Inspiration
Based on the game called Pictionary from a British TV Show we were inspired to create a game where we could see how Humans are able to compete against AI. Nowadays it seems as if it has been able to replace everyone in every job sector. We made this game to prove that humans are still capable of winning even when it seems like AI is the replacement solution.
What it does
We made "outsmart.io". The game where you 'the human(s)' compete against an AI to see who can win. To beat the ai you have to draw to confuse it but make it easy for the humans to understand. Stretching your creativity and making it hard to play without thinking.
In our day to day we tend to think rarely when using ai. This is a fun twist that makes you feel immersed in the AI experience while making it feel like less of a feature but the thing driving the game forward to the win condition.
How we built it
To make the frontend we drew each sprite on pixilart.com and rendered each frame to create a html based frontend which can be ran on any device and all browsers. To make the game more fun we wanted it to be multiplayer. We used a QR Code module for the joining experience reducing join friction. The host joins the game on a large screen allowing players to guess on their phones directly. We used gemini 2.5 flash vision model to view the canvas every 10 seconds. It returns its guess to the frontend and it is parsed to produce a tag on the screen showing what the AI guessed.
Challenges we ran into
- The AI is often too smart and able to guess correctly too early. To make it harder, we mask what the AI can see at any given time
- If the screen is streamed to the AI it is far too hard to beat it on-time due to faster processing so we made it only able to check after the clock passes the screen (10 seconds) making it check every 10 seconds until a round times out or ends
- It is quite hard to draw pixel art on desktop with a touchpad, to solve this we spent a large portion designing the pixel art
Accomplishments that we're proud of
We were able to learn many frameworks within this time and improved our skillset during this hackathon. We learnt how to edit videos better and made many friends during the recording process. Overall was a great experience and we are proud to say we did this.
What we learned
Game Development is not easy, design, code, pitching and submitting. It's a hassle and very hard to have all skills as a developer. We realised during this experience that we needed to learn more skills and improve our range of abilities to be able to create anything we envision easily. AI tools were able to help us during the hackathon such as Claude Opus when designing the backend architecture, however, without it the process would've been extremely tedious. If we had the server side understanding of how to setup multi user connections to servers for gameplay it would have been very simple. During this hackathon we learnt how to do all of this which was not just beneficial but it helped us become resilient.
What's next for Outsmart.io
Outsmart.io was a very fun project to make and we hope to see it being played. We wanted to begin working on using it to fix the solution of data annotation fatigue. If we used data from players to train AI models such as vision models it would improve their accuracy significantly. With this project it would be nice to see how it transforms from a game into a solution to tech.
Built With
- gemini
- javascript
- qrcode
- render
Log in or sign up for Devpost to join the conversation.