become{AGI}

The gameplay revolves around enhancing your AI model by training it with the current Dataset, which is structured into a grid of 32 bytes. Each byte is has a specific type of knowledge, varying in rarity. The scarcer the knowledge type, the greater the amount of knowledge gained by the player. However, while playing, they risk absorbing "Noise," which impedes their model's progression. As various models train on the Dataset, each byte undergoes wear and tear, gradually yielding less knowledge until the Dataset is refreshed automatically every hourl. The most interesting part is the sharing dynamic: players have the ability to share knowledge bytes via Warpcast. By default, both the sharing player and the recipient player benefit from this exchange, gaining knowledge based on the original byte's rarity, condition, and the sharer's experience level. Players can opt to inject "Noise" into their shared bytes, causing recipient models to lose knowledge while their own remains unaffected. When the shared data is displayed on Warpcast, it's content is hidden, adding an element of mystery. The interacting player can change the outcome by liking, recasting or following the sharer: liking doubles the effect, recasting turns you into a source and if there is mutual following the content is revelead. This makes it possible to passively level up by sharing Frames and gives players in-game advantages when engaging with Frames.

This project uses Frames.js in a NextJS app. We use Vercel Storage (Postgres and KV) for storing game data and Cronjobs to launch new datasets every hour.

Frames.js made it very easy to do state management, user flows management and create UI.

We also used Pinata API for fetching players account details and game Analytics. We also integrated Airstack to create more complex queries like player owned POAPs.