The Story of SwellSlots

Jacques du Rand · April 4, 2026

SwellSlots — A Founder's Story

About Me

I'm a 40-something software dev who only "discovered" surfing last year. So I'm a beginner — and then some.

But surfing has already given me so much: joy, fun, exercise… even some real friends.

Why I Built This

Many of the surf forecast apps out there are fantastic, but navigation can be tricky, the information can be dense, and — well — they feel very clinical and cold. No vibe.

SwellSlots was created so I can quickly scan, with just my eyeballs, when it's a great time or a bad time to surf this week at my favourite local spots.

I also wanted to scratch that creative / aspiring-comic-artist itch and build something truly unique and visually appealing.

Growing up in the '90s, the graphics of that era feel nostalgic, vibrant, and distinctive compared to today's sometimes enterprise-y or ultra-realistic (Unreal 5) look and feel.

But most of all, I wanted to surf. And surf a lot.

Design

Oh, I'll be stoned for this, I'm sure — especially if I claim comic-artist aspirations — but AI image generation was really, really helpful here.

I feel like I rediscovered the joy of "building cool things" and "building things I actually want." Many, many times those dreams far outstripped my skills (especially as an artist). AI gave me a new brush and canvas to better express my vision.

Image AI (Gemini, ChatGPT)

Of course, like a kid in a candy store, I tried all the big AIs: ChatGPT, Gemini's Nano Banna, and some local ones.

The good: I got some really great results early on — exactly the look I wanted. I mentioned Street Fighter 2 and uploaded images of their logo and art style as examples. Both ChatGPT and Gemini (Nano Banna) did well, but ChatGPT was better at following prompts and making small corrections.

The bad: Both of them struggled to follow the requirement of placing image assets on a transparent background. It took many, many tries to get there, but by then there were other nasty side effects: crossed eyes, non-human expressions, and so on.

Bottom line: it's still better than what I can draw, but patience is key.

Surf Spot Database

The challenge was finding a way to get a list of real surf beaches — with their coordinates, aliases, and popularity — worldwide, while trying not to include general swimming beaches and focusing on surf beaches only.

Naïve First Try: Google Maps

I thought I'd just be able to query Google Maps and add the word "beach," "surf beach," etc. Boy, was I wrong. It found some, of course, but Google Maps just wasn't up to the task. It kept including general swimming beaches and missing some of the local surf hotspots.

AI to the Rescue: First Try

I then ran a basic query across most of the big three again: Anthropic (Claude), ChatGPT, and Google's Gemini. I had high hopes for Google, given their Maps and Places background. Again, it found some, but the accuracy was lazy-daisy at best — too many "just beaches" or swimming beaches were included.

Funnily enough, ChatGPT with GPT-5.3 and extended thinking was the "better" of the three. And I'm saying this as a Claude-is-my-fav person.

Second Try

I broke it up into three phases:

  • Countries
  • Regions
  • Surf Spots

We first generated a list of about 150 regions across 10+ countries, and then in a loop asked the AI APIs for the surf spots in those regions. Prompt iterations were tedious but necessary, and tuning seemed never-ending.

The main insight was this: this AI method will get you maybe 75–80% of the way there, but the rest is up to other methods and fine-tuning.

As a solo product owner (are we still using this term, or can I just say "builder"?), I'm okay with the output versus the effort.

The Surf Spot DB spun out into a separate internal project, Surf Scout, and is now an ongoing and fun side project.

AI Coding

Claude Code is simply amazing, not going to lie. Is it perfect? Does it make dumb decisions sometimes? Absolutely. Do I have the same flaws and more?

But I've learned to work with the AI-coding warts the same way my co-workers learned to work with me and my warts.

I really do mean that. I feel like a kid again, discovering "building cool projects." Coding is just the medium here, but now, with AI help, I get much closer — and I dare say faster — to my vision.

I know for some people the coding itself is the joy and the vision, and that's okay too.

Tech Stack

Nothing fancy here. We kept it SvelteKit and Tailwind, with silly bash scripts to deploy to Hetzner.

I still think Hetzner has the best cloud UX out there!

Wait, No App Store Apps?

Oh, joy…

Being on Linux and Android, I of course first targeted the Android app store — even paid for my "Developer Account Verification" at the cost of $25!

But to this day, it still seems unable to verify my address. I live in a complex (gated community), and most of my accounts have the internal roads listed (inaccessible to Google Maps, I guess). No matter what account or bill I use, the Google Developer Account process keeps saying rejected.

Not a great experience. Guess it's only Apple and web for now.

What's Next

Getting it into as many surfers' hands as possible, getting some critical feedback, and hopefully some compliments about the UX (haha — oh ego, be still!).

  • More surf spots!
  • Better UX and sharing options

AI Outlook and Opinion

My view on AI (coding, images, tasks) is like this: it's hard to say in a binary way whether AI will replace this or that, or me or you. I think it's much easier to talk in trend lines. So the better question is:

  • In 6 weeks, will there be more or fewer AI tools used in these disciplines and industries?
  • In 6 months, will there be more or fewer AI tools used in these disciplines and industries?

I think we can all agree the trend is up and more.

But… "AI is wrong," "bad coders use it," and all that — sure, that's true. And so are you, sometimes.

But your competitors — in life, love, coding, products — are definitely figuring out how to get value from these new tools, warts and all.

The genie is never going back in the bottle.

You might as well learn to work with them and their faults, because they're only going to get better and more popular.

Final AI-Assisted Life Thoughts

If the old adage was:

"Move fast and break things…"

…then the new AI adage is:

"Move fast and ship more things" (while they might break too).

Surf more!


← All Articles