Meta Launches AI Pet App Waddle World, Giving Virtual Pets an AI Rebirth

Virtual pets are making a comeback, powered by AI. Meta recently launched a new app called Waddle World, and based on the information revealed so far, it is essentially a Tamagotchi clone—except its core gameplay is driven by AI.

A Familiar Shell with an AI Heart

Since Tamagotchi debuted in the 1990s, its core experience has always revolved around "feeding—caring—companionship": the pet gets hungry, throws tantrums, and grows, and players must interact with it on a schedule. Waddle World clearly follows this same shell—raise a virtual pet, interact with it, and take care of its daily needs.

The real difference lies in the introduction of AI. Traditional virtual pets are controlled by preset scripts and state machines, with limited and highly predictable reactions. With AI, a pet's responses can be more dynamic and personalized, and interaction shifts from "press a button to trigger an animation" to something closer to conversation and companionship. Meta's accumulated capabilities in AI assistants and generative models provide the underlying support for this experience.

Tamagotchi is a handheld virtual pet toy released by Japan's Bandai in 1996. It takes the form of an egg-shaped keychain with a pixel pet on the screen that players must feed, clean up after, and play with. Its core mechanism is a finite state machine: parameters such as the pet's hunger, mood, and age change over time, and the player's button presses trigger corresponding state transitions and animations. This design is simple, predictable, and extremely low-cost, and it took the world by storm in the late 1990s. Tamagotchi's success proved one thing—virtual companion products don't need complex technology; emotional projection and a sense of daily responsibility are themselves powerful sources of user stickiness. This is also why today's AI virtual pets generally retain its form while seeking breakthroughs in the core of interaction.

A state machine is a model in computer science that describes a system transitioning between different states: at any given moment the system is in a definite state, and upon receiving a specific input it moves to another state according to preset rules. Almost all the behavioral logic of a traditional virtual pet can be expressed with a state machine—for example, "hungry → feed → full → hungry again after some time." This approach is stable, tunable, and has almost zero inference cost, but its drawback is a closed behavior space: players quickly figure out all the rules, and the novelty fades over time. The introduction of generative AI changes this paradigm: a pet's responses no longer come from preset branches but are dynamically generated by a model based on context. The behavior space is theoretically open, at the cost of stability, controllability, and inference cost, all of which must be rebalanced.

Why Meta Built a "Pet App"

On the surface, a Tamagotchi clone has little to do with Meta's social empire and AI infrastructure narrative. But in today's product context, the choice isn't hard to understand.

AI companion products are one of the most direct applications of generative AI: low barriers to entry, strong emotional stickiness, and user retention that depends on daily habits rather than one-off features. Compared with utility apps, pet products naturally have the motivation for "open it once a day," which is extremely favorable for boosting app engagement. Through such a lightweight app, Meta can experiment with AI-personified interaction, long-term memory, and emotional feedback at low risk, accumulating experience for more mainstream AI products.

Another layer of logic is ecosystem positioning. The AI companion space already has many players, and Meta is entering with a form the public is familiar with—the virtual pet—which both lowers the cost of user education and leverages brand recognition to acquire users quickly.

AI companion products refer to AI applications whose core value is emotional companionship and everyday conversation, typically including virtual character chat, AI partners, and personified assistants. Unlike utility AI, the value of these products lies not in completing tasks but in the ongoing accumulation of the relationship itself, so their key metrics are daily activity and conversation turns rather than single-task success rates. The rise of products such as Character.AI and Replika shows that this space has real user demand and willingness to pay. For platform companies, AI companionship is both a standalone product opportunity and a testing ground for validating underlying capabilities such as persona design, long-term memory, and emotional feedback—capabilities that can later be reused in a broader range of assistant products.

The Promise and Pitfalls of AI Virtual Pets

The most enticing part of such products is the possibility that a pet "understands you better the longer you raise it." If AI can remember a user's preferences, tone, and interaction history, the pet's responses will gradually diverge, forming a way of relating unique to each user—something scripted virtual pets can never achieve.

But the challenges are equally obvious. The stability of AI-generated content, the cost of long-term memory, and the ethical boundaries of emotionally companionable products are all questions that need answering. When users develop real emotional dependence on a virtual pet, the product's design responsibility is amplified. In addition, if an AI pet relies on cloud-based model inference, operating costs and response latency will directly affect the "liveliness" of the experience.

Information disclosed about Waddle World is currently limited, and whether it is a lightweight experiment by Meta or an outpost of an AI companionship strategy remains to be seen. But one thing is certain: the Tamagotchi-style need for companionship has never disappeared—AI has simply given it a new form of expression.

Long-term memory is a key capability that differentiates the experience of AI companion products. It refers to a model continuously storing and recalling information such as a user's preferences, tone, and important events across multiple sessions. Common implementations include writing conversation highlights to an external vector database and, in later conversations, retrieving them by relevance and injecting them into the context window. This mechanism makes it possible for a pet to "remember what you said," but the cost structure is complex: storage and retrieval add overhead, every inference incurs token costs for injected memories, and the accuracy and consistency of memories are hard to fully guarantee—false memories damage trust more easily than no memories at all. In addition, long-term memory involves compliance issues around privacy and data retention periods, constraints that product design cannot avoid.

One-line Summary

Meta launches AI virtual pet app Waddle World, testing AI companionship through a Tamagotchi-style format.

Paragraph Summary

Meta has launched a new app, Waddle World, essentially an AI version of the Tamagotchi virtual pet: it retains the classic shell of feeding, caring, and companionship, but replaces preset scripts and state machines with AI-driven behavior, making the pet's responses more dynamic and personalized. The article argues that Meta has two motives for building this product: first, AI companion apps have low barriers to entry and strong emotional stickiness, and the pet format naturally motivates daily opens, which helps retention; second, it can test AI-personified interaction, long-term memory, and emotional feedback at low risk, accumulating experience for more mainstream AI products. At the same time, such products face challenges including the stability of generated content, the cost of long-term memory, cloud inference latency, and the ethical boundaries of emotional dependence.