A daily crypto, finance, and AI vocabulary puzzle covering payment rails, collateral safety, agent orchestration, and model serving.
Stablecoin Rails
These words describe how a stablecoin is backed, created, verified, and exchanged for its reference asset.
Reserve BackingMintRedeemAttestation
- Reserve Backing: Reserve backing is the pool of assets that supports a stablecoin's claimed value.
- Mint: Minting is the process of issuing new tokens according to a protocol or reserve rule.
- Redeem: To redeem is to return a token and receive its underlying value or collateral back.
- Attestation: An attestation is a signed statement that reserves, balances, or data have been verified.
Credit Health
These words describe how lending capacity, safety buffers, and forced sales are measured.
- Collateral Factor: Collateral factor is the share of posted collateral that can count toward borrowing capacity.
- Loan-to-Value: Loan-to-value compares the size of a loan with the value of the collateral securing it.
- Health Factor: Health factor is a risk score that shows how close a borrowing position is to liquidation.
- Liquidation Threshold: Liquidation threshold is the collateral level at which a position can be forcibly closed.
Agent Workflows
These words describe the control loop that helps an agent decide, remember, act, and stay within bounds.
- Tool Calling: Tool calling is the pattern where an AI system asks external functions or services to do work.
- Planner: A planner is the component that breaks a task into steps and chooses what to do next.
- Memory: Memory is stored context that helps an agent or model reuse relevant information later.
- Guardrail: A guardrail is a rule or check that limits unsafe, off-scope, or invalid agent behavior.
Inference Efficiency
These words describe the techniques that reduce latency, increase throughput, and save compute during serving.
- Batching: Batching combines multiple inference requests so they can run more efficiently together.
- KV Cache: A KV cache stores key and value tensors so a model can reuse attention history.
- Quantization: Quantization reduces numerical precision to make model serving faster and lighter.
- Speculative Decoding: Speculative decoding speeds generation by drafting tokens with a smaller model and verifying them with a larger one.