The Second Sidetrack

At the end of my last post, I mentioned that Fitting Room, my paper-doll outfit experiment, was sitting quietly on my laptop, untouched while I got completely swallowed up by building this blog.

A week later, I have an update on that: it is still sitting there. Quiet as ever.

Fitting Room received exactly zero minutes of my time this week. Right after wrapping up the blog build and writing a few new pieces, I got sidetracked all over again. Or, to be entirely honest, I circled back to an older sidetrack that started before all of this.

Back in late August, while I was back home in Taiwan, I started trying to build a website for my Instagram account, @love.in.hongkong. At the time, I had never even heard the term "vibe coding." I was simply curious to see what a new AI agent could do.

It didn't go well. I hit wall after wall, got hopelessly stuck, and felt genuinely disappointed because I had been so excited about having a new site. Seeing my frustration, my brother suggested I learn Figma first to sketch out wireframes. But a brand-new teaching semester was right around the corner. The thought of adding a steep software learning curve on top of intense teaching prep was exhausting. So I shelved the whole project and told myself I'd get to it later.

Then came this blog. Building it with Claude in seven days proved what kind of execution power Claude actually had.

So I dug that shelved Instagram project back out and handed it over to Claude. That is where all my hours went this week.

Why build a standalone website for an Instagram account in the first place? On @love.in.hongkong, I take people exploring Hong Kong on foot, with all the tricky transit and walking logistics already mapped out. But Instagram is a terrible library: once you have dozens of reels, it is nearly impossible for followers to search specific neighborhoods or browse past routes. A website solves that instantly. It acts as both a curated walking guide to Hong Kong and a clean, searchable portfolio for my work.

The account is currently sitting at over 9,000 followers. My goal is to officially launch the site when I hit the 10,000 milestone. At the current growth rate, that should happen sometime in November, which means the site needs to be ready before then.

As for Fitting Room? It will just have to snooze in my laptop a little longer.

The Ride to the Forest

Even after this blog went live, one lingering question stayed on my mind: how on earth did the first website come together so fast?

Seven days from a casual Wednesday chat to a live website, complete with three archived Chinese books and three new blog posts, built by someone who cannot write a single line of code.

The answer clicked on Sunday.

I was on my way to a forest-bathing workshop, and shared a car with the forest-bathing guide and Agnes, one of my design students. Winding down the road, the conversation turned to my latest obsession outside of Instagram: vibe coding.

When I mentioned that I built and launched the entire blog in just seven days, both of them gasped in disbelief. Neither of them comes from a tech background, so hearing that an absolute non-coder built a website in a week sounded like straight-up black magic.

Watching their amazed faces, I chuckled and shook my head.

"It really isn't about understanding code," I told them.

It wasn't magic, and I certainly hadn't turned into a computer whiz overnight. If you pasted the site's raw source code in front of me right now, I still wouldn't be able to tell you what the lines mean. That speed had nothing to do with programming. It had everything to do with how my day job has trained me to work for more than twenty years.

How Landscape Architects Work

By profession, I am a landscape architect. I have more than twenty years of professional experience, many of them spent in fast-paced international design firms, juggling tight deadlines and complex, high-pressure master plans.

In landscape architecture, there is one non-negotiable reality: we almost never work alone. Teamwork isn't corporate fluff you put on a slide; it is simply how we survive every single day.

Decades in that environment gave me some very specific instincts:

First, I have an eye that hunts down reference images at lightning speed. When a concept sparks in my head, I don't waste time trying to describe abstract geometry in thin air. I immediately pull together the right reference images that capture the exact mood, materiality, spatial proportions, and human feeling of the space.

Second, I know how to give crisp directions and design markups. I will not drop a pile of pretty pictures on a junior designer's desk in total silence, expecting them to read my mind. I pull out a red pen, circle specific zones, mark them up, and articulate why I want a change—whether for aesthetic balance or for how a person will actually experience walking through that space.

Third, I always gauge a teammate's capabilities before handing off a brief. When a new designer joins the team, I don't dump tasks on them blindly. I ask questions, test what tools they know, and figure out their strengths and limits. Only then do I know how to frame the task, explain my expectations, and set realistic benchmarks.

I'm also completely comfortable working with people who aren't sitting next to me. During the pandemic, we collaborated remotely as second nature, and I have spent years teaming up with colleagues stationed in other offices. An AI agent is really no different: a colleague you've never met in person, collaborating across a screen.

So when I sat down to vibe code, I wasn't acting like an intimidated beginner desperately memorizing syntax. I was acting like a creative director briefing her team.

A Very Good Question

That dynamic showed up clearly while building the landing page for this blog.

When Claude and I started, my very first question was not "Can you build this page?"

I asked: "Can you read live external web links?"

I wanted to pull reference sites from the web for it to look at, so I needed to know our communication boundaries upfront. If it couldn't read live URLs, my entire workflow would have had to change—taking dozens of screenshots, typing out tedious layout breakdowns, or finding messy workarounds.

Once Claude confirmed that it could browse live links, it did something that immediately caught my attention. Instead of rushing to spit out generic boilerplate code, it behaved like a seasoned senior designer. It turned around and asked me to explain the specific reasons why I liked each reference site I had shared.

That was a very good question.

Because it asked the right question, I gave it an exact, detailed brief: which visual elements matched my aesthetic, how the user journey should flow, and the overall mood we wanted to strike. With that clarity, Claude nailed the structure.

What followed was the familiar back-and-forth rhythm of my day job: refining layouts, nudging font sizes, tightening margins, and polishing the user experience—exactly like marking up landscape schemes with my design team.

And during that iteration, something fascinating happened. Many of my markups were driven purely by aesthetic intuition: "This feels a bit cramped," "Give this block more breathing room," or "The balance here feels off." I was communicating in sensory, design-led language. Claude would immediately understand, validate my observation, but then reply in its unmistakably machine voice: "Yes, I measured the pixel spacing and set up a rule to maintain that proportion."

I loved that contrast. My human aesthetic eye was being backed up by machine precision. What I felt intuitively but couldn't articulate in technical terms, Claude locked down with exact, pixel-perfect rules.

(Writing has its own version of this workflow—which is how six posts came together in two weeks—but that experiment deserves a post of its own.)

George's Half and the Right Teammate

Earlier this week, when I told my friend George about the seven-day turnaround, his immediate reaction was: "Well, that's because you have a strong AI."

George's half of the argument is definitely true. If I had stayed stubbornly chained to that first AI agent from August, this website wouldn't exist yet.

Yet a powerful AI is still only half the equation. Hand a brilliant AI to someone who doesn't know how to brief, how to select reference images, or how to mark up revisions, and it will just produce polished clutter. On the flip side, more than twenty years of design experience won't get you a live website in seven days if your teammate can't write a single line of code.

My own takeaway comes down to four interlocking pieces: the right tool, the right teammate, good teamwork, and an efficient workflow.

When those four line up, the friction between having an idea and bringing it to life disappears. My professional instincts and my AI teammate complement and unlock each other. Not knowing how to code is no longer the bottleneck holding my creativity back; instead, the ideas in my head are making their way out into the world faster than ever before.

As for my Fitting Room app? It's still resting quietly on my laptop. November is creeping up fast, and I have another website to finish first.