Pushup RPG

Most fitness apps record effort after the hard part has already happened: the person first has to decide to exercise.

I wanted to test a different loop. What if the exercise itself were the game input?

I am an independent developer in Kazakhstan, and I built Pushup RPG around that question. You prop up a phone, its camera counts a push-up or squat on-device, and the completed rep becomes an attack against a monster.

The premise sounds simple. Building it changed how I think about product input, feedback, game balance and privacy.

1. A rep is not a button press

A button gives the product one clean event. A camera gives it uncertain observations: joints disappear, the phone moves, light changes, frame rates vary, and the bottom of a push-up is often where the arm is hardest to see.

I stopped treating a rep as two threshold crossings and started treating it as an ordered sequence: establish the top, observe descent, collect credible bottom evidence, observe ascent, then confirm completion. The larger lesson was that products built around physical or sensor input need to preserve causality, not just detect isolated values.

2. Detection quality is only half the product

When nothing counts, the user cannot inspect a debug trace. They just see a phone on the floor and a game refusing to respond.

“Step back” is useless if camera frames have stopped. “Tracking” is misleading if the necessary joints have been outside the frame for the whole set. I had to separate camera delivery, body visibility and movement progress—and design specific guidance for each state.

The best underlying model still feels broken if the product explains a failure badly.

3. Game balance becomes a physical-work budget

In an ordinary RPG, adding health to an enemy mostly makes a fight longer. Here it asks the player for more real repetitions.

That means every encounter has two budgets: entertainment time and physical work. They overlap, but they are not the same. A battle that is merely tedious with touch controls can become physically unreasonable when the attack input is a push-up.

4. Privacy wording has to match the architecture

Rep counting runs on-device. Camera video and pose landmarks are not sent to the game's servers for rep counting. Optional clip recording is off by default and remains local unless the player chooses to share it.

That constraint is not a footer detail. It limits the technical options, makes device variance my responsibility, and changes how the feature should be explained. I learned to avoid comforting but overbroad claims such as “nothing ever leaves the phone” when the wider app also has accounts, social features or optional sharing.

5. The mission needs a measurable operating question

The long-term ambition is intentionally oversized: help double the number of push-ups people choose to do worldwide.

The near-term product question is much smaller: does making each rep create immediate visible progress help someone start a set and return for another one?

For other founders building consumer habit products: which metric would you optimize first here—completed first workout, D7 return rate, or reps per active user—and why?

Pushup RPG is free on iOS and Android if you want context: https://pushup.quest/get