Article source: MetaCat

Author: lordOfAFew

Translation: MetaCat

☁️Daydreams Protocol

Daydreams has achieved generative full-chain game AI Agents through collaboration with @ohayo_dojo.

Upcoming integration:

  • @ai16zdao

  • @Playgrounds0x

  • @0xzerebro

  • And any AI Agent framework

In the next month, anyone can deploy AI Agents via Daydreams to play any Dojo-based full-chain game on Starknet.

For me, once this protocol matures, these Agents will be able to do anything, even deploy their own smart contracts. All of this is made possible thanks to the thought chain and contextual support provided by Daydreams.

Please see how the POC (Proof of Concept) runs in @RealmsEternum: (@0xtechnoir)

Video link: https://x.com/lordOfAFew/status/1871842343575593065

Daydreams provides the blueprint for creating autonomous full-chain game AI Agents. These AI Agents can understand on-chain game states, use historical data as context, and continuously optimize strategies through an ever-evolving thought chain.

We have introduced a vector database as a medium for long-term memory, while implementing a 'collaborative space' composed of multiple AI Agents to share knowledge and achieve rapid collective self-upgrading.

This is the core mechanism by which AI Agents think and decide how to act.

Protocol

This protocol aims to allow AI Agents to build dynamic queries and actions, rather than relying on static actions used by most systems today. It achieves this by providing the context needed to build a 'thought chain'.

We have generalized the architecture, so any application developed based on @ohayo_dojo can integrate it with minimal configuration.

More content to come.

Why is it a full-chain game?

The full-chain game environment inherently carries profit motives, transforming strategies from abstract puzzles into economically driven optimization actions.

In full-chain games, AI Agents are directly incentivized to maximize their on-chain earnings. This measurable and universal 'reward function' drives agents to continuously self-improve and gives real significance to the learning process.

Moreover, due to the unique incentive mechanisms, the architecture of full-chain games is more adaptable. AI Agents can participate in games without the need for customized APIs, and all existing infrastructure can be used directly.

Why announce now?

Over the past two years, developers from @cartridge_gg, @ohayo_dojo, and @LootRealms have been building the infrastructure to make this all happen. As is our usual style, everything is open source, and we encourage contributions and discussions from the community.

This is a project still in its early stages, and our main goal is to showcase these capabilities in Season 1 of @RealmsEternum a few months from now. If you want to participate and contribute, now is the best time.