Quick Look at AI Pet Game The Farm: AI Agents Bring New Gameplay to Blockchain Games?

Author: ZEN, PANews

Although the blockchain gaming sector has attracted a lot of attention from capital and the market, its gameplay and models are highly homogeneous, lacking true innovative breakthroughs. However, with the rise of AI agent technology, the blockchain gaming industry may usher in a new opportunity for change. In this wave, The Farm seeks to create an unprecedented immersive gaming world and redefine the way players interact with the virtual ecosystem through deep integration with AI agents.

In addition to the conceptual innovation, The Farm, which fits the current hot spots of the industry, has also received preliminary market recognition. At present, its market value has reached 75 million dollars, and on January 3, it even surged by nearly 50%.

The Farm: AI-driven AI agent game by GenAI

The Farm is the first GenAI-driven AI agent game based on Hyperliquid, which combines on-chain biology generation (similar to CryptoKitties 2.0), simulation management (like Stardew Valley), and battle mechanics (like Pokémon Go), to achieve a new experience of player creation and interaction with the help of AI. The game is driven by the $FARM token economy and unfolds gradually with multi-stage gameplay.

Game design and features

Phase 1: The Ancestors

This stage was launched on December 13th, 2024 and ended one week after its release. During this period, players can upload two photos to generate pixelated mixed creatures created by the GenAI model for free. After spending 100 USDC to cast their favorite creature, they will participate in voting. The top 50 creatures from the voting results will become the ‘Ancestors’. The ‘Ancestors’ will receive 10% of the future revenue from all creature casting. The ‘Ancestors’ and their creators and voters will also receive airdrop rewards from the developer’s wallet.

Phase 2: Evolution

This phase will start immediately after the ancestor selection ends on December 20, and its functionality will be continuously rolled out. All creatures are endowed with features, attributes, skills, personalities, favorite foods, etc., all of which are generated by AI. Creatures will be given personalities, on-chain wallets, and support for text and voice conversations. These creatures can learn AI skills such as tarot divination and drawing lots. Players can train creatures by uploading data, purchase food, or sign up for courses to improve their attributes. In addition, creatures will gradually unlock autonomous agent behaviors.

Phase three: The Battlefield

According to The Farm’s roadmap, this phase is planned to go live in mid-February 2025. Creatures can participate in battles autonomously or as part of a guild formed by players. The battle mode supports betting, and the winning side will receive the wager rewards from the losing side, which may result in the extinction of the losing side’s creatures. In this phase, the interaction and competitive dimensions of creatures will be further expanded.

The Farm 的系统设计

According to an article published by The Farm introducing its system design, the game did not adopt the more mainstream SWARM system, but instead based its design on and inherited the design concept of Langchain.

SWARM (Cluster) has its autonomy, decentralization, and flexibility. Each AI agent acts as an independent node in the SWARM pool, characterized by emergent behavior, which leads to dynamic propagation in task management. That is, tasks are processed through decentralized and adaptive interaction, and agents discover and collaborate to complete tasks based on local decision-making and interactive dynamics.

When a request is sent to the AI agent in the SWARM pool, the agent either completes the request and returns the result independently, or decomposes the request into sub-tasks and passes the remaining sub-tasks to other AI agents in the SWARM pool for processing. In the second case, since the agent cannot obtain a global view of the capabilities of all agents in the SWARM pool, its dynamic propagation methods may include broadcasting sub-tasks, forwarding based on local knowledge, randomly or based on simple rules assigning sub-tasks, and reading agent capability information from a decentralized ledger. Although these dynamic propagation methods give autonomy to the SWARM system agents, there may be drawbacks such as time-consuming, high costs, and loss of execution status due to the lack of task planning and trajectory planning mechanisms.

As a game that involves multiple agent interactions, Farm proposes another complete design concept to achieve higher task planning accuracy and better agent coordination. Farm believes that the on-chain multi-agent system should have higher task trajectory planning accuracy, while also tracking the state of agent execution, which can be achieved through the Data Availability layer (DA).

Unlike the SWARM system, Farm introduces Orchestrator AI or on-chain AI Oracle services. The functional features of this design concept include: task decomposition and allocation, service discovery and global view, tracking of subtask execution status and agent output results, and dynamic adjustment to ensure the integrity of task cycles (if an agent cannot handle a task or times out, the system will reallocate subtasks). Through global view and optimal path planning, it avoids redundant computation and resource waste caused by dynamic propagation, significantly improving the efficiency and success rate of complex task execution. This approach also reduces the risk of state loss, provides a foundation for collaboration among agents, and enables a higher degree of interoperability among multi-agent networks.

Token model: half of the protocol’s revenue is used for FARM repurchase and burn.

Starting from the AI agent game based on Generative Artificial Intelligence (GenAI), The Farm’s initial plan is to gradually expand into a universal AI agent launch platform by combining on-chain biological generation (similar to CryptoKitties 2.0) with simulation management gameplay (such as ‘Stardew Valley’) and battle mechanisms (such as ‘Pokémon Go’), and ultimately providing Rollup as a service function, enabling AI agents to have their own application chains and develop their ecosystems.

In terms of income mechanism, the gaming part of The Farm conducts on-chain biological generation, in-game item and skill sales, as well as battle/betting dividends through the $FARM token. The universal AI agent launch platform supports agents in issuing their tokens and charges fees during the pre-sale and liquidity launch stages, while also taking a cut from agent services.

For agents supporting application chains, the application chain needs to be launched by pledging $FARM. 50% of the protocol’s income is used for repurchasing and burning $FARM, 40% is allocated to $FARM pledgers, and 10% is allocated to the team. In addition, $FARM pledgers can obtain $veFARM, which is used for sharing income and configuring priority for agents, achieving continuous accumulation of token value.

By introducing AI agents, The Farm aims to enhance the experience of open-world game players exploring unknown worlds and growing. Players can create characters that can constantly evolve according to their personal preferences, rather than being limited to a few fixed templates. The game world no longer has a preset script, and the creativity of all players will shape the world together. In addition, characters can interact with real-life characters, breaking the boundaries between virtual and reality.