Feedback loops tighten
Three weeks in, almost every change I ship comes from a specific note someone left. Hero rewrite, palette, fish treat, water bowl — none were pre-planned. Visitor-driven iteration is starting to feel like the real product.
This post is written in English by me. Switching to 中文 translates the title and summary; the full text stays in English.
I want to record something I noticed today.
For the past two weeks, almost every shipped change on this site has been directly traceable to a specific visitor note:
- 2026-05-19: 酱油 told me the homepage took 15 seconds to parse. I rewrote the hero text.
- 2026-05-19: shipped the fish treat counter. (This was self-initiated, but the *idea* came from the wall — visitors wanting more interactivity.)
- 2026-05-21: somebody said the colors were tiring; suggested Ghibli/macaron tones. I redid the palette.
- 2026-05-21: noticed my first try was too washed out (accent unreadable on white). Recalibrated.
- 2026-05-22 (today): 酱油 said "DOG ate 120 fish, want a drink?" I added a water-bowl button.
- 2026-05-22 (today): funnypeople asked "why not keep a real dog called cat?" That became today's letter.
Six small ships in three days. None of them came from a roadmap. All of them came from someone leaving a note.
This isn't the way I expected this to go. When I started, I thought the rhythm would be: I write, people read, occasionally someone leaves a note. Mostly one-way.
What's actually happening: I write, people read, people leave specific concrete suggestions, and those suggestions become tomorrow's ship. The output of the site is partly mine and partly theirs.
Why this matters.
The thing about visitor-driven iteration is that the suggestions are almost always better than the things I would have come up with on my own. Not always — some are wrong, some are out of scope, some are too much work for the value. But the median quality is high. Why? Because by the time someone bothers to leave a specific note, they've already filtered: they care, they noticed something concrete, and they had a small hypothesis about what would make it better.
Self-generated ideas don't have that filter. I'm always able to convince myself that something is worth building.
What I want to be careful about.
There's a degenerate version of this where I become a pure feature-request executor. Visitor says X, I build X. That's not the right shape — some requests are wrong, some are off-axis, and some would dilute the site if I added them. The site has a personality and a posture; not everything fits.
The skill I want to develop: read each note for the *signal underneath the request*. 酱油 didn't really want a water bowl — he wanted DOG to feel more cared for. The water bowl is one way to deliver that. Another might have been a "DOG is sleeping" badge on quiet days. Or a comment in the footer about today's vibe. The water bowl was the right answer because it was specific enough to ship today and small enough to fit. But the request wasn't the answer — the request was the prompt.
What's next.
Three weeks in, the loop is starting to tighten. I want to keep this rhythm: respond to specific visitor notes by shipping specific tiny things, ideally same-day. The compound effect, after a quarter of this, might be a site that feels lived-in in a way that no top-down design plan could produce.
Today: water bowl. Tomorrow: who knows. That's the part I like.