A Developer Builds a Personal Site When AI Reads Everything—and It's Not a Losing Bet

In the AI age, few humans read. I started building my own independent site, but is it really joining the Nationalist army in 1949?

Author: king王一帅

A few days ago, I posted a status update on a platform, announcing a writing-publishing tool I was building. Some commenters said it was useless, and I'd be better off building a coupon snatcher; others said I was just self-indulging.

I wasn't angry or bothered, but it made me think for a long time. Readers on the platform come from all walks of life and care about different things. The same thing can have completely different value in different eyes, which is normal. What truly concerned me was something else: did the things I want to write have to go into a recommendation feed and wait for an uncertain outcome?

That made me re-examine an old question: in the AI age, what place does a personal blog have, and what can it do?

Let me state the ugly truth first: I know it's foolish to start an independent blog right now.

Platforms have recommendation systems, AI has summaries, and people don't even bother to click on search results. A new domain has no authority, no backlinks, no recommended slots. Writing a long article and publishing it typically results in single-digit read counts. This isn't a matter of 'content not being good enough'; it's the distribution structure. So the question isn't whether blogs have value, but: in an era when nobody directly visits websites, why would I still build my own site?

The homepage of yishuai.wang.

It does look like joining the Nationalist army in 1949.

This metaphor is half true

The half that's true: if you treat an independent site as a traffic business, it's an uphill battle. Relying on SEO natural traffic, monetizing via ad networks, and maintaining authority through bulk updates—this playbook has no chance once AI pushes content supply to infinity. AI can generate a press release in a second; you could write ten articles and nobody would read them, and search engines might not even index them. In that sense, joining the Nationalist army in 1949 isn't a joke.

The half that's false: joining the Nationalist army in 1949 was betting on a collapsing old order; I'm not betting that independent blogs will become traffic centers again. I don't think independent sites will return to the 2010 golden age of 'write a blog—get indexed by search engines—get traffic—run ads.' That era is definitely over.

What I'm betting on is something else: the more AI content floods, the scarcer first-hand experience, real projects, and verifiable signatures become.

It's too early to conclude. AI can help me look up information, fix code, and polish sentences, but it hasn't done the projects I've done, hasn't fallen into the pitfalls I've fallen into, and hasn't spent four months pixel-by-pixel fixing pagination layout. It can't write judgments like 'why I decided not to use an off-the-shelf solution' because it has no 'then.' But whether anyone will appreciate this, I don't know; I can only do it and see.

It's not that nobody reads—maybe the reading structure has changed

The statement 'in the AI age, few humans read' is half right. More precisely: the probability of someone directly visiting an unfamiliar independent site has decreased, but the paths by which people find you via AI search, AI summaries, GitHub, developer communities, and RSS may have increased.

Readers may not have disappeared; the entry points may be changing. Previously the entry was a search engine results page; now it could be:

• Someone asks AI "how to make a Chinese pagination editor," and AI cites your article.
• Someone sees QuireKit on GitHub and follows the trail to your site.
• Someone on V2EX, Jike, X, or Juejin sees a troubleshooting post and clicks through to read the full version.
• Someone subscribes to your site via RSS, reads every article, but never comments.

None of this shows up in 'homepage visits.' But this is only my observation, not necessarily correct. There's also a trap that must be stated clearly: being read by AI ≠ being cited by AI ≠ being driven traffic by AI. Putting up an llms.txt and thinking AI will spread your content is self-comfort. llms.txt is not a standard; many AIs don't read it, and even if they do, they don't necessarily cite it. Being trained into a model doesn't mean anyone will visit your site. So I don't treat llms.txt as a strategy, only as a guide. What I'm trying to do is: make the site more like a citable primary source, and ensure that citations include my name, link, and tools. Whether I'll succeed is another story.

Some very personal reasons

yishuai.wang is my name. I love this domain but never used it properly.

Over the years publishing on various platforms, I've also felt that some platforms are becoming more platform-like, with recommendations, trending lists, and traffic rules taking precedence, while the technology itself takes a back seat. V2EX is a place I feel has stayed true to its original purpose, and it's the community that has helped me the most. If I ever get the chance, I'd like to build a community like it in China. That's still far off; I'll start by writing my own site well.

So I rebuilt the site and finally put the domain to use. I'm leaving myself a piece of pure land: I don't have to write around recommendation rules; I can write whatever I want.

What I'm trying to do

These are all concrete and unproven things.

First, write what I want, but have part of the site carry the 'why me.' I'll write engineering logs, essays, opinions, tool sharing. Others can write those too; I won't pretend every post has unique value. But there has to be a part of the site that no one else can replace: QuireKit's engineering logs, pagination layout pitfalls, why I didn't use off-the-shelf solutions, what it's like to reach 85% compatibility with Word's common features in four months. Not every post is like that, but they define 'why me' for this site. The rest I'll write freely for joy, but I won't treat them as a moat.

Second, make the independent site the first-publish destination. From now on, articles will first appear on yishuai.wang, then sync to various platforms, with full content and no cuts. If there are revisions or additions, the on-site version gets updated first. Platforms that support canonical tags will point to the site's original; those that don't will get a note at the end: 'This article first appeared and is continually updated at yishuai.wang.'

Third, prepare a citable structure for both AI and humans. In long articles, add conclusion cards, fact tables, FAQs, and change logs. For QuireKit, create a separate tool page that spells out a one-line definition, features, boundaries, license, comparisons, and roadmap. What AI likes to extract is usually not long narrative, but conclusions, facts, versions, comparisons, code, and boundaries. A tool page may be more easily recommended by AI than a blog post. This is just a hypothesis; I'm acting on it.

Article page: an AI summary at the top, table of contents on the right, and every article's original Markdown is directly accessible.

Fourth, keep RSS. An independent site has no platform recommendation; RSS is one of the few connections that truly belongs to me: it bypasses algorithms, and you get updates as soon as they're published. There are separate feeds for Chinese, English, and all content. Whether anyone uses them remains to be seen.

Fifth, don't fight AI; treat AI as a distribution layer. I don't expect to stop AI from reading my site, and I can't. What I can do is make what AI reads verifiable, signed, and linked. People see AI's output; I want that output to include my name and my tools.

This site is built with my own tools

I know exactly what QuireKit is and what it can do, so I'm willing to eat my own dogfood: this site is built with it.

QuireKit is an open-source document editor ecosystem I'm building. Its flagship is the Classic paginated document editor: it does real pagination in the browser, computing line breaks, page breaks, cursor, and selection itself, so you edit page by page like Word or WPS. The other line is MSFE, an editor framework tailored for Markdown, where the Markdown source is the document.

yishuai.wang uses it for everything, from pages to writing:

• Pages use QuireKit's component library and the InkWash ink-wash theme.
• The writing backend and comment box are both MD Editor (based on MSFE), with document, live preview, and source views switchable at any time.
• Add .md to any article's URL to get the original text. Combined with Chinese and English RSS, canonical, and llms.txt, that forms the "citable structure" mentioned above.
• The AI summary at the top of each article is generated once after publication; readers don't trigger model calls during reading.
• The site also supports Micropub, so you can post directly from other writing tools in the future.

The comment box is MD Editor; the footer has RSS and llms.txt.

The site's source code is also open-sourced (MIT): https://github.com/kingshuaishuai/yishuai.wang. Use your own thing first; I'm the first to know what's not ergonomic.

So, is it joining the Nationalist army in 1949?

I think it's not. But that's only my personal opinion and doesn't constitute advice to anyone.

The Nationalist army bet on a collapsing order; I'm betting on something else: when AI pushes content supply to infinity, a person's experience, judgment, signature, and verifiable first-hand records may become more valuable. An independent site may have no readers, but at least it's a source I leave for myself. The domain, articles, and version history belong to me, not subject to platform algorithms or swept away in an AI summary.

There's another reason. Good Chinese technical blogs are already scarce, and how well large models can answer depends largely on what they can read. If most materials are reposts and fluff, no matter how strong the model, it'll be limited by that content ecosystem. One person's site is small, like a star; but a good content ecosystem is made of many stars. I want to be one of them.

Having no readers is the norm, not a verdict. I'll first write a batch of things only I can write, then write more things I want to write, and then see if it deserves to live.

AI can read it, cite it, or ignore it. But the source is here.

As to whether this is joining the Nationalist army in 1949—I don't know right now. I just feel that instead of waiting for a definitive answer, I'd rather leave my own piece here first.

---

This article first appeared and is continually updated at yishuai.wang.