Overengineering Detector

About the Project

💡 Inspiration

Hackathons are full of ambitious ideas — and overengineered ones.
We’ve all seen simple problems wrapped in Kubernetes, microservices, message queues, and buzzwords before a single user exists.

While preparing for Hack&Roll, we noticed something ironic:
AI is often added where it’s least needed — just because it’s available.

So we asked a simple question:

What if we built a project for “Most Overengineered use of AI”… that proves you don’t need AI?

That idea became Overengineering Detector.

🧠 What We Learned

• Overengineering follows clear, repeatable patterns.
• Many architectural decisions can be flagged deterministically, without AI.
• AI is most useful as an explanation layer, not a decision-maker.
• Clear constraints and fast feedback are more valuable than “future-proof” designs.
• Good tools should help teams delete complexity, not add more.

🛠️ How We Built It

The core of the project is a rule-based heuristic engine that analyzes project descriptions using 40+ weighted engineering signals (e.g. Kubernetes, blockchain, microservices, event sourcing).

Each signal contributes to an overengineering score capped at 100, and is mapped into higher-level architectural anti-patterns such as:

• Premature scaling
• Buzzword-driven development
• Distributed systems without need
• Operational complexity exceeding scope

This engine runs fully offline and returns results in ~95ms.

On top of that, we added an optional AI layer:

• When enabled, AI rewrites the same heuristic diagnosis into human-friendly (or sarcastic) language.
• When disabled, the system behaves identically — proving AI is not required for the core logic.

The frontend is a single-page, demo-first interface designed for fast judging:

• One input
• One button
• Immediate, visual results
• Optional roast mode for fun

⚠️ Challenges We Faced

• Avoiding the “LLM wrapper” trap while still using AI meaningfully
• Designing roast-style feedback that is funny but not offensive
• Keeping explanations short enough for 3-minute judging
• Ensuring the demo never breaks, even if the AI API fails
• Balancing humor with technical credibility

✅ Final Takeaway

Most teams don’t need better infrastructure.
They need clearer thinking.

Final verdict:
Stop building infrastructure. Build a product.

Built With

• next.js