimination round—it is eliminating countries, and it is also eliminating individuals.
OK, that’s all for this episode. If you want to understand AI, want to become a super individual, and want to find like-minded people, come join our newtype community. See you in the next episode!