How We Won VIBE-ATHON 2026 — Building Read X in 6 Hours
Roni sarkarSome days just hit different. May 16th, 2026 was one of those days.
Our team walked into the VIBE-ATHON 2026 hackathon at Techno India University with a boilerplate, a plan, and a lot of nervous energy. Six hours later, we walked out with first place.
This is the story of how we built Read X — a regional Indian script OCR, translation, and preservation tool — in a single day.
What Is VIBE-ATHON?
VIBE-ATHON 2026 was not your typical coding contest. Organized by Technothon at Techno India University, West Bengal, it was a 6-hour vibe coding hackathon where creativity mattered as much as technical execution. AI tools were explicitly allowed. There were 15 curated problem statements and each team received theirs through a lottery system. No pre-built templates. No shortcuts. Just raw building.
The format was simple and brutal at the same time — 6 hours, 3 people, one problem statement, and a jury watching at the end.
Our Team
We were a team of three. I handled the full-stack development — backend API routes, database, and integration. My second teammate focused entirely on the frontend, building every page and component. Our third teammate, who does not come from a coding background, owned the design, the presentation, and quality assurance throughout the build.
That last point matters more than people think. A lot of teams underestimate the non-technical member. We did not. She was testing the app continuously, writing the pitch script, and building the slide deck while we were deep in the code. When it came time to present to the jury, she delivered a 3-minute pitch that was clear, confident, and compelling. That presentation was a significant part of why we won.
The Problem Statement — Problem 11
When the lottery assigned us Problem 11, we knew immediately that this was ours to lose.
The problem described a very real and very human challenge: people often encounter old handwritten notes, signboards, or printed material in unfamiliar regional scripts and cannot understand or preserve the information easily. The expected outcome was a system that could extract text from handwritten or printed regional scripts, translate it, and make it easy to save and share.
The requirements included:
- Detect text from images automatically
- Support multiple Indian writing systems
- Convert extracted content into readable text
- Allow translation into familiar languages
- Handle blurry or uneven photographs
- Enable quick sharing and saving of results
The moment we read it, we looked at each other and said — Gemini Vision API. That was the entire technical insight that unlocked our solution.
What We Built — Read X
Read X is a web application where a user uploads a photo of any handwritten or printed text in a regional Indian script — Bengali, Hindi, Tamil, Telugu, Kannada, Malayalam, and more — and the app automatically extracts the text, detects the language, translates it to any target language, and lets the user save and share the result.
The demo moment that impressed the jury most was uploading a photo of real handwritten Bengali text and watching it transform into extracted text and an English translation in seconds. That single interaction told the entire story of the product better than any slide ever could.
The Tech Stack
We built Read X on a stack I was already deeply familiar with, which was a deliberate choice. Under time pressure, familiarity is a competitive advantage.
Frontend: Next.js 15 with App Router, Tailwind CSS, shadcn/ui
Backend: Next.js API Routes
Database: Supabase (PostgreSQL) — we actually switched from MongoDB mid-event due to connection issues, migrated to Supabase, and it was one of the best decisions we made all day
Authentication: NextAuth.js with email/password, OTP validation, Google, and GitHub OAuth
Image Storage: ImageKit
AI Core: Google Gemini 1.5 Flash API for OCR, language detection, and translation in a single API call
Deployment: Vercel
The architecture was straightforward by design. A user uploads an image, it goes to ImageKit for storage, then the base64 version goes to Gemini which returns the extracted text, detected language, and translation in one structured JSON response. The result is displayed instantly, and the user can save it to Supabase and generate a public shareable link.
The Preparation That Made It Possible
Winning a 6-hour hackathon is not about what you build in 6 hours. It is about what you prepare in the days before.
Before May 16th, I had already built a complete boilerplate with authentication, database connection, reusable components, API route templates, protected routes, and a deployment-ready Vercel configuration. When we received our problem statement, we did not spend a single minute on setup. We cloned the boilerplate and started solving the actual problem immediately.
That preparation gave us roughly 90 minutes of advantage over teams that were still configuring authentication and database connections during hour one.
The lesson is simple — the boilerplate is not a shortcut. It is evidence of preparation.
The Hardest Moment
Around hour three we hit a wall with MongoDB connection issues that were costing us time we could not afford to lose. The decision to cut MongoDB entirely and migrate to Supabase mid-build was stressful. It felt like a risk. But Supabase's clean API and zero configuration overhead made the switch faster than staying with a broken setup would have been.
That decision — made under pressure with 3 hours remaining — is one I would make again without hesitation.
The ability to make a clean technical call under stress, without overthinking it, is a skill you only develop by building things.
What the Jury Said
The jury evaluated on four criteria — innovation and creativity, technical implementation, problem-solving approach, and presentation clarity.
Our edge on innovation was the Gemini Vision integration handling OCR, language detection, and translation in a single API call rather than three separate systems. Our edge on presentation was a non-technical teammate who had been testing the app for four hours and could explain exactly what it did and why it mattered to real people.
The combination of technical depth and presentation clarity is what ultimately separated us from the other teams.
What I Learned
Preparation beats talent under time pressure. The team with the best boilerplate almost always outlasts the team with the best ideas.
The non-technical teammate is not a liability. In a hackathon format where presentation is 30-40% of the score, the person who owns communication is as valuable as the person who writes the most code.
Scope ruthlessly. We picked the core features, committed to them, and built them well. We did not try to implement everything on the problem statement. A clean, working demo of three features beats a broken demo of ten every time.
Switching tools under pressure is okay. The MongoDB to Supabase migration mid-event felt scary. It was the right call. Do not stay loyal to broken tools when a better option is one install away.
Your demo has one job. It needs to make the jury feel the problem and feel the solution in under three minutes. Our opening line to the jury was — "Someone finds an old handwritten letter from their grandparent in Bengali and cannot read it. In ten seconds, Read X extracts and translates it." That one sentence did more work than any technical explanation.
What Is Next for Read X
Read X is not going into a folder somewhere. The problem it solves is real — India has hundreds of regional scripts and millions of documents that are inaccessible to people who cannot read them. A tool that bridges that gap has genuine value beyond a hackathon demo.
I am planning to polish the codebase, improve the image handling for lower quality photographs, add support for PDF uploads, and deploy it properly as an open-source project. If you are interested in contributing or using it, watch this space.
Final Thoughts
VIBE-ATHON 2026 was the most intense and rewarding six hours I have spent building something. The pressure was real, the stakes felt real, and the win feels real.
To my team — thank you. Every role mattered. Every decision mattered. We earned this one together.
To anyone preparing for a hackathon — start your boilerplate today. Know your stack cold. Divide roles clearly. And when the problem statement drops, scope fast and build clean.
See you at the next one.
Read X was built at VIBE-ATHON 2026, Technothon, Techno India University, West Bengal.
Stack: Next.js · Supabase · Gemini 1.5 Flash · ImageKit · NextAuth.js · Tailwind CSS · shadcn/ui · Vercel
Read X was built at VIBE-ATHON 2026, Technothon, Techno India University, West Bengal.