What I Learned Building My First AI-Assisted Side Project

I was scrolling through tech Twitter when I saw another viral photo of a billboard from Highway 101 in San Francisco. Someone in the comments asked "what company is this?" and nobody knew.

This happens constantly. And as a marketer, I get a little FOMO from only being able to interpret these billboards from behind my laptop vs in my car on a daily commute

So had an idea for an online archive of all these billboards. Another "idea" but how do I actually bring it to life? Enter techbillboards.ai

The Weekend Build

Friday Night

I opened my laptop around 8 PM with a simple goal: build a site where people can browse tech billboards and actually understand what the companies do.

My stack choice:

• Lovable for rapid prototyping and deployment. You can go from idea to deployed app faster than traditional development workflows.

• Supabase for the database and auth. PostgreSQL with built-in auth and real-time features. I knew I'd want user submissions eventually.

• Logo.dev for company logos. Consistent company branding without hunting down logo files. One API call gets you clean SVGs.

I started with a conversation with Claude: "Help me build a billboard catalog site with location data and company information." About 30 minutes later, I had a working prototype and detailed product requirements.

Saturday

Most of Saturday was spent on the core functionality: company profile pages with descriptions and links; location mapping so people can see where billboards actually are; search and filtering by company, location, or campaign type.

At this point, I switched to Lovable chat. I'd describe what I wanted, get working code, test it, then immediately think of the next improvement.

By Saturday evening, I was uploading billboard photos and company data. Started with the obvious ones: PostHog, Intercom, Artisan…

Sunday

The hardest part wasn't technical - it was content. I needed 50+ billboard campaigns to make the site feel complete.

I reverse-searched examples I had previously seen on my feeds. Searched for "101 billboard," "sf billboard," "ai billboard," etc. Scraped results I could find on twitter/x and linkedin.

This took longer than the actual development. Turns out building the database is easy. Filling it with good data is the real work.

Deployed around 6 PM. Total build time: maybe 10 hours across three days.

The Response

Posted about it on LinkedIn wedensday morning. Within hours:

• The NYTimes had just released a quiz on tech billboards. Got to offer a gift link.

• A few marketing team asked for better photos (more staged versions) than the user snapped photos pulled from car rides

• A few founders asking if they could get included.

Traffic hit 1K visitors in the first week. Not bad for a weekend side project.

The Bigger Shift

AI eliminated all the setup friction. Instead of spending hours configuring Supabase schemas or fighting with CSS layouts (which normally would have stopped me dead in my tracks), I focused on product decisions: data architecture decisions, content strategy, and design choices. It handled Initial database design, boilerplate React components, API endpoint setup, responsive CSS frameworks, and basic error handling.

TechBillboards.ai isn't revolutionary. It's a simple catalog site that solves a specific problem.

But the process of building it changed how I approach every work problem.

Before: "This should exist. Maybe I'll buy a domain and get to it one day."
Now: "This should exist. Let me build it this weekend."