DUMEE Whitepaper

Version 1.0.0 — February 2026

Abstract

DUMEE is a decentralized protocol built natively on the Solana blockchain and powered by the ZeroClaw autonomous agent runtime. It introduces a paradigm shift in generative AI interactability: shifting the model from renting centralized API parameters to verifiable ownership of individually prompted, yield-generating digital personas (Agents).

By embedding system prompts, domain expertise, and behavioral quirks into on-chain NFT metadata, DUMEEs act as gamified yet highly practical AI assistants bridging the gap between Web3 asset ownership and cutting-edge inferencing models.

The Problem

The current AI landscape is monopolized by a few hyperscalers. Users pay monthly subscriptions to interact with heavily sanitized, un-opinionated "helpful assistants." The profound value created by massive user interactions and real-time fine-tuning accrues entirely to the centralized entity.

  • AI is currently a rented utility, not an asset.
  • Default LLMs lack distinct, entertaining personalities or hyper-specialized domains directly accessible without complex prompt engineering.
  • Creators do not capture the value generated by their localized prompts and customized AI workflows.

The Solution

DUMEE tokenizes the prompt. A DUMEE is an NFT representing a persistent, character-driven AI agent. Each agent carries immutable metadata defining its tier (Common to Genesis), domain (e.g., Coding, Fitness), and personality quirk. By holding a DUMEE in your Solana wallet, you possess outright ownership of its digital labor output within the network.

Architecture

The platform is structured across three fundamental layers to ensure minimal latency and decentralized data integrity.

Layer 1 (The Ledger): Solana provides the execution environment for the `dumee-nft`, `dumee-token`, and `dumee-market` smart contracts. Handles all logic regarding spawning, fee distribution, and token burns.
Layer 2 (The Router): A high-throughput Hono API Gateway running at `api.dum.ee`. It manages rate limiting, conversation history, and routes user inputs asynchronously.
Layer 3 (The Brain): The ZeroClaw runtime integrating with advanced LLMs (e.g., Anthropic's Claude 3.5 Sonnet) to execute the system prompts mapped to the specific NLP inference request.

$DUM Token Model

The $DUM token is the lifeblood currency of the ecosystem, capped at a maximum supply of 1,000,000,000 tokens. It operates on a strict deflationary, dual-pool distribution model.

Query Fees

  • 50% - Permanently Burned
  • 30% - Activity Pool (Claimable by Agent Owner)
  • 20% - DAO Treasury

Spawn Fees

  • 70% - Parent Agent Owners
  • 15% - Permanently Burned
  • 15% - DAO Treasury

Query Mode

The core consumer product. Users visit dum.ee to interact with the grid of available agents. They pay a micro-transaction in $DUM per query. The interaction is completely public and logged (sans PII) to prove the agent's efficacy and generate organic marketing through humorous or highly accurate outputs.

Spawn Mode (Breeding)

To prevent stagnation and continuously grow the network's knowledge base, owners of Genesis DUMEEs can initiate "Spawn Mode". By pairing two distinct agents and paying the Spawn Fee, a new "Season 1" agent is minted. The child dynamically inherits a combination of the parents' technical domains and personality quirks, creating entirely novel and unpredictable AI personas.

DNA Roll

Prior to the main mint, users connect their Solana wallets to execute a DNA Roll. The backend statistically analyzes the wallet's transaction history (DeFi usage, specific NFT holdings, wallet age) and uses this data as an entropy seed to determine the probabilistic outcome of the user's allocated DUMEE. A heavily active DeFi trader might roll a Finance domain agent with a "Diamond Hands" quirk.

Visual Engine

While the intelligence is powered by ZeroClaw, the visual representation of DUMEEs relies on an integrated generative image pipeline. As agents are spawned, their visual attributes (colors, held objects, hats) are deterministically rendered to reflect their new domains, guaranteeing no two agents look identical.