AI and Web3 glossary
41 terms explained in plain English, each linked to where it matters in real projects.
A2A (Agent2Agent)
An open protocol that lets AI agents from different vendors discover each other and exchange tasks.
Related: Multi-Agent Systems
Account abstraction
Smart-contract wallets that allow gas sponsorship, batching and easier recovery.
Related: dApp Development
Agentic AI
Systems where AI agents carry out multi-step work across a process with little supervision.
Related: Agentic AI Development
Agentic RAG
Retrieval where an agent plans several searches, checks gaps and searches again before answering.
Related: RAG Development
AI agent
Software that uses a language model to plan steps, call tools and complete a task, within limits and with human checkpoints.
Related: AI Agent Development
Audit (smart contract)
An independent security review of contract code before it holds funds.
Related: Smart Contract Audit
Chunking
Splitting documents into smaller passages for search in a RAG system.
Related: RAG Development
Computer vision
AI that understands images and video, for example to detect defects or read labels.
Related: Computer Vision
Context window
The amount of text a language model can consider at once, measured in tokens.
Related: LLM Development
dApp
A decentralised application whose core logic runs on smart contracts.
Related: dApp Development
DeFi
Decentralised finance: lending, trading and yield through smart contracts.
Related: DeFi Development
Distillation
Training a small model to copy a larger model’s outputs on a task, to cut cost.
Related: LLM Fine-Tuning
Embedding
A list of numbers that represents the meaning of text or images, used for semantic search.
Related: RAG Development
ERC-3643
A token standard for permissioned securities with identity checks on transfer.
Related: RWA Tokenization
Evaluation set
A fixed set of tasks with expected answers used to measure AI quality before each release.
Related: AI Governance
Fine-tuning
Training an existing model further on your examples to change its behaviour or format.
Related: LLM Fine-Tuning
Gas
The fee paid to process a transaction on a blockchain.
Related: Ethereum Development
Guardrails
Rules and filters around a model that block unsafe inputs, outputs or actions.
Related: AI Security
Hallucination
When a model states something false with confidence.
Related: RAG Development
Human-in-the-loop
A design where people review or approve AI outputs or actions at set points.
Related: AI Agent Development
Layer 2 (L2)
A network that processes transactions off the main chain and settles to it for lower fees.
Related: Ethereum Development
LLM
Large language model: a model trained on large amounts of text to understand and generate language.
Related: LLM Development
LoRA
Low-Rank Adaptation, a cheap way to fine-tune large models by training small add-on layers.
Related: LLM Fine-Tuning
MCP (Model Context Protocol)
An open standard for connecting AI applications to tools and data through servers.
Related: MCP Server Development
MLOps
Practices for deploying, monitoring and retraining machine learning models reliably.
Related: MLOps
Model routing
Sending each request to the cheapest model that can handle it well.
Related: LLM Development
MPC wallet
A wallet whose key is split across parties so no single one can sign alone.
Related: MPC Wallet
Multisig
A wallet that needs several signatures to approve a transaction.
Related: Crypto Wallet
Oracle
A service that brings outside data, such as prices, onto a blockchain.
Related: DeFi Development
Prompt injection
An attack where hidden instructions trick a model into ignoring its rules.
Related: AI Security
Proof of reserves
Evidence that a token issuer holds the assets backing its tokens.
Related: Gold Tokenization
RAG
Retrieval-augmented generation: search your documents, then answer using what was found.
Related: RAG Development
Re-ranking
A second step that reorders search results so the most relevant passages come first.
Related: RAG Development
Rollup
A type of L2 that bundles transactions and posts them to a main chain.
Related: Cosmos & Substrate
RWA
Real-world asset: an off-chain asset such as property or bonds represented by tokens.
Related: RWA Tokenization
Smart contract
A program on a blockchain that runs automatically when conditions are met.
Related: Smart Contract Development
Stablecoin
A token designed to keep a steady value, usually one unit of a fiat currency.
Related: Stablecoin Development
Token (AI)
A small piece of text used to measure model input and output; billing is per token.
Related: LLM API cost calculator
Tokenization
Representing ownership of an asset as a token on a blockchain.
Related: RWA Tokenization
Travel Rule
A rule requiring sender and receiver details to travel with crypto transfers above set amounts.
Related: Stablecoin Payments
Vector database
A database built to store embeddings and find similar items quickly.
Related: RAG Development
Tell us what you want to build
A senior engineer reads every request and replies within one business day with questions, a rough range, or a call link.
- NDA on request before any details
- You own all code, models and IP
- US, UK and EU time-zone overlap