Pepper

Pepper Pet World designed by Eliza Pan

Day Care 🏫 (every pet in the world)

Pets nobody has played with for a week rest here, safe and looked after. Tap one and it comes straight home to the ocean.

Pepper's character prompt (You're the author!)

History

Every save lands here. Tap one to read it — you can always bring an old Pepper back.

Pepper's memory book (what your pet remembers between chats)

When you press ✨ new chat, Pepper rereads the finished conversation and writes the important bits here — names, favorites, promises. You can fix or add memories yourself, too.

Old chats

Every finished conversation lands here. Tap one to read it again.

Pepper's makeover 🎨

Tap anything! It changes right away — on every computer, and in your bubble on the who's-playing ocean.

🎨 body

LLM Transcript

Pepper is played by an LLM — a large language model (claude-haiku-4-5). Every reply is one HTTPS request to the Anthropic API. Each entry below is one request, in full: the system prompt (the model's standing instructions), the messages array it read (role user is your side, role assistant is Pepper's), and the raw text it generated. Lines stamped [EVENT] were injected by the game — not typed by you. The model must answer with the JSON {"say": "...", "do": "..."} — say is the words ("" means it chose to say nothing) and do is a move for Pepper's body to act out ("jump", "blush"... or "" for none).

🐾 The Pet

Feed and play from the shelf to keep your pet's tummy and happiness up — overfeeding a full tummy, or letting it hit empty, turns your pet sick until you give 💊 medicine. 🌙 Bedtime slows the drain for the night, a tap on the tank is a tickle, and 🎨 Makeover restyles color and shape anytime. Your pet lives in the cloud, so it's the same on every device.

💬 The Chat

Type in the chat to talk with your pet; it reacts to whatever happens in the game, too. Pressing ✨ New chat consolidates the conversation into 🧠 Memories — which you can view and edit — then starts fresh. ✏️ Author rewrites your pet's character prompt, and 🔍 Debug shows the request behind each reply.