A daily crypto, finance, and AI vocabulary grouping puzzle for beginners.
DeFi Collateral Controls
These concepts describe collateral mechanics in DeFi without recommending any lending strategy.
- Collateral Haircut: A collateral haircut reduces the value assigned to pledged assets when calculating borrowing capacity or risk exposure.
- LTV Cap: A loan-to-value cap sets the maximum amount that can be borrowed relative to the accepted value of collateral.
- Oracle Price Band: An oracle price band defines an acceptable range for external price inputs before a protocol flags or limits their use.
- Liquidation Threshold: A liquidation threshold is the risk level at which a lending position may be closed or partially sold to repay debt.
Stablecoin Liquidity Rails
These terms explain liquidity and settlement concepts around stablecoins at an educational level.
Mint Redeem WindowPool DepthPayment StablecoinSettlement Cutoff
- Mint Redeem Window: A mint redeem window is the period when eligible participants can create or redeem stablecoins with an issuer.
- Pool Depth: Liquidity pool depth describes how much asset supply is available in a pool before trades begin to move prices materially.
- Payment Stablecoin: A payment stablecoin is a token designed to support transfers or settlement while targeting a stable reference value.
- Settlement Cutoff: A settlement cutoff is the deadline after which a transfer, redemption, or payment may be processed in the next settlement cycle.
Agent Audit Trails
These ideas help describe how agent actions can be recorded, reviewed, and constrained.
Agent Action LogPolicy EvaluationHuman Review QueueCredential Scope
- Agent Action Log: An agent action log records the tools, decisions, and outputs produced by an AI agent during a workflow.
- Policy Evaluation: Policy evaluation checks whether a proposed agent action follows the rules, permissions, or risk limits defined for a system.
- Human Review Queue: A human review queue holds agent outputs or requests that require a person to approve, reject, or adjust them before completion.
- Credential Scope: Credential scope defines which resources, actions, or time periods an access key or identity can use.
AI Data Pipeline Ops
These terms connect AI data operations with quality, provenance, and model performance monitoring.
- Data Provenance: Data provenance describes where data came from, how it was collected, and what transformations it passed through.
- Feature Store: A feature store manages reusable data attributes that machine learning systems can use for training or inference.
- Label Quality: Label quality describes how accurate, consistent, and useful human or automated annotations are for model training and evaluation.
- Drift Monitoring: Drift monitoring tracks changes in input data, outputs, or model behavior that may reduce reliability over time.