• 

  Loading Screen (Working version)

• 

  Landing Page (Just Play, Choose Categories, Settings)

• 

  Setting Page 1

• 

  Settings Page 2

• 

  Settings Page 3

• 

  Choose Category Page 1

• 

  Choose Category Page 2

• 

  Example 1 (A/B Choice)

• 

  User Chose A

• 

  The user click "What Would Sonar Do?" and got an a humorous AI response

Would You Rather - Road Trip Edition

Inspiration

On road trips, my kids play "Would You Rather" with questions like "Would you rather eat 100 bees or 1000 red ants?" One choice is usually less horrible than the other. I thought it would be more interesting to make it educational with questions like "Would you rather put your hand in a particle accelerator beam or in a microwave for 20 seconds?" (The particle accelerator is actually safer, which is counterintuitive and worth discussing.)

The existing "Would You Rather" apps are pretty terrible. They're either monetized to death, have boring questions, or just show you vote percentages when you pick something. Basic stuff. This one is different because the AI explanations are actually informative and sometimes funny.

What it does

It's a mobile game that gives you weird dilemmas and explains why your choice was good or terrible. You get two options, pick one, and can ask "What Would Sonar Do?" to get an explanation. The explanations try to be educational and amusing, like xkcd comics but for decision-making.

The app has different categories (food, sci-fi, horror, etc.) and content ratings from family-friendly to adult. It works offline with pre-loaded questions and online for fresh content. The goal is starting conversations and maybe learning something along the way.

How we built it

React Native/Expo app with a Supabase backend. The interesting part is the caching system: when someone gets an AI explanation for the first time, we save it so the next person doesn't need a new AI call. Keeps costs down and responses consistent.

Used Perplexity AI for generating content and explanations. Had to switch from reasoning models (too verbose) to smaller ones that actually work for quick responses.

Challenges we ran into

The AI response formatting was tricky. Reasoning models gave us comprehensive but incredibly wordy explanations that killed the pacing. Smaller models worked better for the casual gaming experience.

Balancing offline versus online functionality took some work. The app needed to be useful in areas with bad cell coverage but still benefit from cloud features when available.

Content rating was harder than expected. Even my teenagers thought the initial G-rated content was boring, so we added edgier options with proper filtering.

Also, we had to shape the AI responses appropriate to the content rating. We didn't want to tell an 8 year old they are a douche-nozzle.... Well, we do but not this way ;)

What we accomplished

Built a functional app in three days that works offline and online. The caching system reduces AI costs while maintaining quality responses. It's actually entertaining enough that my normally serious Lithuanian fiancée found it amusing.

The database architecture automatically tries to reuse community-generated questions before creating new ones, which should scale reasonably well.

What we learned

Supabase is quite good for rapid prototyping. Perplexity can generate genuinely funny content when prompted correctly. Reasoning models aren't great for gaming applications - simpler tools for simpler tasks.

Also learned that "family-friendly" content can be legitimately boring if you're not careful about it.

What's next

Working on user-generated content with AI moderation. People can submit their own questions, AI checks them for appropriateness, and good ones get added to the general pool.

Planning to expand this into a collection of road trip games:

• Location-based "Would You Rather" questions triggered by GPS
• "What's around us?" using AI and location APIs for local information
• License plate identification game using computer vision. Collect all 50.
• Local trivia based on current location
• "Two truths and a lie" generator

Need to figure out sustainable monetization that doesn't involve ads or subscription hell. I hope that I can figure out monetization that seems reasonable without ads. I'd rather not put it out if I have to use ads.

---

A slightly smarter version of a car game that might actually teach you something.

Built With

• cursor
• expo.io
• perplexity
• react