Aug 15, 2026

Onchain Settlement, Market Depth, Agent Memory, and Compute Economics

An educational crypto, finance, and AI vocabulary puzzle about blockchain settlement, market structure, agent context, and model efficiency.

Onchain Settlement

These terms cover completion, irreversibility, inclusion, and short-range chain history changes.

SettlementFinalityConfirmationReorg
  • Settlement: Settlement is the stage where assets or obligations are formally completed.
  • Finality: Finality means a blockchain state is treated as effectively permanent.
  • Confirmation: Confirmation is evidence that a transaction has been included in a block.
  • Reorg: A reorg happens when a blockchain replaces part of its recent history with a different branch.

Market Depth

These terms cover available size, the bid-ask gap, execution drift, and tradability.

  • Depth: Market depth is the amount of buy or sell interest available near the current price.
  • Spread: The spread is the difference between the best bid and best ask.
  • Slippage: Slippage is the difference between the expected trade price and the actual fill.
  • Liquidity: Liquidity is how easily an asset can be bought or sold without moving price much.

Agent Memory

These terms cover context limits, fetching references, external actions, and stored recall.

  • Context Window: The context window is the maximum amount of text a model can consider at once.
  • Retrieval: Retrieval fetches relevant documents or data so an AI system can use them.
  • Tool Calling: Tool calling lets an agent ask external functions or APIs to do work.
  • Memory: Memory is stored information an agent can reuse across turns or tasks.

Compute Economics

These terms cover response delay, work rate, grouped processing, and smaller-number representations.

  • Latency: Latency is the time between a request and a response.
  • Throughput: Throughput is how much output a system can produce over a given period.
  • Batching: Batching groups multiple inputs into one processing pass.
  • Quantization: Quantization compresses model numbers into lower precision to save memory and speed up inference.