An educational crypto, finance, and AI vocabulary puzzle about token plumbing, price feeds, agent controls, and model serving.
Stablecoin Settlement
These terms cover issuance, redemption, reserves, and the target price peg.
MintRedeemReservesPeg
- Mint: Minting is the act of creating new stablecoins or other tokens under the rules of a system.
- Redeem: Redemption is the process of exchanging a token for the asset or value that backs it.
- Reserves: Reserves are assets kept aside to support redemptions or maintain confidence in a token's value.
- Peg: A peg is the target price relationship a stablecoin tries to hold against a reference asset.
Oracle Safety
These terms describe how feeds are aggregated, checked, and paused during bad data.
OraclePrice FeedMedianizerCircuit Breaker - Oracle: An oracle supplies external information to a blockchain or smart contract.
- Price Feed: A price feed is a data stream that reports the current value of an asset or reference price.
- Medianizer: A medianizer combines several price inputs and uses the median to reduce the effect of outliers.
- Circuit Breaker: A circuit breaker pauses trading or execution when data looks wrong or volatile beyond a set limit.
Agent Permissions
These terms cover how an agent chooses actions, uses tools, and stays inside guardrails.
Function CallingToolSandboxApproval
- Function Calling: Function calling lets an AI model ask a system to run a specific tool or function.
- Tool: A tool is an external capability an agent can use to do something beyond text generation.
- Sandbox: A sandbox is a limited environment that restricts what a model or program can affect.
- Approval: Approval is a signoff step that allows an agent to continue with an action or request.
Serving Capacity
These terms describe the hardware and scheduling choices that limit or expand request volume.
GPUQueueingAutoscalingWarm Pool
- GPU: A GPU is specialized hardware that can process many math operations at once.
- Queueing: Queueing is the process of holding requests until resources are available to handle them.
- Autoscaling: Autoscaling adjusts the amount of compute available based on demand or load.
- Warm Pool: A warm pool is a set of prestarted workers kept ready to reduce startup delay.