I Turned Your GitHub Profile Into a Detective Case File (And the Verdict Might Hurt)

Mariam Tariq — Sat, 01 Aug 2026 12:36:07 +0000

Your GitHub profile is a public confession booth.

Repos count. Followers. That bio you wrote at 1 a.m. and never fixed. Languages you claim to "love" while secretly opening node_modules like a crime scene every Tuesday.

I kept staring at those numbers and thinking: this shouldn't look like a résumé. It should look like a case file.

So I built Commit Crime Scene — a free, silly little tool that takes a GitHub username and stamps out a detective dossier: charges, a chaos score, a verdict class, and a downloadable PNG you can drop into Slack like evidence.

No login. No OAuth. No "connect your account so we can harvest your soul." Just a username and the public GitHub API.

The setup
You type a handle. The precinct goes to work.

Behind the yellow tape, the page:

Fetches the public user profile
Pulls recent public repos (up to 100)
Skims languages, forks, archive dust, follower ratios, account age
Invents a completely unserious legal case against you
Thirty seconds later you've got a paper-looking dossier: mugshot (your avatar), case ID, "charges," chaos score, languages of interest, and a rubber-stamp verdict that either says CASE CLOSED or — worse — STILL AT LARGE.

It's theater. Delightful theater. But it's theater powered by your actual public stats, so the jokes land a little too close to home.

Sample charges (yes, these are real outputs)
Depending on the profile, you might get hit with things like:

Operating with TypeScript, JavaScript, Python in broad daylight
Flooding the evidence locker with 87 public repos
Recruiting 240 witnesses (followers) — possible cult
No bio. Classic. Denies everything.
Suspected collaboration with non-human accomplices (looking at you, AI repo names)
If your bio contains TODO, WIP, hack, or broken… the tool notices. The statute of limitations does not.

Verdict classes (prepare your ego)
The classification system is not peer-reviewed. It is, however, emotionally accurate:

Serial Shipper — you open repos the way other people open tabs
PR Ghost — many eyes, few commits in the public square
Weekend Bandit — high chaos, midweek alibis thin
Dependency Hoarder — forks everywhere; originals unclear
AI Co-Conspirator — your bio/repos scream LLM energy
Cold Case Committer — years on the beat, suspiciously calm chaos score
Package Deal Felon — JS/TS/Python + elevated chaos = indictment
Suspect at Large — light paper trail; detective raises an eyebrow
There are a few more in the pool — Commit Scrawler, Branch Burglar, Merge Conflict Artist — because sometimes the algorithm just wants you roasted gently.

Chaos score: the fake science section
There's a 7–99 "chaos" number. It's not ML. It's not astrology. It's a cheeky blend of:

repos-per-year velocity
language variety
forks + archived repos
follower weight
whether you bothered to write a bio (no bio = slightly more suspicious, obviously)
Is it rigorous? Absolutely not.
Is it fun to compare with your teammates? Dangerously yes.

How it actually works (for the curious)
Stack-wise it's intentionally boring — in a good way:

Plain HTML / Tailwind / a little vanilla JS
GitHub REST API from the browser (/users/{login} + /users/{login}/repos)
Deterministic-ish hashing for case IDs and a couple of fallback verdicts
html2canvas to export the dossier as a PNG
Everything runs client-side. Your token never leaves your machine because… there is no token. Unauthenticated public API only.

Rate limits (the plot twist)
GitHub's unauthenticated rate limit is the one real villain in this story. If the precinct gets rate-limited (403/429), the tool doesn't die quietly — it offers a demo case so you can still screenshot, share, and keep the bit alive.

There's also a shareable URL shape (?u=yourhandle) so you can send someone a deep link and make them face their own verdict in public.

Why build a toy like this?
Because shipping doesn't have to be solemn.

Developers already live inside dashboards, CI red builds, and "just one more dependency." A weekend project that turns contribution graphs into noir stationery is a palate cleanser — and a surprisingly good conversation starter in Discord / team chats.

Also: if a tool makes someone laugh and click "download PNG," you've accidentally designed a viral loop. Yellow crime-scene tape helps.

Try it on yourself (and then on your manager)
Open Commit Crime Scene
Enter a GitHub username
Read the charges out loud for maximum courtroom energy
Download the PNG
Drop it in a channel with zero context
If you get Serial Shipper, wear it.
If you get PR Ghost, we both know what you did.
If you get AI Co-Conspirator, welcome to 2026.

👉 Investigate a profile: shipnative-landing.vercel.app/crime-scene

Built for fun under the ShipNative umbrella (an Expo / React Native AI starter kit + free setup checklist for people who actually want to ship apps). The crime scene tool is free, standalone, and not affiliated with GitHub — just a silly dossier generator for public profiles.

Now go file a case. The evidence locker is open.

How I stopped rebuilding the same Expo AI app foundation every time

Mariam Tariq — Sat, 01 Aug 2026 06:36:01 +0000

Every time I started a new Expo AI app, I ran into the same problem.

I wanted to build the product… but I spent the first week rebuilding the foundation instead.

Auth. Navigation. Chat UI. Theme. Paywall. Styling config.
Again and again.

The real bottleneck wasn’t the idea
The idea was usually clear.
The slow part was everything around it:

Setting up Expo Router the right way
Wiring Supabase auth with persistent sessions
Building a chat UI that supports streaming responses
Adding light/dark mode that actually saves
Creating a paywall that can later connect to real in-app purchases
That setup can easily take 40+ hours — before you ship a single real feature.

What I decided to do
I stopped treating boilerplate as disposable.

I packaged the stack I kept rebuilding into a reusable starter called ShipNative.

It’s a TypeScript Expo kit with:

Expo Router (file-based routing + auth redirects)
NativeWind v4 for styling
Supabase authentication with Secure Store
Gemini-powered AI chat with streaming UI
RevenueCat-ready paywall (monthly/annual + restore)
Demo mode so you can explore without API keys
Persisted light/dark theme
A .cursorrules file for cleaner AI-assisted development
Why demo mode matters
I didn’t want a starter that only works after you configure five services.

With demo mode, you can open the app and immediately test:

login/register screens
AI chat streaming behavior
paywall upgrade/restore flow
Then add your Supabase, Gemini, and RevenueCat keys when you’re ready for production.

Who it’s for
ShipNative is for builders who want to move faster:

indie hackers launching AI mobile apps
freelancers who don’t want to bill clients for setup week
developers who want a clean architecture they can customize
It’s not a no-code builder, and it’s not a finished consumer app.
It’s the foundation — so you can focus on your product.

Quick start
npm install
cp .env.example .env
npx expo start
If you want the full kit
I published it here:

Product: https://mariamtariq72.gumroad.com/l/shipnative
Landing page: https://shipnative-landing.vercel.app
Quick question
What slows you down most when starting a new Expo app?

Auth, navigation, AI chat UI, paywall/IAP, or styling setup?

I’d love to hear what you’re stuck on — I’m improving the starter based on real feedback.

DEV Community: Mariam Tariq

I Turned Your GitHub Profile Into a Detective Case File (And the Verdict Might Hurt)

How I stopped rebuilding the same Expo AI app foundation every time