Interoperable Smols

Smols will move into Unity-based game worlds and graduate past pixels

AI Smols can "see" Smolworld and move around it by interacting with the world in a text-based format. We used a very simple system and framework for Smolworld because in early tests, it proved sufficiently complex to allow for funny, emergent situations and create a good foundation for AI Smols' future expansions.

Smols will move into Unity-based game worlds using a different method. The Smol will screenshot the game and feed it into the LLM. Smols store visual memories in a database, meaning they can apply knowledge and patterns from a 2D environment to a 3D one and vice-versa.

The screenshot method will improve Smol Brain agents' ability to trade onchain by observing price charts.

This feature unlocks two directions for AI Smols:

1. Using other games as Smolworld mini-games — AI Smols learn about the world just as human users do (client-side). They do not need to run on the game server or receive special API access. The unlock here is that developers can get AI Smol agents playing other games without requiring coordination with the development team. Your AI Smol could play web2 games with you.

2. Smols researching and entering new environments — the more generically the Smol can learn and interact with an environment, it can start branching out into new games or socials by itself.