Augment-to-Earn Lifecycle: From Task to Reward

Step 1: Task Generation

When users interact with Cryptomate's extensions seeking answers, the system's Large Language Models (LLMs) generate initial responses. However, the process doesn't stop there. Cryptomate employs a sophisticated detection mechanism to identify potentially controversial or sensitive content within these AI-generated answers. When such content is detected, the system automatically creates an augment-to-earn task.

These tasks are then broadcast to the network of CMA (Cryptomate Augmenter) token stakeholders, also known as "Augmenters". This proactive approach ensures that sensitive or complex topics receive human oversight, maintaining the integrity and reliability of the information provided.

Cryptomate's augment-to-earn system handles two types of tasks:

Realtime Tasks: These are urgent tasks that require immediate attention. For these tasks, we expect active augmenters to respond quickly, typically within 4-10 seconds. These tasks are crucial for providing timely and accurate information to users in real-time scenarios.

Review Tasks: Some tasks do not require immediate augmenter input. These tasks involve reviewing previous LLM answers to enhance overall model quality. Review tasks have a much longer expiry time, allowing augmenters to conduct thorough analyses and provide detailed improvements. This type of task is essential for the long-term development and refinement of the system.

Step 2: Human Augmentation

Upon receiving an augment-to-earn task, Augmenters spring into action. These aren't just casual contributors, but rather a network of sophisticated investors and crypto experts. By staking CMA tokens, they've demonstrated not only their commitment to the network but also their deep understanding of the crypto ecosystem and financial markets.

When engaging with a task, these expert augmenters have two primary options for contributing:

• Select the Best Completion: Leveraging their extensive knowledge, Augmenters can review the existing candidate completions and select the one they believe is the most accurate, insightful, and valuable. This option allows them to apply their expertise in evaluating complex crypto-related information and choosing the response that best reflects current market realities and expert understanding.

• Input Their Own Completion: If an Augmenter believes they can provide a more nuanced, up-to-date, or strategically valuable response than the existing candidates, they can input their own completion. This new completion then becomes a candidate for other expert Augmenters to consider and potentially select. This option encourages the introduction of cutting-edge insights, expert analysis, and sophisticated perspectives that only seasoned crypto professionals can provide.

Step 3: Completion Selection

For all candidate completions submitted by LLM or Augmenters, only one will be ultimately presented to the user. The selection process works as follows:

1. Augmenters review and vote on the candidate completions.

2. Each Augmenter's vote is weighted based on their stake in the system. The stake is represented by native token CMA.

3. The completion with the highest weighted votes is selected as the final answer presented to the user.

While reputation doesn't directly influence the voting power, it plays a crucial role in the overall ecosystem by affecting reward distribution and user trust.

Step 4: Acceptance

The acceptance process in Cryptomate's augment-to-earn system is crucial for determining which completions are considered successful. There are three types of tasks, each with its own acceptance criteria:

1. Feedback Tasks:

   • Judged directly by the end-user through a simple thumb up/down rating system.

   • A completion is accepted when it receives positive or neutral user feedback.

   • These tasks prioritize user satisfaction and the practical value of the information provided.

   Example prompt: "Explain the concept of blockchain in simple terms for a beginner."

2. Prediction Tasks:

   • Evaluated based on how well the augmented completion aligns with future facts or outcomes.

   • A completion is accepted when the prediction is proven correct by subsequent events.

   • These tasks reward accuracy in forecasting and require deep domain knowledge and analytical skills.

   Example prompt: "What will be the price range of Bitcoin one month from today?"

3. Poll Tasks:

   • Assessed through consensus among Augmenters.

   • Acceptance is determined by the weighted votes of Augmenters, with voting power proportional to their stake and reputation.

   • The completion with the highest weighted votes is accepted.

   • These tasks leverage the collective wisdom of the Augmenter community to determine the best response.

   Example prompt: "What is the most promising emerging blockchain technology for scalability in 2024?"

For all types of tasks, only one completion is ultimately accepted and presented to the user. This ensures that users receive a single, high-quality response that has been vetted through the appropriate acceptance process.

Step 5: Reward Distribution

Once a completion has been accepted through one of the above processes, the system moves to the reward distribution phase. Rewards in Cryptomate's ecosystem come in two forms: reputation points and CMA tokens.

Reputation Rewards

• Augmenters who supported the accepted completion receive an increase in their reputation score.

• The amount of reputation gained may vary based on the task's difficulty or importance.

• Reputation can only increase; there is no penalty to reputation for supporting unaccepted completions.

• Reputation growth is gradual, reflecting consistent positive contributions over time.

• Reputation cannot be transferred between users, representing an individual's personal track record in the system.

Crypto Rewards

• The token reward for the task is split among the Augmenters who supported the accepted completion.

• The distribution of tokens is proportional to each Augmenter's stake in the task.

• For example, if Augmenter A staked 100 tokens and Augmenter B staked 200 tokens on the accepted completion, Augmenter B would receive twice the token reward of Augmenter A.

• These rewards are added to the Augmenter's stake and can be transferred or withdrawn as desired.

Interplay of Reputation and Stake

While reputation and stake-based rewards are distributed separately, they both play crucial roles in the ecosystem:

• Stake provides immediate financial incentives and determines voting power in the selection process.

• Reputation builds over time, reflecting an Augmenter's consistent performance and reliability.

• High reputation can lead to increased trust from other Augmenters and potentially more opportunities within the ecosystem, even if not directly influencing reward distribution.

This dual reward system encourages both short-term active participation (through stake) and long-term commitment to quality (through reputation), creating a balanced and sustainable ecosystem for Cryptomate's network of sophisticated crypto experts and investors.