DOG BONE JUMP

This post is my submission for DEV Education Track: Build Apps with Google AI Studio.

What I Built

I built A delightful 1920x1080 desktop landscape dog jumping game where players tap to guide their dog to catch bones.

A Bone Collector Dog arcade game leveraged Google AI Studio to turn a more than 12-step architectural blueprint into a fully functional, event-driven game complete with a score tracker, sound effects, and clean state management.

My Preplanned Instructions

To generate the core logic of the application, I provided Google AI Studio with a highly specific, 12-step foundational prompt:

1. Make a 1080x1920 canvas.
2. Spawn Dog emoji as player at bottom.
3. Spawn bone at randon place at screen.
4. Add touch screen for user to jump dog to the position of bone.
5. Store the score how many bones the dog has got.
6. Show the score at the top of screen on left side.
7. Show the score with bone emoji (example: score 3 will show 3 bones at top left screen.)
8. Make start Screen to start the game
9. Add start game button on start screen.
10. Add exit game button on top right corner at screen on gameplay.
11. Add Exit button at top right at screen at start screen to exit the game.
12. Add audio in game from audio/ folder using mp3 files.

Demo

Play

My Experience

For this i'm using Gemini 3.7 Flash

First Prompt

1. Make a 1080x1920 canvas.
2. Spawn Dog emoji as player at bottom.
3. Spawn bone at randon place at screen.
4. Add touch screen for user to jump dog to the position of bone.
5. Store the score how many bones the dog has got.
6. Show the score at the top of screen on left side.
7. Show the score with bone emoji (example: score 3 will show 3 bones at top left screen.)
8. Make start Screen to start the game
9. Add start game button on start screen.
10. Add exit game button on top right corner at screen on gameplay.
11. Add Exit button at top right at screen at start screen to exit the game.
12. Add audio in game from audio/ folder using mp3 files.

---

1. It suggested me 2 more instructions:

1. Ensure the audio loops during gameplay for a seamless experience.

2. Implement sound effects for each bone collection interaction.

3. It suggested me few tampletes to choose for my game UI. I chose the last one.

4. It stopped after doing UI, Didnt follow all the instructions and steps to make finished game.

5. It did the change but i could not see it.

6. I gave new prompt to make screen size 1920x1080 for desktop.

7. I had to refresh the entire screen then i it has followed all the instructions already.

8. I tested the game.

9. I gave another new prompt "after 10 bones +1 achivement and reset bone collection stack for the next 10 bones for next achievement."

9.It updated the game and did auto refresh this time and gave prompt notification got 2 errors.

1. I clicked 'Fix' button.

2. It fixed the errors and explained it in prompt window and updated the game with auto refresh again.

1. I gave a new prompt "dog should jump at the user's screen click/tap position".

2. It updated the game again with auto refresh and stored last score on screen shown.

3. First time it got a glitch that i hid start game but it gone back to start screen back once. This glitch happend 2nd time here coz it happened before at an update too.

4. I gave new prompt "keep the dog on ground, so remove the gap between dog and shadow."

5. It updated the game again in 58 seconds.

6. I gave new prompt "keep the dog on ground, so remove the gap between dog and shadow."

7. It centered the dog on shadhow horizontal only.

8. I manually updated the code myself by clicking **"Vew changes"** button in prompt that redirected me to the updated code showing me diff.
   

            transform: `translate(-50%, -85%) scaleX(${dogPos.scaleX}) rotate(${dogPos.rotation}deg)`,
            transformOrigin: '50% 85%',

to

            transform: `translate(-50%, -100%) scaleX(${dogPos.scaleX}) rotate(${dogPos.rotation}deg)`,
            transformOrigin: '50% 100%',

1. Game updated and got auto refresh but this time no glitch on start game.

2. This time game just ran without glitch and i realized "that glitch was there before coz i started game before it was fully ready in prompt.

3. I could not see it is not fully ready coz there was not progress indication in progress bar for that. It looked like it's ready. Here is the possiblity to improve it for Google AI Studio, if i understood it correctly.

4. I gave a new prompt "change the dog avatar to new random dog image at each achievement."

5. I ran for 172 seconds and updated the game as i wanted without that glitch this time coz i knew when to hit the "Start Button" this time.

1. On the prompt it shows Checkpoint "Vew changes" window after that it's ready.

2. Next prompt i gave "make shodow size and position update according to the dog position on the screen. (Example: if dog not on ground then shadow size decrease, if dog changing x position shadow also changes it to keep it always beneath the dog."

3. Again Gemini updated the game in 65 seconds but i got a bug or i would say an enhancement posibility in game logic.

Bug 1. It resets the dog image back to the very first image if clicks "Start Button" but keep the new dog if press .

Bug 2. It gives dog foot prints as the dog avatar image at achievement.

Bug 3. It repeats the old dog image at achievement.

1. I gave new prompt "Store the last updated avatar image of the dog."

2. Gemini updated the game in 110 seconds. Bug fixed.

3. I gave new prompt "do not repeate the old avatar new achievement."

4. Gemini updated game in 153 seconds. Bug fixed.

5. I gave new prompt "dog avatar at new achievement must be a full body dog image."

6. Gemini updated the game in 157 seconds and fixed the bug.

1. I gave new prompt "bone shouldn't float at one place in air. it should come from left going to right side if dog is in the right side of the screen center and should come from the right side going to the left if dog is in the lef t side of the screen. if the dog could not hit the bone in air then bone doesnt count."

2. Gemini fixed the bug in 124 seconds but found another bug to be fixed "dog isn't always facing front when jump".

1. I gave new prompt "dog should always face front when jump".

2. Gemini fixed the bug in 156 seconds.

1. I gave new prompt "dog should idle animation on ground by face and tail tilting."

2. Gemini updated the game in 114 seconds.

3. I gave new prompt "remove face tilt and add tung animation for idle state."

4. Gemini updated game in 114 seconds.

5. I gave new prompt "Move mouth triangle bihind the tung." Gemini failed here.

6. I gave new prompt "move the z index of tung up over the mouth." failed again.

7. Then i realized maybe that yellow triangle is also the tung from old design. so i gave new prompt "move the z index of tung up over the mouth
   "

8. Gemini ran for 9 seconds and ran out of free generations for today.

45. Finally i restored back to face tilt animation no tung animation for idle to remove the duplicate tung.

---

After finishing:

I had Google AI Studio accound for a mounth but i didnt started using it since i felt it looks easy but maybe i will not easy to use. For that reason of dout i just didn't even started testing it just coz i didn't have any specific reason to use it. But then saw the "Dev Education Tracks" post on the Dev.io Home page and i click on it to read and i just tried it right after that.

Google AI Studio proved incredibly effective at translating exact gameplay specs into a working code structure.

The Best Part: The AI perfectly handled the visual translation of the score system—converting a standard numeric counter into a dynamic row of visual bone emojis (3 = 🦴🦴🦴) precisely as requested.

What did i learn: At the time of sharing the project i realized preinstructions was empty in the right side window, still it worked without it. It could help in some common instructions like "give me reply in the for of diffs so can add the code my self." and "Reply to in devnagari." and so on.

Key Takeaway: Clear, structured requirements yield incredible results. Breaking my game loops down into 12 granular steps allowed the AI to deliver production-ready logic on the first try.

It was fun to use it for me. Thanks for the Dev Education Tracks.