A daily crypto, finance, and AI vocabulary puzzle about stablecoin plumbing, credit limits, agent workflows, and model serving tradeoffs.
Stablecoin Rails
These terms describe how a cash-like token is backed, redeemed, and kept near its target value.
- Backing Reserve: A backing reserve is the pool of cash, cash equivalents, or other assets used to support a token's value.
- Redemption Window: A redemption window is the period or rule set that determines when holders can swap tokens back for underlying value.
- Peg Band: A peg band is a narrow range around a reference price that a stable asset is expected to stay within.
- Issuer Attestation: An issuer attestation is a statement or report from the issuer describing reserves, liabilities, or backing status.
DeFi Credit Controls
These terms describe how borrow capacity is set and when positions become risky.
- Collateral Factor: A collateral factor is the portion of collateral value that a protocol allows a user to borrow against.
- Borrow Cap: A borrow cap is the upper limit on how much debt can be created within a protocol or market.
- Health Factor: A health factor measures how safely a borrow position is collateralized relative to its liquidation point.
- Liquidation Threshold: A liquidation threshold is the collateral level at which a position becomes eligible for forced closure.
AI Agent Workflow
These terms describe planning, tool use, memory, and limits that keep an agent on task.
- Tool Calling: Tool calling is the ability of an AI agent to request an external function, API, or tool to complete a task.
- Planner: A planner is the component or behavior that breaks a goal into steps and decides what to do next.
- Memory Window: A memory window is the amount of recent context an agent can keep available while it works.
- Guardrails: Guardrails are rules or checks that keep an AI agent within safe, useful, and approved behavior.
Compute Serving Metrics
These terms describe how systems batch requests, measure speed, and stay within capacity.
- Batch Size: Batch size is the number of requests or tokens processed together in one serving step.
- Inference Latency: Inference latency is the time between sending a prompt and receiving a model's output.
- Throughput: Throughput is the amount of model work a serving system can complete over a given period.
- Utilization Rate: Utilization rate is the portion of available compute capacity that is actively being used.