Crypto × AI Glossary

Key terms from crypto, AI and where they meet, explained briefly and plainly.

Crypto basics

Blockchain: A shared ledger where transactions are recorded in linked blocks that cannot be altered. No single company keeps the record; the network’s computers do it together.
Wallet: An app or device that manages your crypto. It actually stores the keys that control your coins, not the coins themselves.
Seed phrase: The 12 or 24 words that restore a wallet. Anyone who has them can take your funds; never give them to any site or person.
Stablecoin: A token whose value is usually pegged to 1 US dollar (e.g. USDT, USDC). Traders use it like cash.
CEX / DEX: A CEX is a centralized exchange run by a company. A DEX is a decentralized exchange run by smart contracts, where assets stay in the user’s wallet.
Layer 1 / Layer 2 (L1 / L2): An L1 is a base blockchain (Bitcoin, Ethereum, Solana). An L2 runs transactions faster and cheaper and settles the result on the L1 (e.g. Base, Arbitrum).
Gas: The fee paid to the network to process a transaction. It rises when the network is busy.
Staking: Locking tokens to help secure a network in exchange for rewards. Lock-ups and penalty risks can apply.
DeFi: Financial apps that run on smart contracts without intermediaries like banks: lending, swaps, yield.
TVL: "Total value locked": the dollar value of assets deposited in a protocol, used as a size indicator.
RWA: Real-world assets: stocks, bonds or commodities represented as tokens on a blockchain.

Airdrop: A project giving free tokens to its users. Who qualifies is usually unknown in advance; "claim your airdrop" messages are a common scam.
Points program: A project rewards activity (trading, liquidity, tasks) with points and says they will count if it launches a token. No token is guaranteed.
TGE: "Token generation event": when a project’s token is first issued and usually starts trading.
Perpetual futures: A derivative with no expiry that can be traded with leverage. Funding payments keep its price close to spot.
Funding rate: A periodic payment between longs and shorts on perpetuals; the crowded side pays.
Leverage and liquidation: Leverage opens a position several times the collateral. If price moves far enough against you, the position is force-closed (liquidated) and the collateral is lost.

Large language model (LLM): An AI model trained on huge amounts of text that can read and write language (e.g. Claude, GPT, Gemini).
Token (in AI): The small chunks a model splits text into, roughly 3-4 characters. Prices and limits are measured in tokens.
Context window: How much text, in tokens, a model can read and consider at once.
Reasoning model: A model that "thinks" step by step before answering. Better on hard problems, but can be slower and pricier.
AI agent: An AI that plans and carries out multi-step tasks on its own, using tools such as the web, code or apps.
Benchmark: A standard test used to compare models (e.g. GPQA, SWE-bench). A single test may not reflect real-world use.
Open-weight model: A model whose trained weights are published, so anyone can run and adapt it (e.g. DeepSeek, Qwen).
Parameters: The numeric values a model learns in training. Their count shows model size but does not alone decide quality.
Hallucination: When a model confidently produces false or made-up information. Important facts should be checked against sources.
RAG: Retrieval-augmented generation: the model first fetches relevant documents or web results and answers from them, improving freshness and accuracy.

On-chain AI agent: An AI agent with its own wallet that can transact on a blockchain: pay, swap and interact with contracts.
DePIN: Decentralized physical infrastructure networks, where people who share resources such as GPUs, storage or wireless coverage are rewarded with tokens.
Decentralized GPU market: Blockchain-based markets that rent out idle graphics-card compute for AI training and inference (e.g. Render, Akash).
Decentralized AI network: Systems where independent participants run models and are rewarded by the network for quality output (e.g. Bittensor and its subnets).
AI token: Crypto assets tied to AI projects (e.g. TAO, FET, RENDER). Prices often react strongly to AI news and are volatile.
Oracle: A service that brings outside data (prices, weather, AI outputs) onto a blockchain, connecting smart contracts to the real world.