A daily crypto, finance, and AI vocabulary grouping puzzle for beginners.
Rollup Settlement Ops
These concepts explain operational pieces of rollup settlement without endorsing any chain or token.
Rollup SequencerBridge FinalityData Availability SamplingFraud Proof Window
- Rollup Sequencer: A rollup sequencer orders layer-2 transactions before they are bundled and posted or proven against a base chain.
- Bridge Finality: Bridge finality is the point at which a cross-chain transfer is considered settled enough that the receiving side can rely on it.
- Data Availability Sampling: Data availability sampling is a method for checking that enough transaction data is published for others to verify a chain or rollup state.
- Fraud Proof Window: A fraud proof window is the period during which participants can challenge an incorrect rollup state update.
Stablecoin Duration Risk
These ideas describe how reserve assets, maturity timing, and rate changes can matter for stablecoin operations.
Duration GapT-Bill LadderReserve Transparency ReportRate Pass-Through
- Duration Gap: A duration gap describes a mismatch between how quickly stablecoin holders may redeem and how quickly reserve assets mature or can be sold.
- T-Bill Ladder: A Treasury bill ladder spreads reserve maturities across dates so cash becomes available at regular intervals.
- Reserve Transparency Report: A reserve transparency report summarizes information about assets, liabilities, or controls behind a stablecoin or similar cash-like product.
- Rate Pass-Through: Rate pass-through describes whether changes in interest earned on reserves are reflected in product economics, fees, or user-facing yield.
Agent Permission Design
These terms help describe safer AI agent workflows where access is scoped, recorded, and reviewable.
Model Context ProtocolTool Call AllowlistAgent Memory StoreApproval Checkpoint
- Model Context Protocol: Model Context Protocol is a standard way for AI applications to connect models with tools, data sources, and structured context.
- Tool Call Allowlist: A tool call allowlist defines which tools or actions an AI agent may use during a workflow.
- Agent Memory Store: An agent memory store keeps selected context that an AI agent can reuse across steps, sessions, or tasks under defined rules.
- Approval Checkpoint: An approval checkpoint is a required review step before an AI agent can take a sensitive action or send an instruction to another system.
Inference Capacity Economics
These concepts connect AI model serving with capacity planning, latency, and compute utilization.
GPU Cluster UtilizationInference BatchingServing LatencyCompute Reservation
- GPU Cluster Utilization: GPU cluster utilization measures how much available accelerator capacity is actively used for training, inference, or related workloads.
- Inference Batching: Inference batching groups model requests so hardware can process them more efficiently, often with a latency tradeoff.
- Serving Latency: Model serving latency is the time between a request reaching an AI serving system and the model response being returned.
- Compute Reservation: A compute reservation is an agreement or allocation that keeps infrastructure capacity available for expected future workloads.