Three days ago I had an idea. Today it's live, indexed on Google,
and has real visitors from Bing and Twitter.
I'm a last-year Computer Science student. I know JavaScript
(React, Node), a bit of Java and C. I had never shipped a real
product before.
Here's exactly what I built, how I built it, and what I learned.
The Problem I Wanted to Solve
A friend of mine is a life coach. After every client session she
spends 45 minutes writing up summaries, action plans, and
follow-up emails. Every. Single. Session.
I looked around for tools that solve this. Everything I found
either required a bot to join your call and record it, or was
a massive platform with 50 features she didn't need.
She does in-person sessions. She takes her own notes. She just
needed something to turn those notes into professional documents
fast.
So I built it.
What I Built
CoachWriter AI — a web app where coaches paste their raw
session notes and instantly get:
- Session summaries
- Client action plans
- Follow-up emails
- Progress reports
No recording. No bot. No audio processing. Just notes in,
professional documents out.
The Stack
- Frontend: React + Vite + TailwindCSS
- Backend: Node.js + Express
- AI: Google Gemini API (free tier)
- Auth + DB: Supabase
- Payments: Stripe
- Deployment: Vercel (frontend) + Railway (backend)
Total infrastructure cost to start: €0.
(Railway has a 30-day free trial, everything else has a free tier)
How Claude Code Changed Everything
I used Claude Code — Anthropic's CLI coding tool — as my
primary development partner throughout the build.
The workflow was simple:
- Plan and strategize in Claude.ai chat
- Build in VS Code with Claude Code running in the terminal
- Repeat
Claude Code would read my entire codebase, understand the
context, and make changes across multiple files at once.
Things that would have taken me hours took minutes.
Some examples of what I delegated entirely to Claude Code:
- Full Supabase auth integration (signup, login, protected routes)
- All 4 Gemini API prompt templates
- Stripe subscription flow
- Usage limiting system (free tier: 5 generations/month)
- Complete test suite
- Full documentation
I focused on decisions, product direction, and catching mistakes.
Claude Code focused on implementation.
What Surprised Me
The speed. I went from idea to deployed product in 3 days
while also doing all the non-code work — domain purchase,
SEO setup, Google Search Console, social media posts.
The SEO setup. I'd never done SEO before. Within 24 hours
of launch my site was indexed on Google and Bing. I got my
first organic Bing visitor the same day I submitted my sitemap.
How much there is beyond coding. Building the product was
maybe 40% of the work. The rest was: choosing a domain,
setting up DNS, configuring CORS for the custom domain,
fixing duplicate meta tags, submitting to directories,
writing blog posts, posting on Reddit and Twitter.
Nobody tells you about all that stuff.
The Mistakes I Made
CORS. When I moved from the Vercel subdomain to my custom
domain trycoachwriter.com, I forgot to update the allowed
origins in my Express backend. Every API call failed. Took
me longer than I'd like to admit to figure out why.
Duplicate meta tags. I added SEO tags in both index.html
and via react-helmet-async. Bing flagged me for duplicate
title tags, description tags, and canonical tags. The fix
was simple once I understood the problem — static tags in
index.html for crawlers, Helmet only for dynamic page titles.
Sitemap serving. I spent way too long trying to serve
sitemap.xml as a static file from Vercel with a React SPA.
The catch-all rewrite kept intercepting it. Eventually solved
it but it cost a few hours.
Current Status
- ✅ Live at trycoachwriter.com
- ✅ Indexed on Google and Bing
- ✅ First organic visitors from search
- ✅ Custom domain configured
- ✅ Blog live with first post
- ⏳ First paying customer (working on it)
- ⏳ €350/month MRR target (my rent)
What's Next
Getting the first paying customer is the only metric that
matters right now. I'm reaching out directly to coaches on
LinkedIn and Instagram, posting in coaching Facebook groups,
and writing SEO blog posts targeting keywords coaches
actually search for.
The product works. The infrastructure is solid. Now it's
purely a distribution problem.
Would I Do It Again?
Absolutely. The combination of Claude Code for building and
Claude.ai for strategy is genuinely powerful for a solo
developer. I shipped something real, with auth, payments,
AI integration, tests, and documentation, in 3 days.
If you're a CS student or junior developer sitting on an
idea — just start. The tools available right now make it
more achievable than ever.
If you're a coach or consultant dealing with post-session
admin hell, try it free at
trycoachwriter.com —
no credit card required.
And if you're a developer curious about the Claude Code
workflow, drop a comment — happy to go deeper on any part
of the build.
Top comments (0)