Everybody Lies

What This Game Is Teaching Me

By Joey Bess··3 min read·
Lesson

I started Everybody Lies as a mystery game idea that sounded simple in my head.

A town. A crime. A cast of suspicious citizens. Everybody with a secret. The AI would create the town, the people, the clues, and the little lies that make a mystery fun.

That was the dream version.

The real version has been more like: the town won’t generate, the voices won’t play, the browser refuses to autoplay, the phone runs out of storage, and I sit there wondering if I have accidentally built a very expensive loading spinner.

I say that with affection. Mostly.

This project has taught me that building with AI is not the same as waving a wand. It is more like working with a very fast, very confident assistant who sometimes hands you a beautiful cake and sometimes hands you a plate of flour.

The first big lesson has been make the first version smaller than your imagination.

My imagination wanted a rich mystery world with unique towns, unique citizens, unique images, unique voices, and a proper investigation board. All of that still matters. But the first version needed to be simpler: can a person start a mystery, meet suspects, collect clues, and move the investigation forward?

That is the core loop. In plain English, the thing you do over and over that makes the game a game.

If that does not work, the fancy parts do not save it.

The second lesson is that AI can create too much, too easily.

That sounds like a nice problem, but it is still a problem. A mystery game needs structure. The town can be different every time, but it cannot be nonsense. The citizens can be strange, but they still need to fit the story. The secrets need to matter. The crime needs to make sense.

At one point the town generation broke because the app was being too strict about what kind of information it would accept. I had to loosen that up and let the AI’s answers be cleaned up instead of rejected outright. I do not fully understand every technical piece of that, but I understand the lesson: when you ask AI to be creative, you also need a way to tidy up after it.

The third lesson is that the magical parts are often held together by very practical plumbing.

I added AI-generated images for each town. Lovely idea. Then I hit a storage problem because the app was trying to save too much image data on the phone. The fix was not glamorous. It was basically: stop saving the heavy version of the picture in local storage. Local storage is just a small cupboard inside the browser. I had stuffed it like a holiday closet.

I added unique voices for citizens, which felt wonderfully alive when it worked. But then some voices did not work. Then fallback voice logic had to be added, which is a plain way of saying: if the fancy voice fails, use another one instead of going silent.

Then autoplay caused trouble because browsers often block sound until the person taps something. This is not the app being dramatic. It is a safety rule, so websites cannot start yelling at you the moment you arrive. The fix involved unlocking audio after user interaction and reusing one audio player behind the scenes.

Again, not glamorous. Necessary.

The fourth lesson is that “almost done” is a dangerous phrase.

I can get something working on my screen and feel a little puffed up. Then I look at what it would need for the App Store and come back down to earth. Smooth page transitions. Better handling when something fails. Clear first-time experience. Less roughness. Less “please understand, I’m still learning.”

A real person using the app should not have to know my excuses.

What would I tell someone like me, starting this later in life with no coding background?

I would say: you are allowed to be confused. You are allowed to ask simple questions. You are allowed to build ugly first versions. You are allowed to think something is broken forever and then discover it was one small rule in the wrong place.

But I would also say: do not let the AI talk you into building the whole cathedral on day one.

Build the door. Open it. See if anyone wants to walk through.

Small takeaway: the magic is real, but the patience is still required.