1. The Weekend That Got Hijacked by AI
It’s rare, but it finally happened — a weekend all to myself. My agenda was basically:
- Drink coffee
- Maybe play some games
- Hang out with my dog
But then I stumbled across a dangerous notification in my Cursor IDE:
GPT-5 is free inside
cursor-cli
.
OpenAI clearly doesn’t want people noticing Google’s release of Genie 3 — it’s become a trend that they try to steal the spotlight whenever Google releases anything.
But that will be a different article. For now, thanks for the free compute.
That one discovery transformed my weekend from low-effort bliss into “let’s see how much free usage I can squeeze in before they close the gates.”
Within an hour I had multiple terminals running and I was prototyping things like:
- React UI components
- A Pokémon battle simulator
- SVG-generated icons
Fun? Absolutely. Useful? Questionable.
But these are the usual stress-tests I run whenever a new model drops. That’s when I realized — if I’m going to keep building random things like this, I need a central hub. Somewhere to document my experiments, share my builds, and laugh at the occasional disaster.
2. The Plan That Changed Mid-Flight
Originally, I set out to make a generic business site template — clean, CMS-ready, client-friendly. Something I could repurpose for different people.
But here’s the thing about AI: It doesn’t just generate code really fast — it also allows you to explore many different solutions in such a short amount of time.
Which is very dangerous for someone like me, because I’m genuinely curious. Suddenly I had 10 more features than I needed, and my “simple” template was mutating into a chaotic mess.
Pretty soon, my “client-friendly template” had become a grab-bag of features I thought were cool in the moment but provided little to no real value. I was on track to adding it to the graveyard of apps I never finish.
At that point, I decided: no more playing around. I’m going to deploy this, strip it down, and get the fundamentals right before adding anything new.
3. Why I Picked Cloudflare (And Not Vercel or Netlify)
Sure, I could’ve gone with Vercel — Next.js and Vercel are basically best friends. But Cloudflare had been sitting in my “I should try this” list for months.
Here’s what sold me:
- Speed: Their edge network delivers content almost instantly, no matter where the visitor is.
- Security: Automatic SSL and DDoS protection without me lifting a finger.
- Global reach: Feels like my site is hosted in a cyberpunk server room in the sky.
Plus, developers I trust have been raving about it.
I was also excited to use wrangler
for managing a minimal local test and deploying straight to production from the terminal.
4. The First Build… And First Facepalm
My first attempt was peak feature creep. It had:
- A bunch of ShadCN components that looked cool but added no real value
- Inconsistent UI spacing
- A colorful theme that screamed “try hard”
The AI-generated “slop” was such a mess that there was no way I was going to use that repo.
I’d fallen right into the trap AI makes so easy: adding things because I can, not because I should.
5. Hitting Reset: The “Keep It Simple” Rule
Instead of wasting hours fixing problems I’d created, I decided to start fresh.
The new plan:
- Grab a minimal Next.js starter template from Vercel
- Strip it down even further
- Only include pages that serve a clear purpose
I ended up with:
- Landing Page → Who I am + what I do
- Blog → For builds, thoughts, and experiments
- Projects Page → Portfolio of AI-assisted creations
- Contact Page → Basic, no carrier pigeons required
If it didn’t help visitors, it didn’t make the cut.
Example: Starting the project
# Create a new Next.js project
npx create-next-app@latest my-blog
# Move into the folder
cd my-blog
# Install dependencies
npm install
6. Deploying to Cloudflare
With the site ready, deployment was next — and Cloudflare Pages made it weirdly painless.
I didn’t even need to push to GitHub. That still blows my mind.
Here’s what I did:
- Created a
wrangler.json
file. - Updated my
next.config.ts
. - Added a couple of scripts to
package.json
.
From there:
npm run preview
npm run deploy
This let me test the repo locally exactly as it would run on Cloudflare.
It was that easy — and now I don’t have to pay a monthly fee like I would with Vercel. Plus, I can manage deployments entirely from my terminal.
7. Why This Worked
Keeping things minimal saved my weekend.
The win list:
- Fast load times thanks to Cloudflare’s edge network.
- A clean Next.js setup that supports
generateStaticParams
and edge functions without the bloat. - A design I can expand without tearing it all down later.
The biggest win? I actually shipped something instead of leaving it in “coming soon” limbo.
8. Lessons Learned
- AI accelerates scope creep. Amazing tool, but it needs guardrails.
- Deploy early. You’ll find the real problems faster.
- Every feature has a cost. If it’s not worth maintaining, skip it.
9. What’s Next
This is just the starting point. I’ve got plans:
- Build Log → Public updates on experiments
- Training AI Models → Documenting a few product ideas and their progress
- Knowledge Graph → Training a personal AI using all the site’s content
- SaaS Ideas → Behind-the-scenes of starting a small venture
It’s going to get weird, experimental, and probably a little over-engineered — but that’s half the fun.
If you’re building your own portfolio or blog — especially with AI in the mix — remember:
Start small. Deploy fast. Improve later. A live imperfect site beats a perfect one nobody can visit.