An educational crypto, finance, and AI vocabulary puzzle about collateral, short-term cash instruments, agent coordination, and model serving.
Onchain Credit Risk
These terms cover collateral levels, account health, liquidation triggers, and borrowing limits.
- Collateral Ratio: The collateral ratio compares pledged collateral with the debt or exposure it supports.
- Health Factor: Health factor is a risk score that shows how far a position is from liquidation.
- Liquidation Threshold: The liquidation threshold is the level at which collateral is no longer enough to keep a position open.
- Borrow Cap: A borrow cap is the largest amount a protocol or policy allows someone to borrow.
Money-Market Instruments
These terms describe very short-dated instruments and pooled cash products used in treasury management.
Treasury BillRepurchase AgreementCommercial PaperMoney Market Fund
- Treasury Bill: A treasury bill is a short-maturity U.S. government security often treated as a cash-like instrument.
- Repurchase Agreement: A repurchase agreement is a short-term loan backed by securities and repaid at a later date.
- Commercial Paper: Commercial paper is unsecured short-term debt issued by companies to fund working capital.
- Money Market Fund: A money market fund invests in short-term debt and cash-like instruments.
Agent Workflow
These terms cover instructions, tool use, planning, and scratch-space for intermediate steps.
- Tool Calling: Tool calling lets an AI model request a function or external action during a run.
- System Prompt: A system prompt sets the highest-priority instructions that guide a model's behavior.
- Planner: A planner breaks a task into steps and chooses the next action.
- Scratchpad: A scratchpad is a temporary place for intermediate reasoning, notes, or state.
Serving Capacity
These terms cover queue management, smaller batches, temporary warm capacity, and routing decisions.
MicrobatchingQueueingWarm PoolLoad Balancing
- Microbatching: Microbatching groups a small number of requests together to improve hardware efficiency.
- Queueing: Queueing is the process of lining up requests until capacity is available.
- Warm Pool: A warm pool is reserved capacity kept ready so new requests can start quickly.
- Load Balancing: Load balancing distributes requests across available workers to keep utilization even.