A daily crypto, finance, and AI vocabulary puzzle covering settlement rails, borrowing safety, search systems, and model serving.
Stablecoin Mechanics
These words describe backing, issuance, redemption, and whether a stablecoin is drifting from its peg.
- Reserve: A reserve is the backing held against issued tokens or liabilities.
- Mint: Minting issues new tokens according to the protocol's rules.
- Redeem: Redeeming returns a token for the underlying asset or claim value.
- Depeg: A depeg happens when a stablecoin or similar asset moves away from its reference value.
Credit Risk
These words describe the safety margin that protects lenders and the forced closing that can happen when it disappears.
- Collateral: Collateral is an asset pledged to secure borrowing or another obligation.
- Health Factor: Health factor estimates how close a lending position is to liquidation.
- Liquidation Threshold: The liquidation threshold is the risk level at which a position may be closed or sold.
- Overcollateralization: Overcollateralization means the pledged assets are worth more than the loan or obligation they support.
Retrieval Systems
These words describe how text is represented, split, and ranked before a model answers.
EmbeddingChunkingRerankerVector Store
- Embedding: An embedding is a numerical representation of data that captures semantic meaning.
- Chunking: Chunking breaks long text into smaller pieces for indexing and retrieval.
- Reranker: A reranker scores retrieved items again to improve relevance ordering.
- Vector Store: A vector store keeps embeddings so similar items can be searched efficiently.
Inference Serving
These words describe the runtime tradeoffs that shape model responsiveness and serving efficiency.
- Latency: Latency is the delay between a request and the start or completion of a response.
- Batching: Batching processes multiple inputs together to improve hardware efficiency.
- Throughput: Throughput is the amount of work a system completes over a given time period.
- Quantization: Quantization reduces numeric precision to make model serving faster or cheaper.