Welcome to Carb Load – a regular collection of my thoughts on technology, cycling, food, fashion, flowers, and whatever else I can throw in that catches my attention.
This week’s issue is a little shorter than normal — I’m sharing how I rebuilt my website for Wave Floral, my floral business, using AI via Claude Code.
I’m heads down in San Francisco for work, so if you’re itching for more carbs, I encourage you to read my piece published last week in the The Sweat Lookbook and while you’re at it — subscribe to Lee's excellent coverage of all things sports, fashion, and storytelling, with the sharpest eye in the game.
My main goal this fall was to get Wave Floral’s name out there. For small business owners, it’s common to hear, “wow, I love your stuff—I had no idea you did that!” The adage holds: if you’re not tired of your message, you’re not saying it enough.
I treat Wave as a marketing lab: experimenting with pop-up bouquet bars, installs for restaurants, and occasionally going viral on social media. But none of that matters if people can’t find or use your website to actually order flowers.
So, I decided to climb back into the sandpit that is redoing a website.
Instead of faffing with Squarespace and endless tutorials, redesigning my website was easy and fun thanks to working with AI.
Over a few late nights and one weekend—and less than ten hours total—I shipped a new version of the site that felt eye-catching, useful, and most importantly, representative of the Wave Floral brand and what we want to offer to our customers.
Our new website, paired with a flyering campaign through downtown, has resulted in consistent inbound traction (new inquiries, regularly!) the last few weeks, vs. prior when it was a ton of outbound and demand generation (reaching out to businesses and repeat clients.)
I don’t think this would have been accomplished in the timeframe needed — if not for working with an AI coding assistant.
I share this partly to push back against the two extremes I see: the “996 grindset” technologists trying to both escape and create the “permanent underclass” vs. folks, typically non-technical, who broadly dismiss AI as labor-erasing. I straddle both and want to build a bridge.
I didn’t steal anyone’s job—I was never hiring a developer for this. Could I have done it without AI? Sure—but it would’ve taken months, because realistically I wouldn’t have had time to build a new website for Wave with a full-time job and tons of hobbies.
This is the second essay in my series on learning to program and building in the midst of a software and tech paradigm shift.
Rebuilding showed me the real leverage: combining (1) design thinking (what), (2) fundamentals and execution (how) and (3) curiosity (why) to build something – because a website doesn’t just showcase your work, it makes it real.
free range apps
This is the first essay in a series exploring my journey as a self-taught programmer, going from non-technical founder to shipping production code as co-founder of Whimsy. This piece is my vision for how AI-assisted development could reshape not just who builds software, but what kinds of pr…
If you’ve ever:
Struggled with Squarespace or didn’t want to pay for it
Been semi-technical and wanted help building your own application
Been non-technical but curious about learning to program
Had a side project you want to ship/publish
This one’s for you!
Step 1: Get familiar with web design fundamentals
Before working with Claude Code and other agentic programming tools, I spent time understanding web design basics and learning programming “from scratch” by taking courses. This made it much easier to ship working code and get a website running quickly, and I don’t think people can or should skip this before using AI to build something external-facing.
Essential knowledge to acquire first:
HTML & CSS fundamentals - Not everything, but enough to understand structure vs. styling and not stare at an index.html document (the backbone to every website). Codecademy's Web Development courses are a great resource.
Basic file management - How to organize folders, work with file paths, understand how websites are structured. The Odin Project's Foundations course is excellent for this.
Command line basics - You'll need to push code to GitHub and troubleshoot issues. Learn basic commands: cd, ls, git add, git commit, git push. Once again, highly, highly recommend The Odin Project for the most comprehensive tutorial into how to work with your computer and start writing code.
Design principles - Study websites you love. What makes them work? How do they use color, spacing, typography? Start noticing patterns before you try to build them. Some websites I like to use:
Step 2: The tools I used
Armed with a basic understanding of fundamentals and familiarity with writing basic HTML/CSS, now you're ready to actually build.
Below is a list of the tools I use to ship simple sites quickly (shoutout to Canzhi and Nishant Kheterpal for setting me up on these years ago before I started learning how to program):
Essential Tools:
GitHub account + Pages: Code repository, free hosting for your website.
VSCode: My preferred IDE, have also used Cursor and Windsurf
Claude Code: $17/month; other coding agents rates vary
Squarespace (custom domain): $15/year (optional but worth it if you want your website to appear as www.yourwebsite.come vs. www.yourwebsite.github.io)
Hosting via GitHub Pages: You write code, push it to GitHub, enable Pages in settings, and your site is live. Sites on the free tier have public repositories (i.e. everyone can view your code) but you can upgrade to keep your repo private.
If you’re looking for a more robust yet simple option for hosting, I use Vercel for my startup and Netlify for my side projects (e.g. hawkhillpr.com).
Step 3: Curiosity drives, AI accelerates
After getting set up with your tools and firing up a coding assistant in your IDE of choice, what’s next?
This is where most people go wrong. They treat AI like a magic button instead of a collaborative tool that amplifies their understanding. It’s important to come in with a clear spec of what you want your site to do, mock-ups of rough layouts, and specific design elements.
It’s tough, because you have to both have an understanding of where it goes wrong, what it can do beyond your limitations, and when it does something beyond your limitations, you need to have the discipline, time, and curiosity to investigate it, reproduce it, and create something ultimately even better.
But as you do this, it’s an incredible compounding process that I believe, accelerates learning for most in their early stages of development.
I started by collecting inspiration, and I think this is one of the biggest differentiators and accelerators in AI development. Being able to identify and say certain characteristics on a website and web design is integral to prompting and working with AI partners correctly.
What I wanted to get from the inspiration was:
Layout of landing pages — full-screen images vs. grid-style galleries
Navigation: Types of menus (lists vs. hamburger), types of links
Service pages: How different florists describe their services to the public
Some sources of inspiration:
For each site, I noted specific elements: "I love how this landing page uses full-screen flower images" or "This navigation is clean but bold."
Building a new website
A scientific method, if you will:
Collect inspiration – be specific and take notes + screenshots.
Inspect what works – use browser tools to see how a site achieved that hover effect or grid layout. Copy snippets, test them on your own site, learn.
Understand principles – why does this choice matter? What user story is it serving? What questions is this answering for the user/customer?
Prompt AI agents with clarity – “make it look good” vs. “center content, grid images, sticky nav” will get you different results.
Stay curious – ask Claude why it chose certain design decisions (using your fundamentals that you’ve learned), dig into its reasoning.
Test + iterate – every change teaches something about responsiveness, compatibility, or UX.
The old website
My original site was mostly (80+%) hand-coded and served its purpose as a landing page when I only did pop-up events. It was a single page that didn’t feel very “Wave”, but it worked as something quick and dirty to be live.
Wave Floral 2.0
I wanted something that would reflect my creative energy with my favorite colors (hot pink and cobalt blue), feature large, attention-grabbing images of my work, include expanded service offerings with dedicated pages, work beautifully on mobile devices, and allow me to quickly test new landing pages and marketing ideas.
My goals for a new website:
More images and expanded service offerings
Move the ordering form to a separate page and optimize for mobile
Have the ability to test new landing pages quickly
Landing Page
I wanted a bold, simple entry point, so I built a full-screen photo layout inspired by florist websites I admire. It’s straightforward but effective: one striking image of flowers sets the tone and makes the brand feel alive the moment you land on the site.
Floral Services Page
This page was the hardest because it forced me to clarify pricing and what I actually wanted to offer. I settled on a bullet-point list of services paired with photos, since floristry is as much about educating clients on what’s possible as it is about delivering the work.
Along the way, I experimented with reusable icons, started a mini design system, and even discovered “neobrutalism” when researching button designs.
Neobrutalism is utilitarian, sharp, and bold – and something I was drawn to, but hadn’t created before while programming. I read up on the design style via blog posts, figured out the colors and rough shape that I wanted my neobrutalist headings and bullet-points to be, and prompted Claude, which did a pretty good job!
UofM Students Landing Page
This was my most experimental page and part of a hyperlocal marketing strategy in Ann Arbor. I designed funny, student-focused flyers with QR codes that linked to a dedicated landing page—within a week, 15 people had scanned and inquired.
This flyering campaign was actually the entire genesis behind why I ended up redesigning my website. A week before classes started in August, I had an idea to do these funny flyers and put them up around town. I realized I wanted a student-specific, loud, and cheeky landing page that people would see when they scanned the QR code.
This is when design and research, as well as some CSS fundamentals, really came into play. I created a few mockups of the rough layout in Canva and pasted them into my Claude Code session. It took several iterations and a lot more hand-written code than I expected, but we got there eventually with this mobile-first style. Along the way, I learned more about mobile-responsive design (something I often don’t think about when just using Tailwind or React libraries that already account for this at my day job) as well as how to even design something that looked kitschy and retro like this by inspecting Claude’s iterations and adjusting.
There’s still plenty to improve on my site—standardizing navigation, cleaning up CSS, and adding an auto-updating gallery—but this project reinforced something I believe deeply: AI’s real power is in augmenting curiosity and understanding, not replacing them. The people who thrive won’t be the ones chasing quick prompts, but those who use AI to deepen knowledge and build real expertise.
I’m super bullish on this because I’ve literally seen how I’ve been able to get an uptick in business, aided by AI. Compare this to hiring a web developer (easily $2,000-5,000 for a custom site) or premium website builders like Squarespace ($200+/year), and the economics for a small business owner or an independent operator become clear. But beyond cost, there's something powerful about building your own thing, with a tireless assistant.
Think about it like learning a language with an exceptional tutor. You don't just memorize phrases – you learn grammar, understand context, and develop intuition for what sounds right. Claude became my design and development tutor, helping me understand the underlying principles that make websites work.
This connects to what Anu Atluru writes about in "The Self Is The Platform" – we're entering an era where individual capability can rival collective coordination, but only when that capability is built on real understanding. The future belongs to people who use AI to become genuinely better at things, not just faster at producing mediocre outputs.
I hope this piece inspired or changed at least one person’s mind on who “gets” to work with AI and how we can continue to be a part of the future by working, interacting, and building with it.
If you enjoyed reading this and want more, you can find more of my work on the following:
Hotel wrangler, code pusher, demo queen, & co-founder: Whimsy
Flower slinging: Wave Floral
Algorithm worshipper: TikTok