A daily crypto, finance, and AI vocabulary puzzle about how stablecoins settle, credit is constrained, retrieval works, and model serving is measured.
Stablecoin Settlement
These terms describe how backing is tracked, how holders exit to cash-like value, and how disclosures support confidence.
Backing MixMint WindowRedemption QueueReserve Report
- Backing Mix: Backing mix is the blend of assets, cash equivalents, or other reserves used to support a stablecoin.
- Mint Window: A mint window is the period or rule set during which new tokens can be issued against eligible backing.
- Redemption Queue: A redemption queue is the order or delay holders may face when exchanging tokens for the underlying value.
- Reserve Report: A reserve report summarizes the assets and liabilities that support a token's backing claims.
DeFi Credit Rails
These terms describe how on-chain loans are limited, priced, and protected from liquidation.
Loan to ValueBorrow CeilingRisk BufferOracle Price
- Loan to Value: Loan to value compares the size of a loan with the value of the collateral securing it.
- Borrow Ceiling: A borrow ceiling is the upper limit on how much value a position can borrow from a protocol.
- Risk Buffer: A risk buffer is the margin that helps a position absorb market moves before liquidation risk rises.
- Oracle Price: An oracle price is the external market price feed a protocol uses to value collateral or assets.
Retrieval Stack
These terms describe how systems turn documents into searchable chunks, match meaning, and refine results.
Semantic SearchDocument ChunkCross EncoderQuery Expansion
- Semantic Search: Semantic search finds text by meaning or intent instead of matching only exact keywords.
- Document Chunk: A document chunk is a smaller passage split from a larger document so it can be embedded and retrieved.
- Cross Encoder: A cross encoder scores a query and document together to produce a more precise ranking signal.
- Query Expansion: Query expansion adds related words or concepts to a search query so retrieval can cover more relevant text.
AI Compute Economics
These terms describe request latency, token output, reuse of cached work, and the memory footprint of a model.
Inference SpeedToken RateCache Hit RateModel Footprint
- Inference Speed: Inference speed describes how quickly a model can produce an answer after receiving a request.
- Token Rate: Token rate measures how many output tokens a serving system can produce in a given amount of time.
- Cache Hit Rate: Cache hit rate is the share of requests that can reuse stored work instead of recomputing it.
- Model Footprint: Model footprint is the memory or storage size a model needs while being deployed or served.