Last January I had 847 supporters on Tpylo. Made around $4,800/month. Felt great until I calculated the fees.

William Olsen — Sun, 14 Dec 2025 10:01:18 +0000

Here's what I was actually paying:

Patreon: 8-12% platform fee

Stripe: ~3% payment processing

Total: Losing $500-600 every single month

That's a used car every year. Gone.

So I spent 3 days researching every creator platform:

Ko-fi (still 5% on memberships)

Buy Me a Coffee (5% too)

Gumroad (10% 💀)

Then found something different. A platform taking only 5% total. Not 5% + processing. Just 5%.

I switched everything in February.

Same supporters. Same content. Same memberships.

8 months later:

Saved $3,200 in fees

Added digital product sales (wasn't easy on Patreon)

Direct chat with top supporters (game changer for retention)

My take-home: 95% instead of 85%

The migration took one weekend. Sent one email. Lost 3 supporters, gained 47 new ones in the first month because the checkout was cleaner.

What actually mattered:

Transparent pricing (supporters see exactly where money goes)

Multiple revenue streams in one place (donations + memberships + digital products)

Direct connection features (not buried behind tiers)

I'm not saying this works for everyone. But if you're paying 10%+ in fees and wondering where your money goes, do the math.

Someone asked me in DMs how to calculate if switching is worth it:

Monthly revenue × 12 × (current fee% - new fee%)

Example: $5,000 × 12 × (13% - 5%) = $4,800/year saved
If that number is bigger than what you'd spend on a good laptop, maybe worth exploring.

Happy to answer questions.

I Built a Full SaaS Platform in 3 Months Without Knowing How to Code (Here's My Story)

William Olsen — Fri, 28 Nov 2025 10:57:08 +0000

Let me start with something uncomfortable: I don't know how to code from scratch.

I'm not one of those people who learned programming at 12, built websites for fun, or has a computer science degree.

I'm just someone who's been frustrated by the same problem over and over: every time I want to try a new tool or start a project, I waste hours scrolling through endless best tools lists, reading random blog posts, and still ending up confused.

Three months ago, I had an idea: what if there was a platform that actually helps you decide which digital tools to use, based on real reviews and clear comparisons?

*Not just another messy directory, but something that feels alive and useful.
*

I called it *Jozaz *— a decision engine for digital tools.
But here's the problem: I had no idea how to build it.

The moment everything changed discovering AI coding tools

I knew what I wanted Jozaz to do. I even sketched the pages on paper: a homepage with categories, product pages with reviews, a ranking system that shows trending tools, user profiles, subscriptions… all of it.

But when I opened a blank code editor, I froze.

Then I discovered Cursor and Windsurf AI coding tools that help non-programmers build software. I typed: "Create a Next.js app with a homepage showing products." It worked. Not perfectly, but enough to start.

Failures and surprises

After a few days, I discovered the disasters.

My entire approach was fundamentally wrong. When I said "build the homepage" I wasn't following best practices for a project of this scale.

That's when the hardest part of the journey began: understanding how to actually build a complex platform.

I started reading and learning, even if just at a basic level. Eventually, I realized something crucial: *this isn't simple at all, even with AI tools. *

The big problem is that AI will indeed build what you ask for. But I discovered it's not just about building it's about how you structure and organize code in a hierarchical, balanced way so everything works together.

Month 1: Pure chaos. I didn't understand databases or authentication. I asked Cursor endless questions like How do I let users sign in Sometimes it worked instantly. Sometimes I debugged for hours.

By month's end: working sign-in, database, basic product pages.

Month 2: I built the core features — ranking algorithms, dynamic badges, category filters, real-time notifications, and subscriptions via Polar.sh.

I explained what I wanted in plain language; the AI generated code. I tested, broke it, fixed it, repeated.

Month 3: Polish turned into panic. Bugs, performance issues, design problems everywhere. The real issue? Me thinking "it's not ready yet."

And believe me, the process of discovering errors, protection, and fixes was 100 times more difficult than building the project.

For example, when I run npm run lint. This was the real shock: over 544 problems, ranging from warnings to major errors. But now i fixed them almost all.

Snyk.io, It greatly helped me in discovering security problems.
So I fixed critical bugs, optimized performance, and shipped.

1.5 weeks ago: launch day

I deployed Jozaz and posted on Reddit, indie communities, and Discord. Not check out my startup but honest: I'm not a developer.

What I learned

Artificial intelligence will not replace the developer
Shipping beats perfecting real feedback is invaluable
Honest storytelling attracts real support

Three months ago, I couldn't code. Today, Jozaz is live with real users. If I can do it, you can too.

https://jozaz.com/

DEV Community: William Olsen

Last January I had 847 supporters on Tpylo. Made around $4,800/month. Felt great until I calculated the fees.

I Built a Full SaaS Platform in 3 Months Without Knowing How to Code (Here's My Story)

The moment everything changed discovering AI coding tools

Failures and surprises

1.5 weeks ago: launch day

What I learned