A daily crypto, finance, and AI vocabulary puzzle about issuance, liquidity, agent control flow, and inference throughput.
Stablecoin Operations
These terms describe how a stablecoin is backed, issued, and redeemed.
- Reserve Ledger: A reserve ledger records the assets that back outstanding tokens.
- Mint Authorization: Mint authorization is the permission to create new token supply.
- Redemption Window: A redemption window is the period during which holders can exchange tokens for backing assets.
- Attestation: An attestation is a signed statement that backing assets or balances match a reported amount.
Exchange Microstructure
These terms describe how order books, spreads, and price formation work in liquid markets.
Order Book DepthBid-Ask SpreadTick SizePrice Discovery
- Order Book Depth: Order book depth is the amount of buy or sell interest visible at different price levels.
- Bid-Ask Spread: The bid-ask spread is the difference between the highest bid and lowest ask in a market.
- Tick Size: Tick size is the smallest increment a market allows for quoted prices.
- Price Discovery: Price discovery is the process by which markets form a current price from trading activity and new information.
Agent Orchestration
These terms describe how an AI agent decides what to do and where to send each step.
- Planner: A planner breaks a task into steps and chooses what the agent should do next.
- Router: A router chooses which tool, model, or workflow should handle a request.
- Tool Calling: Tool calling lets an AI model trigger an external function or API.
- Scratchpad: A scratchpad is temporary working space the agent can use while reasoning or planning.
Model Serving
These terms describe common ways to cut latency and compute during model inference.
- Batching: Batching combines multiple requests so a model can process them more efficiently.
- Prefill: Prefill is the inference stage that processes the input prompt before generation starts.
- Prompt Cache: A prompt cache stores repeated prompt work so the system can reuse it on later requests.
- Quantization: Quantization reduces numerical precision so models use less memory and often run faster.