The question friends and colleagues have been asking most lately: Vibe Coding sounds pretty cool, but where exactly do you start?
It's a good question. AI Coding is a productivity tool, not an entertainment tool. If you try to tinker with it in daily life, you'll mostly be at a loss. The vast majority of daily needs have already been saturated by cheap or even free apps over the past decade or so. Want bubble tea? There's Meituan. Want to edit videos? There's Jianying (CapCut). Why vibe-code another one yourself?
My answer only offers two paths: either start from work, or start from building your own digital identity. There is no third path.
1. Start from Work
Let me first explain why work.
Tools should grow wherever human attention is focused. Whether it's eight hours or ten hours a day, that's the time you're compelled to produce value. The pressure is high, the feedback is direct. There's no better testing ground for transforming your work paradigm.
Scenarios like playing ball, cooking, or binge-watching shows seem more "free," but the actual time you invest weekly is pitifully thin, and your attention isn't sufficiently attached. Vibe coding doesn't easily take root in these places because your mind isn't really there.
Work is the opposite. You spend the vast majority of your time immersed in it. Off-the-shelf tools mostly cost money and aren't particularly good. Output gets validated by others, feedback doesn't disappear. Room for improvement is concrete, not imagined.
My own metric is AI penetration during work hours. Early on, I was probably only spending 10% of my time interacting with AI, with the remaining 90% being my original way of working—that state basically equals zero penetration. Later I gradually pushed it above 80%. When penetration is high, the only things left in my day that still use the old paradigm are purely human activities like meetings, signing off, and talking with clients.
This number is actually closer than you might think. In Stack Overflow's 2025 Developer Survey, 84% of developers are already using or planning to use AI tools, with 51% using them daily. Anthropic's own Economic Index data is even more direct: 46% of conversations on Claude.ai are work-related, and on the API side it's 74%. Heavy users treat it as a work partner, not a toy.
But pushing penetration from 10% to 80% isn't just a matter of adding AI to your workflow. It requires breaking apart your current tasks and redistributing them: which segments let AI write, which let AI research, which let it run on its own, and which must be monitored by humans. This restructuring is the most natural training ground for getting started with vibe coding.
2. Start from Building Your Digital Identity
If you can't find an opening at work right now—say the process is too rigid, or the team isn't ready yet—the second path works just as well. Build your own personal digital identity.
Why place it second, outside of work? Because its ROI is absurdly high.
The investment is so small it's barely worth calculating. A personal website can now be built with vibe coding in a day or two.
The potential payoff is compounding. What you write is yours, not the platform's. The same piece published on Zhihu, WeChat Official Accounts, Geekbang, X, and Xiaohongshu (RED) represents five completely different exposure formats. Each platform's algorithm is different. The probabilities of hitting a viral piece stack together, making it far more stable than betting on a single platform.
I cold-started for just over a month, averaging about 2,000 daily views across all platforms. This number actually means very little. What matters is the distribution. I had an article about software bidding and procurement that got pushed to over 20,000 reads and hundreds of comments on Zhihu. Many other articles sit at just dozens to hundreds of views on each platform. But as long as your inventory is large enough, one or two will eventually hit. That's the compounding effect of distribution.
Here's a counterintuitive fact. profy.dev surveyed over 60 technical hiring managers: 93% will look at a candidate's portfolio website, but 51% flat-out said "not having a portfolio website doesn't lower a candidate's chances." What does that mean? Static showcase "vanity sites" are largely useless. What actually works are vehicles that let content gradually accumulate. GitHub profiles, blogs, project notes—these things with "output traces" are what constitute a digital identity.
Applied to vibe coding, this path is now much easier to walk. AI has compressed the most tedious parts of the content pipeline to near-zero cost: writing, reformatting, adapting for different platforms, generating different media. What used to take half a day for one person to spread an article across five platforms now takes ten-plus minutes. Even investing just 30 minutes daily adds up significantly over the long term.
As for "why you should build a digital identity," I wrote a previous piece that covers it more completely, so I won't repeat it here.
3. The Prerequisite: It Must Be the Real You
My friend followed up with a second question that I think deserves more elaboration than "where to start."
He said: I feel a bit hesitant about putting my own stuff out publicly online. What should I post? How should I post it?
My answer has only one rule: post, but what you put out must be the real you.
Don't perform, and don't chase traffic by imitating a style you don't even endorse yourself. This is where long-term ROI most easily crashes.
Many people misunderstand, thinking that building a personal identity means "performing a better version of yourself." This path only works for full-time influencers who depend on traffic for their livelihood. That's their survival mode. For most people, digital identity is a compounding bonus beyond work, not their main livelihood. Once you start performing, you incur an invisible liability: every piece you publish is a "raw archive" that may be dug up and cross-examined someday in the future.
The fact that the internet has memory is, I think, grossly underestimated.
The classic example is Justine Sacco. In 2013, she casually tweeted before flying to South Africa, with only 200 followers at the time. Eleven hours later when her flight landed, she was trending worldwide amid outrage and was fired upon returning to her company. Follower count is no shield. James Gunn was fired by Disney in 2018 over old tweets from 2008–2009—a decade prior. Kevin Hart stepped down as Oscars host the same year, also over decade-old jokes that were dug up.
Once your influence crosses a certain threshold, past content gets examined under a magnifying glass.
So before I publish anything, I ask myself: at that point in time, is this the real me? Is this an opinion I'm willing to claim later when I look back?
If yes, publish. Even if your thinking changes later and you find it naive—that's fine. People naturally change, and looking back at youthful naivety is normal; it's part of your history. But if you couldn't even endorse what you wrote at that point in time, looking back it becomes a stain. These stains accumulate and become constraints holding you back.
Posting your authentic self has another hidden benefit. It forces you to think clearly about "what is the real me." The value of this process itself may exceed the exposure the content generates.
Not Posting Means Zero
Back to my friend's original question.
Two paths: starting from work is the most natural, since your attention is already there; starting from digital identity is the most cost-effective, with small investment and compounding returns. Whichever one you can start doing immediately, take it—it's far more effective than agonizing for six months.
On digital identity, I also said in my last piece: if you don't do it, the opportunity just hangs there waiting. The probability of any event hitting you is zero. If you don't buy a lottery ticket, the probability is zero; if you buy one, at least there's a probability.
But rather than grand narratives like "digital identity is leverage," what I want to emphasize more is: start today, don't wait. The barrier is already ridiculously low. Few things are so universally applicable across industries and effective for everyone. There's only one prerequisite: that it's the real you.
Everyone should do it. Musk, Lei Jun—they're all doing it, even Trump is doing it. You're using the same infrastructure, the same medium as them. There's no reason you shouldn't do it.
References
- Stack Overflow Developer Survey 2025: AI Tool Usage
- Anthropic Economic Index — January 2026 Report
- profy.dev: Portfolio Websites Survey — 60+ Hiring Managers Surveyed
- Justine Sacco Tweet Incident (2013) — CNN
- James Gunn Old Tweets Dug Up (2018) — Variety
- Kevin Hart Steps Down from Oscars (2018) — NPR
- GitHub Copilot Adoption in Fortune 100 — GitHub Official Docs
- Cursor Growth Data — Sacra Analysis