AI Games and Virtual Pets: From Creatures to Generative Agents

What this article covers

• What powers games like Creatures, Tamagotchi, and The Sims.
• The history of artificial life in gaming, from Norns to today.
• The current generation 2026/2027: Alife remakes, evolution sandboxes, and AI pets.
• Generative Agents: the research project that gives Sims real AI brains.
• What you can play, run, or build yourself.

Introduction: Does this actually exist?

Short answer: yes, and it has for decades. The idea of breeding, raising, and watching digital creatures grow intelligently predates the term “AI game” itself. What changed: we used to simulate intelligence with neural plasticity and rule sets; now creatures learn through reinforcement learning, physics simulation, or entire language models. The field is called Artificial Life, or Alife for short, and 2026/2027 marks a genuine revival.

The Predecessors: Creatures, Tamagotchi, and The Sims

Tamagotchi (1996): The prototypical virtual pet. A pixelated creature that gets hungry, sleeps, and dies if you ignore it. No real AI involved, just need bars and randomness, but it laid the foundation: emotional attachment to code.

Creatures (1996): The serious Alife game, technically revolutionary for its time. Your Norns had a simulated genome, a biochemistry with hormones and organ responses, and learned through reward and punishment (praise and scolding via mouse clicks). Intelligence was inheritable: clever parents produced clever Norns. No script dictated what a Norn did; behavior emerged from the simulated nervous system. Still the standard against which true Alife games are measured.

Little Computer People, Dogz/Catz, Seaman, Nintendogs, Black & White: The extended family. Seaman on the Dreamcast responded to voice input, Black & White let a godlike creature learn from your actions, Nintendogs recognized voice commands through a microphone.

The Sims (2000): The most successful variant. Sims have needs, personalities, and make autonomous decisions, but according to fixed rules. No learning, no memory, no real AI; the magic lies in simulating social situations.

The New Generation 2026/2027: Alife Returns

This exact genre is making a comeback right now:

• Creatures (Steam Remake, 2026): The original returns with genuine Alife mechanics: breed, teach, and guide Norns through new worlds, including rewind and turbo functions.
• Precursors: Origins of Folklore: The spiritual Creatures successor, in development: real genetics, real biochemistry, real AI cognition, species drawn from world folklore. Intelligence must be bred, not programmed.
• The Neeblarium (2026): A peaceful Alife sandbox. You place small creatures (Neeblies) into biomes and watch families, lineages, and mutations emerge without direct control. Emergent evolution instead of scripted behavior.
• ANLIFE: Motion-Learning Life Evolution (Feb 2026): Physics-based evolution. Block creatures learn to move, efficient movers survive and pass on their bodies. Observe, don’t control.
• FlyMuse (2026): A “bio-grade desktop pet.” A simulated fruit fly connectome (a real mapped nervous system) as a pet on your computer that learns, dreams, reproduces, and remembers you.
• Desktop Pets in general: Shimeji successors and LLM-based desktop companions combine that Tamagotchi feeling with real conversational AI.

Generative Agents: When Sims Get Real Brains

The most exciting project isn’t a commercial game but research. The Stanford study Generative Agents (2023) populated a small Sims-like town with 25 LLM-driven characters. Each agent had memory, reflection, and planning. They woke up, cooked breakfast, went to work, formed opinions, made acquaintances, and independently organized a Valentine’s Day party based on a single prompt.

The whole thing is open source (GitHub: joonspk-research/generative_agents) and runs locally: a Python server plus a browser view of the simulated town. You can watch agents and intervene using natural language. For BotServ readers, this is doubly interesting: the architecture (memory stream, reflection, planning) powers modern AI agents. Smallville is basically an agent playground disguised as a game.

Follow-up projects and commercial adaptations use the same pattern: LLM-powered NPCs in games that don’t rely on dialogue trees but actually converse.

What Makes “Real” AI in Games?

Three technical generations in overview:

1. Rule-based AI (Tamagotchi, The Sims): State machines and need logic, predictable, no learning.
2. Adaptive AI (Creatures, Black & White, FlyMuse): Genetic codes, reinforcement learning, physics-based evolution. Behavior emerges rather than being scripted.
3. LLM AI (Generative Agents, modern desktop pets): Memory, planning, natural language, characters with genuine personality and recollection of you.

Play and Experiment Yourself

• Original Creatures runs on modern hardware via GOG/Steam and the upcoming remake.
• Generative Agents run locally: Python plus an LLM key, Smallville in your browser, for anyone interested in agent architecture.
• The Neeblarium and ANLIFE are Steam purchases for observers rather than players.
• Desktop pets like FlyMuse for a permanent digital companion.
• Build your own: a Tamagotchi-like creature with a local LLM as its “brain” is a doable weekend project. See our article on AI bots with Ollama for the technical foundation and IRC-Coding.de for coding tutorials.

Common Pitfalls with AI Games

Check “AI” in marketing carefully. Many games tout AI but mean rule-based automata. Real Alife shows itself through emergent, heritable behavior.

LLM pets require maintenance. Language model companions forget without memory systems, and dialogue costs API tokens. Local models via Ollama solve this.

Alife means loss of control. If you want to steer everything, you’ll be disappointed. The appeal lies in observation; creatures do what their simulation drives them to do.

Don’t underestimate emotional attachment. It sounds silly but it’s real. You bond with code that remembers you. That’s the whole point of the genre.

Further Reading on AI Games

• IRC-Coding.de - Build your own AI pets, bots, and agents with over 500 coding articles.
• IRC-Mania.de - Technical background and tutorials.
• IRC-Security.de - Security when your pet runs online or your model is exposed.
• IRC-FAQ.de - Quick answers.
• Generative Agents on GitHub - Run Smallville yourself.

Related articles on BotServ.de: AI Agents Learning Path, IRC Bot with Local LLM, and AI in Cybersecurity.

FAQ: AI Games and Virtual Pets - Common Questions

Do games like Creatures still exist today?

Yes. A Creatures remake launches on Steam in 2026, Precursors: Origins of Folklore is the explicit spiritual successor with real genetics and biochemistry, and titles like The Neeblarium and ANLIFE bring evolution sandboxes to market.

How does Creatures differ from The Sims?

The Sims simulate humans using fixed rules and need logic. Creatures simulated life itself: Norns had genomes and biochemistry, learned through reward, and inherited intelligence. The Sims show scripted behavior; Norns show emergent behavior.

Is there a modern Tamagotchi with real AI?

Desktop pets like FlyMuse use a simulated real nervous system (a fruit fly connectome), and LLM-based companions blend pet feeling with genuine conversational ability.

What are Generative Agents?

Stanford research from 2023: 25 LLM-driven characters in a Sims-like town with memory, reflection, and planning. The agents independently organized social events. Open source and runs locally.

Can I build such a virtual pet myself?

Yes. A small program with need values, memory, and a local LLM (Ollama) as personality is achievable. The technical foundation is in our AI bot article.

Are AI NPCs in games already real?

Partly. LLM-driven NPCs exist in mods and indie games; AAA titles increasingly use generative dialogue. Generative Agents proved the potential, and commercial integration is growing.

What is Artificial Life?

A field that artificially simulates life: genetics, evolution, ecosystems, emergent behavior in software. In games, Creatures and Alife sandboxes are the prominent examples.

Where do I start if I like Creatures?

With the Steam remake or the original, then move to The Neeblarium for evolution and ANLIFE for physics-based learning. For the hardcore: run Generative Agents locally and operate a Sims town with real AI.