I'm not a professional developer. But I had an idea on Friday night and
by Sunday I had a live product at troubleshooting.sh. Here's exactly
how it happened.
The problem I kept running into
Every time I hit a cryptic error I'd copy-paste it into ChatGPT and get
back something like "make sure your variable is defined." Technically
correct. Completely useless.
I wanted something that treated me like a developer — told me exactly
why the error happened, gave me the precise fix, and explained the
reasoning so I didn't hit the same wall again.
The idea
A single-purpose tool. Paste your error, pick your stack, get back:
- Root cause — the actual reason it happened
- Exact fix — the command or code change, not a vague pointer
- Why it works — the underlying explanation
- Prevention — one tip so you don't hit it again
That's it. No dashboard, no account, no bloat.
The unexpected constraint: I only had a Chromebook
No MacBook. No Windows machine. Just a Chromebook.
Turns out Chromebooks ship with a built-in Linux container called
Penguin. You enable it in settings, open a terminal, and suddenly
you have a full Debian environment. I had no idea this existed until
I needed it.
From there the setup was straightforward:
mkdir troubleshooting-sh && cd troubleshooting-sh
npm init -y
npm install express anthropic dotenv
The stack
I kept it as simple as possible:
- Node.js + Express — backend server
- Claude API — the AI that actually fixes the errors
- Vanilla HTML/CSS/JS — no framework, one file frontend
- Hostinger — hosting via Git deploy
- GitHub — version control and deployment pipeline
No database. No auth. No unnecessary complexity for v1.
The part that took longest: getting the AI response right
The actual code was maybe 3 hours. The system prompt took longer than
anything else.
A raw API call to Claude gives you a decent answer. But tuning the
system prompt to consistently return structured, useful responses —
root cause first, exact fix second, no fluff — that took a lot of
testing with real errors from Stack Overflow.
This is what finally worked:
You are an expert debugger. When given an error message:
ROOT CAUSE: One sentence explaining exactly why this happened.
THE FIX: The exact code change or command to resolve it.
WHY IT WORKS: One sentence explaining the fix.
PREVENT IT: One tip to avoid this error in future.
Be concise, precise, and practical. No fluff.
Simple. But getting to simple took a while.
Streaming makes everything feel faster
One thing I'm proud of — the response streams token by token so you
see the answer building in real time. No spinner. No waiting for a
complete response before anything shows up.
This is a small UX detail that makes a big difference in how the tool
feels to use.
Deploying from a Chromebook
This is where it got interesting. I couldn't use a traditional FTP
upload or SSH easily from ChromeOS. Instead:
- Pushed code to a private GitHub repo from the Linux terminal
- Connected that repo to Hostinger via their Git import feature
- Hostinger auto-detected Express, ran npm install, and deployed
The whole deployment pipeline took about 20 minutes to set up and now
every git push auto-deploys. From a Chromebook.
What I skipped intentionally
A lot of "proper" SaaS features didn't make the cut for v1:
- No database (added later for shareable fix links)
- No authentication
- No payment system yet
- No analytics beyond basic page views
I added a simple in-memory rate limiter — 5 fixes per day per IP —
to protect against API cost blowouts on launch day. That took 10
minutes and was worth doing before going live.
Everything else can wait until real users tell me what they actually
want.
The result
troubleshooting.sh is live. It works. Real errors go in, useful fixes
come out.
Built entirely on a Chromebook, in one weekend, with no prior
deployment experience on this stack.
If you're sitting on an idea and waiting for the perfect setup —
the right machine, the right time, the right amount of knowledge —
this is your sign that you don't need any of that.
You just need a terminal and a weekend.
Try it at https://troubleshooting.sh — paste your nastiest error
and see what comes back. I'd love to know what breaks it.
What's next: shareable fix links, a Pro plan for unlimited fixes,
and better handling of Rust and Go errors. Building in public —
follow along if you're interested.
Top comments (0)