BUILD or KILL

• 

  Form for SaaS idea

• 

  Output for SaaS idea

Inspiration

There are a lot of AI idea validators and site graders online, but most of them give generic and overly positive feedback. They don't actually help founders make a decision.

What it does

Build or Kill is a structured AI decision engine for SaaS ideas.

Users input a problem, solution, target audience and optional context. The system evaluates the idea using weighted scoring model across factors like market demand, competition, monetization, distribution and founder fit.

It returns a clear verdict: BUILD, PIVOT, or KILL, along with: numerical score, brutal truth analysis, reality check from a business perspective, a contrarian insight, basic execution plan for viable idea.

How we built it

The project was build using: Next.js for the frontend and API routes, tailwind CSS for UI, OpenRouter API for model access, localstorage for persistence. I did use opencode to write code and most of my effort just went in orchestrating and testing and designing everything.

Challenges we ran into

The biggest challenges were: Prompt engineering: Getting consistent and structured outputs was difficult. Early version, I was trying to combine models which was unnecessary complicated and different models rate the idea differently. API limitations: Since I challenged myself to come up with the product with 0 money spent, I had to figure out how I could get the api to work. The problems i faced with this, free tiers on gemini has very low rate limiting, no real testing can be performed.

Accomplishments that we're proud of

The system produces clear and structed decisions instead of vague feedback. The brutal mode gives direct and realistic critiques that feel closer to real world investor thinking. The UI is simple and focused, making the flow from input to decision easy to follow. At the same time, this is still a heuristic system. The outputs are not benchmarked against real world outcomes, so it should be used as a directional tool rather than a definitive answer.

What we learned

Prompt engineering is critical for turning a general AI model into a structed system. Small changes in prompts can significantly affect tone, consistency, and usefulness. OpenRouter made it much easier to experiment with models an iterate quickly.

What's next for BUILD or KILL

Improve consistency and reduce output variance. Introduce benchmarking against real startup outcomes or datasets. Add more transparency into how scores are generated

Built With

• nextjs
• opencode
• openrouter