APP IDEA #6 · AI · WRITING · EDTECH

Lunchbreak — making AI-written content undetectable, $51K MRR at peak

Lunchbreak takes AI-generated text and rewrites it to pass Turnitin, GPTZero, and 9 other major AI detectors, while keeping the original meaning and tone intact. Peak MRR of $51K, $1.2M all-time revenue, 2,993 subscribers, 70% profit margin — and currently listed for sale at $1.5M.

$51K

Peak MRR (TrustMRR verified)

$1.2M

All-time total revenue

70%

Profit margin

$1.5M

Listed for sale — founder moving to larger ventures

01 / HOW IT WORKS

What the app actually does

Lunchbreak's mechanism is fairly straightforward — AI detection tools look for patterns (sentence uniformity, specific word choices, lack of natural variation). Lunchbreak rewrites text to introduce those natural variations without losing the meaning.

User pastes their AI-generated text into the editor

The input is treated as a draft to be humanized, not published content — the tool's job is the transformation.

A live AI detection score shows the current "AI-ness" of the text

The user can see before any rewriting what score Turnitin or GPTZero would give — this makes the value immediately visible.

One-click rewrite restructures the text to remove detection patterns

Sentence length variation, vocabulary adjustment, and natural rhythm changes are all applied in one pass.

Final scan confirms the text passes all major detectors

User sees a green-light score before submitting or publishing — the loop closes with a verified result, not a hope.

02 / INDIA POTENTIAL

Does this work as an India play

India has 4Cr+ college students who regularly use AI for assignments. Turnitin adoption is growing fast across Indian universities. The demand is real — and the price sensitivity requires a much cheaper entry point than most global tools.

4Cr+

College students in India using AI for assignments — the addressable market is enormous.

Turnitin

Rapidly being adopted by IITs, IIMs, and tier-1 colleges, with tier-2 following — the detection pressure is increasing.

₹199/mo

Realistic India-priced subscription vs Lunchbreak's $14.99-$20/month (~₹1,300+) — opens the mass student segment.

Hindi

A version that humanizes Hindi-language AI text for Hindi-medium assignments and academic submissions is a gap no global tool addresses.

03 / THE WEEKEND BUILD

Friday to Sunday, hour by hour

The core is the AI rewriting call and the detection score display — the rest is UI polish. Both are achievable in a weekend.

Friday

Evening · 3 hrs

7–8 PM Set up a Next.js project, basic Supabase auth, and a simple textarea input for pasting text.

8–9 PM Build the AI detection score display — call a detection API or prompt-engineer a scoring heuristic with Gemini.

9–10 PM Build the one-click rewrite call — prompt Gemini with the text and instructions to humanize while preserving meaning.

Saturday

Full day · 7 hrs

Morning Build the side-by-side before/after editor so users can see original vs humanized text with differences highlighted.

Afternoon Add the post-rewrite detection scan — re-score the humanized output and show the improvement.

Evening Add word count display, preserve formatting (headers, bullets), and build the download/copy output action.

Sunday

5 hrs

Morning Add Hindi text humanization support — test specifically with Hindi AI-generated assignments.

Afternoon Razorpay integration — Rs199/month for unlimited humanization, free tier capped at 500 words/day.

Evening Test end-to-end with a real AI-generated essay, confirm detection scores improve, record the demo for your first post.

04 / APP STACK

What you're actually building with

Next.js 14

Frontend + API routes

Text editor, detection display, and output in one framework.

Supabase

Auth + database

Stores user accounts and usage tracking for the free tier cap.

Gemini API

Humanization engine

Rewrites text to introduce natural variation while preserving meaning.

Detection API

Score display

Third-party API or prompt-engineered scoring that mimics major detector behaviour.

Razorpay

Payments

Rs199/month subscription for unlimited word processing.

Tailwind CSS

Styling

Fast enough to build the editor and score display in a weekend.

05 / WHERE & HOW TO DEPLOY

Going live

Honest heads up: Detection bypass is an arms race — Turnitin and GPTZero update their models regularly. This category requires ongoing prompt tuning to stay effective. Factor this maintenance overhead in.

Push your project to GitHub, import into Vercel — auto-detects Next.js, no config needed.

Add environment variables: SUPABASE_URL, SUPABASE_KEY, GEMINI_API_KEY, RAZORPAY_KEY.

Deploy — Vercel gives you a live .vercel.app URL in under a minute.

Test your humanization prompts against real Turnitin/GPTZero outputs before launching.

Point a custom domain at it from Vercel domain settings.

06 / MARKETING & REVENUE

Getting paying users

How to market it

Post the "94% AI flagged to humanized to submitted to got an A" student story format as a reel.
Run 10-20 reels/day across multiple accounts targeting students, freelance writers, and researchers separately.
Target college student communities on WhatsApp, Telegram, and Instagram — this audience shares tools aggressively.
Offer a generous free tier (500 words/day) so students can test without committing.
Build a dedicated Hindi humanizer flow as a specific product page — essentially zero competition in this sub-niche.

Who pays, and why

College students submitting AI-assisted assignments at universities that use Turnitin or GPTZero.
Freelance content writers whose clients check for AI content before paying.
Academic researchers using AI to draft sections they then need to clean up before submission.

Scenario

Paying users/mo

Revenue/mo

Slow start

500 users x Rs199

Rs99,500

Student community traction + reels

5,000 users x Rs199

Rs9,95,000

10-20 reels/day + exam season spike

25,000 users x Rs199

Rs49,75,000

07 / START BUILDING

Paste this into Claude or GPT

This prompt sets up the full build context so the AI scopes, plans, and starts coding the project with you from message one.

BUILD_PROMPT.txt

I want to build an AI content humanizer for the Indian student market, inspired by Lunchbreak, scoped to ship a working version in a single weekend. Core flow: 1. User pastes AI-generated text into an editor. 2. A live AI detection score shows how detectable the current text is. 3. One-click humanize rewrites the text using Gemini API — introducing natural variation while preserving meaning. 4. Post-rewrite scan shows the new detection score. 5. User copies or downloads the humanized text. 6. Supports both English and Hindi text. Stack: Next.js 14, Supabase, Gemini API, detection scoring mechanism, Razorpay Rs199/month. Deploy: Vercel. Help me step by step: 1. Build the text editor input and basic Gemini humanization call first. 2. Build the detection score display — what approach do you recommend? 3. Build the before/after comparison view. 4. Add Hindi text support and test on Hindi AI-generated text. 5. Add the free tier cap and Razorpay paywall. Keep explanations short and India-context aware. Flag clearly that this is an arms-race category requiring ongoing prompt maintenance. Push me to ship the smallest working version first. If I get stuck, tell me to ask @buildwithkanhaa.