I built a SaaS to compete with tools charging 49-99€/month. Here's what I'd do differently

I'm a solo dev from France. For years I did contract work during the day and built side projects at night. One of those projects needed a changelog, and that's when I fell down the rabbit hole.

I tried Beamer (49€/month). Too expensive for what it does, and the price goes up with your traffic. I tried AnnounceKit (79€/month). No custom domain unless you pay 129€/month. I looked at Headway. Last update: 2020. I checked Canny. Great feedback board, but 79€/month and the changelog is an afterthought.

I ended up analyzing 15 tools. Made a spreadsheet. Compared every feature, every price, every limitation. And I found something obvious that nobody was building for: the indie dev who wants changelog + feedback + roadmap + status page in one tool, for less than 15€/month, hosted in the EU.

So I built it myself. I called it Vershun. 55 sprints. One developer.

I used Claude to write detailed specs and plan each sprint, then Claude Code to implement features, never the core architecture. Just well structured prompts with full context.

But here's the thing: AI doesn't replace knowing what you're doing. My dev background is what made this work. I knew which patterns to use, when to push back on what Claude Code generated, when to refactor, when to throw away and start over. I reviewed every line. I caught security issues, fixed bad architectural decisions, enforced consistent design patterns across 55 sprints. Claude Code is fast, but without someone who knows what good code looks like, you end up with a fragile mess. The AI builds. The developer decides.
Features that took me days before now take hours. Refactoring is almost free. The bottleneck moved from "can I build this" to "should I build this". Turns out that second question is harder.

The stack is boring on purpose: FastAPI, Vue.js, PostgreSQL, Astro for the landing page, vanilla JS for the widget. Hosted on OVHcloud in France. Caddy for auto HTTPS on custom domains. Cloudflare CDN. Total infra cost: 8€/month.
What the product does today: changelog with labels, scheduling, AI generation. Feedback board with anonymous voting. Public roadmap. Releases with AI summaries from GitHub and Linear. Uptime monitoring (HTTP, TCP, Ping, SMTP) with auto incidents and public status page. In-app widget under 18KB. Email notifications to subscribers. All of it in one dashboard, one price.
Pricing: Free forever. 7€/month Starter (6€/month yearly). 15€/month Pro (12€/month yearly). 39€/month Business (32€/month yearly). Flat. No MAU. No tracked users. No per seat.

What I'd do differently:

Start marketing on day one. I spent months heads down coding and zero time building an audience. Now I have a complete product and I'm starting from zero followers, zero email list, zero community. Classic dev mistake.
Write the comparison pages earlier. My /compare pages (vs Beamer, vs Canny, vs AnnounceKit, etc.) are now some of the most visited pages on my site. People actively search for "[tool] alternative". I should have published those way earlier and let them index.

Don't build the status page before launch. I built an entire multi-protocol uptime monitoring system with real-time SSE and 5 notification channels. Nobody asked for it yet. I could have launched months earlier without it and added it based on demand.

Talk to potential users before building. I analyzed competitors obsessively but never talked to a single person who would actually buy. The market research validated the opportunity. Talking to people would have validated the messaging.

Where I am now:

Launching on Product Hunt April 15. Zero customers. Zero MRR. Everything is ready. I know the hardest part starts now: finding the first real users.
You can try it instantly, no signup needed: https://vershun.io

First 50 paying customers get 20% off for life with code EARLY20.
What's your current changelog setup? Paying for one of the expensive tools, using a blog, or just not doing it at all?