An educational crypto, finance, and AI vocabulary puzzle about stablecoin design, derivatives signals, agent orchestration, and model serving efficiency.
Stablecoin Rails
These terms cover the target price, backing assets, cash-out path, and assets used to support issuance.
- Peg: A peg is a target value that a stablecoin is designed to track.
- Reserve: A reserve is an asset pool held to support a token's stability or redemption.
- Redemption: Redemption is the process of exchanging a stablecoin for the asset it tracks or backs.
- Collateral: Collateral is an asset posted or held to support the value of a financial promise or token.
Perp Market Signals
These terms describe recurring payments, position crowding, price reference points, and spread relationships.
- Funding Rate: The funding rate is a recurring payment that helps perpetual futures stay near spot prices.
- Open Interest: Open interest is the total number of derivative contracts that remain open.
- Basis: Basis is the price difference between a spot asset and a related futures contract.
- Mark Price: Mark price is the reference price exchanges use to value derivative positions and reduce manipulation risk.
Agent Tooling
These terms cover sequencing work, choosing actions, retaining context, and invoking functions.
- Planner: A planner is the component that decides the sequence of steps an agent should take.
- Router: A router chooses which model, tool, or path should handle a request.
- Memory: Memory is stored context that helps an agent recall prior information across steps or sessions.
- Tool Call: A tool call is a structured request for a model to use an external function or service.
Inference Speed
These terms describe grouping work, reusing results, response delay, and shrinking model weight precision.
- Batching: Batching combines multiple requests into one processing pass to improve throughput.
- Cache: A cache stores results so repeated work can be served faster later.
- Latency: Latency is the time it takes for a request to receive a response.
- Quantization: Quantization reduces numerical precision to make models smaller and faster to run.