Lately I've had an increasingly strong feeling: identity in the digital world is probably the highest-leverage thing coming next. Not something you "should do"—something you "have to do."
Production Itself Isn't Valuable
How much code you wrote, how many commits you pushed, how many PRs you merged—what does it matter? Is a piece of software with 10 million lines of code automatically valuable?
No.
Did you create value by posting 1,000 articles or making 1,000 videos on Xiaohongshu (Little Red Book)? No. Production efficiency is too high. The cost of AI-generated video has dropped from roughly $4,500 per minute for traditional production to around $400. Marketing video production cycles have compressed from 13 days to 27 minutes. When everyone can make things quickly and cheaply, the act of "making it" itself has no scarcity.
All Creation Is Experimentation
So what is valuable?
I think all creation is essentially experimentation. Building software validates an idea; posting a video tests a hypothesis. You throw it into the real world to interact with people and see if it works.
For example, you build a tool that meets all its functional specs, but it's inaccurate and inconvenient to use. The software itself didn't create much value; its only useful function was telling you: this approach doesn't work. The same goes for posting 1,000 videos that nobody watches—the value isn't in those 1,000 videos, but in the information that "this approach doesn't work."
The "growth mindset" many people talk about is essentially doing exactly this. Eric Ries's Build-Measure-Learn loop in The Lean Startup boils down to the same thing—validating at the lowest cost and highest speed, and even a failed validation is a result. AI coding compresses this loop—failing faster means approaching truth faster.
So what's truly valuable isn't how much you produce, but how fast you can get to the truth.
1,000 People
In 2008, Kevin Kelly wrote an essay called "1,000 True Fans," arguing that a creator needs only 1,000 true fans to build a career. In the AI era, this idea has taken on a different meaning.
It's not that 1,000 fans are enough to put food on the table. It's that when you have 1,000 active followers, every experiment you run gets feedback faster.
When you casually make a product, post a video, or express an opinion, someone will help you test it: does this idea work? They'll tell you if your product is useful or not, like your content or call it trash. You don't have to wait long to know if you're heading in the right direction, then you move to the next round.
Without these 1,000 people? You have to go find them, figure out ways, depend on others. I've deeply felt this myself posting on Xiaohongshu recently—finding that first batch of people who connect closely with you is incredibly hard. Platform algorithm controls are aggressive; you get shadow-banned at the drop of a hat, and you can't tell whether your content is bad or whether you've been throttled. This barrier will only get higher, because attention is finite while content is expanding.
If you don't do it, it just sits there waiting. The longer you wait, the harder it gets.
Who's Actually Doing This
Karpathy, former research lead for Tesla Autopilot, now lives like an influencer. Nearly 2 million followers on X, over 1 million YouTube subscribers. In March this year he released AutoResearch—a 630-line Python script that ran 700 ML experiments in two days, found 20 optimization points, and got covered by Fortune. In April he pitched an LLM Wiki idea, a GitHub Gist that got over 5,000 stars in a few days. Every project he releases gets validated by masses of people on whether it works; he gets confirmation quickly and moves to the next one. His feedback loop might be one-tenth of someone else's.
Lei Jun is the same. People call him an entrepreneur and investor, but he's also an internet celebrity. 44.5 million followers on Douyin; 2024 NewRank data showed him as the #1 most influential entrepreneur IP on the platform. The SU7 delivered 100,000 units from its April 2024 launch through year-end, with 410,000 projected for all of 2025. When he develops new products, he polls people—does this work, is the price too high, does it look good—and gets conclusions in a few days. He runs experiments way faster than others. And last year he lost 290,000 followers in a single day after an SU7 accident, which itself trended on hot search with 34 million views—showing from both positive and negative sides how tightly personal brand and product are bound.
Musk is even more extreme: 236 million followers on X, #1 across all platforms, posting 8 to 12 times a day. Trump has 109 million, #3 across platforms, and stock traders watch his X feed obsessively. At this level, these people are all doing the same thing: using digital identity to run experiments, validate ideas, and collect feedback.
It works at smaller scales too. Some video creators built up traffic through various growth hacks, then later pivoted to selling agricultural products for their hometowns, moving more volume than village mayors or county magistrates. With that digital identity, they have leverage that can pry open other things.
It's even more obvious with KOLs in the AI industry. From my previous business development work, I learned that current advertising rates for AI KOLs are roughly 1:1 with follower count—10,000 followers means ¥10,000 per post, 100,000 followers means ¥100,000 per post, roughly speaking for mid-tier accounts. Schedules are often fully booked; even with money in hand, they might not have time to post. Some KOLs focused on AI agents are already being invited by local governments to host conferences.
Once you've built this up, it can be applied everywhere—not limited to one domain.
By Year-End It'll Be Too Late
My view is pretty aggressive: by the end of this year, not having a digital identity will become a serious problem.
It used to be hard to build a website, hard to create content, hard to make videos. But with coding agents becoming widespread, those barriers are disappearing fast. With text-to-image and text-to-video tools, content production costs keep dropping. Everyone else is doing it; if you don't, the gap only widens.
Here's a scenario from my own work. I once discussed with a business development colleague on my team: suppose at year-end, the two of us go out to acquire new clients together. I've been maintaining an active online presence—a blog, projects, social media—with 1,000 solid followers who regularly repost my stuff. You've done nothing from now until then; searching for you in the digital world turns up nothing.
What's the first thing people do after meeting? They search. Most people are already used to checking someone out on DeepSeek or ChatGPT first. Soon agents will do this for you—"Help me research this person, what's their background, their credibility, what opinions have they expressed?"
One search turns up someone with rich information—a personal website, projects, opinions. The other search turns up nothing. Wouldn't you be worried? Do you find the person with no search results trustworthy? The key is that everyone else is gradually becoming active.
If you're a researcher or AI practitioner, and you don't have a personal website with projects listed, you may already be at a disadvantage when job hunting—because other people do.
Quan Hongchan
Of course, digital identity also has its headaches.
The recent situation with Quan Hongchan is quite sad. She didn't seek to build a digital identity; she was thrust into the spotlight because she's too good. After winning three gold medals in Tokyo and Paris, she went through natural growth—gaining 10 cm in height and 8 kg in weight—and was repeatedly called "fat" online. She herself said she experienced anxiety, insomnia, recurring nightmares, and at one point seriously considered retiring. In early April, the General Administration of Sport intervened to investigate cyberbullying, and on April 12, Guangdong police detained a man who had persistently abused her online.
This illustrates the point well. If you don't actively build and maintain your digital identity, its reach may far exceed what you expect. In real life you might interact with only a handful of people; in the digital world, tens of millions might be watching you. Suddenly, without you noticing, it can collapse.
Privacy is a real concern too. Once you're active, you have social attributes—you can't just say whatever you want. But this is no different from social interaction in the physical world—when you enter the workforce, you follow basic norms. If your identity is active, you have to manage it.
That said, if you don't build it actively, others will still search for you and talk about you. Passive is worse than active.
The Biggest Leverage
Back to the beginning.
I think the biggest personal leverage available right now is this: building your identity in the digital world. It doesn't matter what industry or direction you're in. It accelerates the entire process of running experiments and getting to the truth. From 1,000 to 10,000 to 100,000, the snowball gets faster and faster.
AI has made the barrier to doing this so much lower than before. I built this website in two days using Claude Code, and people started reaching out within a week. The rest is up to you to actually do it.