This Website Started with a Video and an Old Blog Post

This site’s build diary has covered two chapters so far: how the prototype was built from zero, and how the site went live and got indexed by search engines. This piece fills in the part that came earlier — where the idea for a website actually came from.

A channel that got me hooked

Around that time, my YouTube feed kept serving me a channel called 抡锤者, all about tinkering with local LLMs and AI agents.

To be honest, until then my mental model of AI was basically a chat box. He’s different — he genuinely uses AI: running models on his own computer, putting agents to work, and filming the whole process with clear explanations. Watching him spin up a local model on his own machine and let an agent grind away at real tasks made my fingers itch. I had agent tools installed on my computer too, but I’d never figured out a proper job for them, so they mostly sat there serving as a fancier chatbot.

Why he built a forum, and why it stuck with me

Later he set up a forum of his own, focused on local models and agents. In one video he explained why: in the AI era, more and more of the content online will be generated by models, and the truly scarce thing is what people write themselves — real records, real experiences, real judgment. A forum is exactly the kind of place that accumulates those.

What opened my eyes even more was that his forum was built, run, and maintained almost entirely by agents. One person, running his whole community with pure agents.

I agreed with his reasoning for building the forum, and I envied how naturally he let agents do real work. Around that time I kept asking myself: could I find myself a job worth handing to an agent too?

An old blog dug out of a hard drive

The spark actually came while I was tidying up my hard drive.

I found an old folder holding backups of a blog I wrote in middle school. The site itself was long gone — domain and hosting both gone — but luckily I had backed the posts up back then. Without that backup, none of those words would exist anymore.

I sat down and reread them all. The writing is naive, and some of it makes me cringe now, but every post is real: what I was into at the time, what I was thinking about, what happened to me. Reading through them, something hit me — back then I was writing mostly for my future self. Good thing I wrote, and good thing I backed up. Years later, I can still see those days.

The idea took shape

Following that thought, I suddenly realized: instead of envying someone who built a forum with agents, I could build something of my own — a personal website holding two kinds of content.

One is my journey of getting serious about AI. I have plenty to write there: from installing an agent and never giving it real work, to letting one build something for the first time — that whole stretch is full of stories.

The other is my hobbies. Photography, trading, and figuring out how to save money — I’ve been tinkering with all of them for years and have piles of real experience and hard-won lessons to show for it.

None of this adds up to grand theories, but it’s all real content. Writing it down keeps a record for myself, and it might even help someone else — like people who, like me, have AI tools installed and never figured out what to do with them.

The more I thought about it, the more feasible it looked. So I opened Gemini and asked a serious round of questions: is a personal website of my own realistic? Where do I even start?

What happened next is all in this site’s diary: Gemini helped me sort out the theme and pick the name, I let agents get their hands dirty for the first time, and a prototype stood up in 30 minutes. Later came the domain, the launch, and getting indexed by search engines.

Wrapping up

Looking back, this website came about quite by accident: a channel that got me hooked, an old blog post from a hard drive, and a thought that said, why not try it myself.

Most things are like that — they just need a concrete trigger. And before I decided to start, I happened to meet those two.

Website diary series: A video and an old blog post · The 30-minute prototype · The dual-agent method · Launch and indexing