A daily crypto, finance, and AI vocabulary puzzle covering leverage, execution quality, retrieval systems, and agent plumbing.
Collateral Risk
These terms describe how leverage is supported, monitored, and reduced when markets move.
- Collateral: Collateral is an asset pledged to back a loan, position, or other obligation.
- Leverage: Leverage increases exposure by using borrowed capital or synthetic financing.
- Liquidation: Liquidation is the forced closing of a position when collateral is no longer sufficient.
- Haircut: A haircut is a discount applied to collateral value to account for market risk or volatility.
Market Execution
These terms describe how much size a market can handle and what execution costs can look like.
- Liquidity: Liquidity is the ease of buying or selling an asset without moving the price too much.
- Order Flow: Order flow is the stream of buy and sell orders arriving in a market over time.
- Price Impact: Price impact is the amount a trade changes the market price while it is being executed.
- Spread: Spread is the difference between the best available buy and sell prices in a market.
AI Retrieval
These terms describe how systems encode information, search it, and package context before generation.
EmbeddingVector StoreRerankerContext Window
- Embedding: An embedding turns text, images, or other data into numbers that capture similarity.
- Vector Store: A vector store keeps embeddings so a system can search for related items quickly.
- Reranker: A reranker re-scores search results after retrieval to improve the final ordering.
- Context Window: A context window is the amount of text or tokens a model can consider at once.
Agent Workflows
These terms cover planning, tool use, and the boundaries that keep agents under control.
- Agent: An agent is an AI system that can plan, choose tools, and carry out steps toward a goal.
- Tool Calling: Tool calling is the process of asking an external tool or function to do a task for the model.
- Planner: A planner decides which steps an agent should take before or during task execution.
- Sandbox: A sandbox limits what an agent or program can do so experiments stay contained.