The next web, explained in plain English
D
Attack that attempts to prevent legitimate users from accessing a service by exhausting resources.
Attempt to spend the same coins twice; prevented by consensus and confirmation depth.
Risk of linking on‑chain or public activity to a real‑world identity; mitigated with opsec and privacy tools.
Value‑based RL method that uses a neural network to approximate Q‑values for actions.
Regularization technique that randomly disables neurons during training to prevent overfitting.
E
Regularization technique that halts training when validation performance stops improving.
Signature scheme used by many chains; relies on the hardness of the elliptic‑curve discrete logarithm problem.
Deterministic signature scheme with fast verification and strong security, commonly using the Ed25519 curve.
Node deployed close to users for low‑latency access, content delivery, or lightweight verification.
Withdrawal path from a rollup or L2 to its base L1, often subject to a challenge or finality period.
Marketplace for Ethereum economic security where validators can opt‑in to secure additional services via restaking.
Specification process for Ethereum changes and standards; includes core protocol, networking, and ERC token specs.
Introduced chain IDs into transaction signing to prevent cross‑chain replay attacks.
Standard for human‑readable structured data signing with domain separation to reduce phishing.
Public‑key cryptosystem based on discrete logarithms; has homomorphic properties useful in voting.
Communication style for simplifying complex topics into clear, minimal language. Useful for docs and prompts.
Math structure used for efficient cryptography; examples include secp256k1, ed25519, BLS12‑381.
Dense vector representation of text, images, or audio capturing semantic relationships.
Model that converts inputs to vector embeddings used for search, clustering, or retrieval‑augmented generation.
Neural network component that maps inputs to latent representations; often bidirectional in Transformers.
Process of converting plaintext into ciphertext using ciphers and keys to ensure confidentiality.
Decentralized naming system mapping human‑readable .eth names to addresses and metadata.
Combining multiple models to improve accuracy and robustness, e.g., bagging, boosting.
Measure of uncertainty in a probability distribution; also used to describe randomness in cryptographic keys.
Pattern where data is encrypted with a data key, which is then encrypted with a master key for secure storage.
Account controlled by a private key, as opposed to a contract account controlled by code.
Short‑lived cryptographic key used for a single session or message to improve forward secrecy.
Discrete time window used for scheduling roles and computing rewards; includes individual and management epochs.
Seconds elapsed since 00:00:00 UTC on 1 January 1970; common timestamp in APIs and chains.
Token that represents ownership in an entity; may be regulated as a security depending on jurisdiction.
Consensus fault where a validator signs conflicting blocks or messages at the same height/round.
Multi‑token standard supporting both fungible and non‑fungible tokens in a single contract.
Fungible token standard with balances tracked per address, defining transfer and approval interfaces.
Standard enabling smart‑contract wallets with user operations, bundlers, and paymasters, without L1 protocol changes.
Non‑fungible token standard for unique assets, with safe transfers and metadata interfaces.
General‑purpose smart contract blockchain with an execution layer (EVM) and consensus layer (Beacon chain).
Unit measuring computational work in the EVM; users pay base fee plus tip per gas unit.
Lightweight JavaScript library for interacting with Ethereum and EVM chains, wallets, and contracts.
Popular Ethereum block explorer with verified contract source, analytics, and API.
Quantitative measure of model performance, such as accuracy, F1, BLEU, ROUGE, or perplexity.
Log emitted by a contract to index data off‑chain; not accessible to other contracts.
Continuous processing of on‑chain logs into queryable databases for analytics and app backends.
Deterministic runtime for smart contracts that defines bytecode, opcodes, and gas accounting.
Low‑level machine code executed by the EVM, produced by compiling Solidity, Vyper, or Yul.
Low‑level operation executed by the EVM, such as CALL, DELEGATECALL, SLOAD, and SSTORE.
Chain or runtime that supports Ethereum bytecode and tooling with minimal modifications.
The runtime that executes smart contracts on Ethereum and EVM-compatible chains.
Evaluation metric for QA tasks measuring the fraction of predictions that exactly match the ground truth.
Node software that processes transactions and state transitions; pairs with consensus clients.
Layer responsible for running smart contract code and maintaining state, formerly “Eth1”.
Entity that sequences and executes transactions for a rollup, possibly distinct from the prover or validator.
Mechanism allowing users to withdraw funds from a Plasma chain back to L1 using priority queues and challenges.
Techniques that make model predictions interpretable, e.g., SHAP, LIME, saliency maps.
Systems for augmenting LLMs with search, retrieval, or code execution to ground answers in fresh data.
F
Token distribution with no pre‑mine or private allocations; all supply released to the public transparently.
Property of the network where blocks become irreversible quickly under the probabilistic PoS design.
Service that dispenses small amounts of testnet tokens to developers for free.
Change in the statistical properties of input features over time, degrading model performance.
Process of transforming raw data into features that better expose patterns to learning algorithms.
System for managing, versioning, and serving features consistently for training and inference.
Algorithm that averages model updates from clients to produce a global model in federated learning.
Training across many devices or silos without centralizing raw data, aggregating model updates instead.
Protocol mechanism that destroys the base fee to counterbalance issuance and align incentives.
Ethereum fields maxFeePerGas and maxPriorityFeePerGas controlling total spend and miner/validator tip.
Mechanism by which users compete for blockspace through gas prices, tips, or auctions.
Model’s ability to generalize from a small number of examples per class or task.
Supplying a handful of labeled examples inside the prompt to steer an LLM’s behavior.
Encryption allowing computation on ciphertexts to produce encrypted results that decrypt correctly.
Computation designed to be efficient under homomorphic schemes, minimizing multiplicative depth and noise growth.
Government‑issued currency not backed by a physical commodity, e.g., USD, EUR.
Service that converts fiat currency into crypto assets using cards, bank transfers, or wallets.
Stablecoin collateralized by cash and equivalents, with redemptions at par value in fiat.
Decentralized storage network that incentivizes storing and retrieving data with cryptographic proofs.
Point at which a transaction becomes effectively irreversible under the consensus protocol.
Authorization model that restricts actions at the level of resources, fields, or operations.
Adapting a pre‑trained model on task‑specific data, often with smaller learning rates and adapters.
Mathematical field with a finite number of elements; used in ECC, BLS, and zkSNARK arithmetic.
Numeric representation with a fixed number of decimals, common in smart contracts to avoid floating‑point errors.
Uncollateralized on‑chain loan that must be repaid within a single transaction, else it reverts.
Mechanism allowing tokens to be withdrawn before payment in AMMs, enabling arbitrage within one transaction.
Research and tooling ecosystem around MEV mitigation and private transaction relays for Ethereum.
A protocol change that creates a new path of blocks, can be soft (backward compatible) or hard (not compatible).
Algorithm used by nodes to select the canonical chain among competing branches.
Chain that continues on an alternative rule set after a hard fork, forming a separate network and asset.
Mathematically proving that a program satisfies a specification, using model checking or theorem proving.
Large model trained on broad data that can be adapted to many downstream tasks.
Low‑precision number formats that speed up AI training and inference with minimal accuracy loss.
Proof showing that a claimed state transition is invalid; used in optimistic rollups during challenge periods.
Reordering or inserting transactions to profit from observed pending orders; a core MEV category.
User interface of a decentralized application, typically interacting with wallets and RPC providers.
Techniques like commit‑reveal, private mempools, or fair ordering to prevent MEV extraction.
Design goal of achieving rapid, energy‑efficient finality with minimal validator overhead.
Node that fully validates blocks and transactions and stores the complete blockchain state or history.
LLM feature where the model returns structured arguments to invoke external tools or APIs.
First 4 bytes of keccak(signature) identifying the function to call in EVM contract ABI.
Canonical name and argument types of a function, e.g., transfer(address,uint256), used to compute selectors.
Asset where each unit is interchangeable with any other unit, e.g., ERC‑20 tokens.
Automated testing that feeds random and adversarial inputs to expose bugs and invariants violations.
G
Cyclic groups on pairing‑friendly curves used in BLS signatures and zk systems.
Architecture with a generator and discriminator trained adversarially to synthesize realistic data.
Fee paid to execute transactions or smart contract operations, priced per unit of computation and storage.
Maximum gas a transaction or block is allowed to consume; prevents infinite loops and DoS.
Service estimating recommended gas fees based on pending mempool and recent blocks.
Cost per unit of gas, typically expressed in gwei; includes base fee and optional tip.
Per‑unit cost of gas; the base fee is burned under EIP‑1559 while tips incentivize inclusion.
Mechanism reducing effective gas cost when freeing storage or performing specific opcodes; heavily modified by recent EIPs.
Legacy pattern that tokenized gas refunds by creating and later destroying storage; mostly obsolete.
Transaction relayed by a third party or paymaster so the user does not pay gas directly.
Models that produce text, images, audio, or code, such as diffusion models and LLMs.
The first block of a blockchain network; defines initial state and parameters.
Popular Ethereum execution client written in Go.
Widely used smart‑contract wallet with multisig and modular account abstraction features.
Peer‑to‑peer message propagation where nodes randomly relay data to neighbors to reach network‑wide diffusion.
libp2p pub‑sub protocol integrating mesh networking and gossip for scalable message dissemination.
Processes by which a protocol or DAO makes and executes decisions, on‑chain or off‑chain.
Token that grants voting rights or proposal power in a protocol’s governance system.
Transformer‑based language model trained to predict next tokens; adaptable via prompting or fine‑tuning.
Parallel processor used for AI training and inference and, historically, for PoW mining.
Memory‑saving technique that recomputes activations during backward pass to fit larger models.
Technique to limit gradient magnitude during training to stabilize optimization.
Optimization method that iteratively updates parameters in the direction of negative gradient.
Funding mechanism that supports ecosystem builders with non‑dilutive capital or bounties.
Query language often used to fetch indexed on‑chain data from subgraphs and backends.
Strategy that picks the highest‑probability token at each step; fast but can be myopic.
Risk measures such as delta, gamma, vega, and theta used in crypto options.
Adversary incurs small cost to impose larger costs on others, degrading liveness or UX.
Efficient zkSNARK proving system requiring a trusted setup; used in privacy and rollup proofs.
Authoritative labels or facts used for training and evaluating models.
Linking model outputs to verifiable data or tools to reduce hallucinations.
Recurrent neural network cell with update and reset gates; simpler than LSTM.
Denomination equal to 10^9 wei; commonly used for gas prices.
H
Model output that is fluent but factually incorrect or fabricated; mitigated by grounding and verification.
Scheduled reduction of block subsidy in PoW chains, decreasing new issuance rate.
Maximum token supply or fundraise amount that will ever be issued or raised.
Incompatible protocol upgrade that requires nodes to update; may produce a split chain.
Physical device that stores private keys in a secure element and signs transactions offline.
Fixed‑length digest computed from data by a hash function; used for addressing and integrity.
Algorithm mapping arbitrary‑length input to fixed‑length output with properties like preimage resistance.
An input whose hash equals a given digest; revealing it unlocks HTLCs and commit‑reveal schemes.
Number of hash computations per second performed by miners; proxy for network security in PoW.
Contract that locks funds with a hash and a timeout to enable atomic swaps or payment channels.
Hierarchical deterministic wallet deriving many keys from a single seed using paths like m/44’/60′.
Matrix of second‑order partial derivatives; used in optimization analysis and curvature estimates.
Graph‑based index structure for fast vector similarity search in embedding databases.
Community term meaning to hold assets through volatility instead of trading.
Encryption that allows limited computation on ciphertexts; fully homomorphic variants support arbitrary circuits.
Security assumption that most stake, hashpower, or validators follow the protocol.
Wallet connected to the internet for convenience; higher attack surface than cold storage.
Entropy coding used by some layer‑2 systems to compress transaction data.
Workflow where humans label, curate, or approve model outputs to improve quality.
Open‑source umbrella for enterprise blockchain frameworks such as Fabric and Besu.
Configuration parameters of a model or optimizer, e.g., learning rate, batch size, depth.
Search for the best hyperparameters using grid search, Bayesian optimization, or bandits.
I
Protocol for trust‑minimized messaging and token transfers between sovereign chains.
Proof‑of‑authority consensus with Byzantine fault tolerance, used in permissioned EVM chains.
Fundraising event in which a project sells newly issued tokens to the public.
Token sale conducted via decentralized exchanges or launchpads using liquidity pools.
Token sale conducted on a centralized exchange, which manages distribution and listings.
Generative model family that denoises random noise into images through learned reverse processes.
Property that data, once written to the ledger, cannot be altered without consensus violation.
Relative loss incurred by LPs in AMMs when token prices diverge from deposit ratios.
LLM ability to learn patterns from examples provided at inference time without weight updates.
Design principle ensuring participants’ best interests align with protocol rules.
Service or node that processes on‑chain data into queryable databases or subgraphs.
Running a trained model to produce outputs; optimized via quantization, batching, and caching.
Stored key/value attention states used to accelerate autoregressive generation.
Service that hosts models behind APIs with scheduling, KV cache, and autoscaling.
Increase in token supply over time, often rewarding validators or stakers.
Predetermined rules that govern how token supply grows or decays over time.
Succinct proof technique used in Bulletproofs to verify commitments efficiently.
Immutable byte array attached to a transaction that encodes function selector and arguments.
Behavior of models to adhere to natural‑language instructions reliably.
Fine‑tuning LLMs on task instructions and demonstrations to improve following ability.
Portion of context reserved for system or task instructions in prompt engineering.
Compiler form between source and bytecode, e.g., Yul IR in Solidity toolchains.
Ability for independent systems or chains to exchange data and value safely.
Content‑addressed, peer‑to‑peer storage network using CIDs and IPLD data structures.
Data model for content‑addressed objects across distributed systems; foundation for CIDs.
Mutable naming layer for IPFS that maps identities to updatable content addresses.
Decentralized Identifier standard for self‑sovereign identity with verifiable credentials.
Creation of new tokens according to monetary policy; may be inflationary or fixed.
J
Protocol that lets two parties establish a shared key using only a low entropy password while resisting offline dictionary attacks, based on zero knowledge proofs.
Metric for set overlap defined as intersection over union; used for clustering, deduplication, and evaluating retrieval results.
Matrix of first order partial derivatives of a vector valued function; central to backpropagation, sensitivity analysis, and normalizing flows.
Denial of service tactic where an attacker holds HTLC slots or routes with low probability payments to block capacity, mitigated with reputation and fees.
Client library used in web apps to interact with blockchain nodes, wallets, or AI APIs from the browser or Node.js, for example ethers, web3, viem, or official vendor SDKs.
High performance numerical computing library with composable transformations like JIT compilation, vectorization, and automatic differentiation, widely used in research and large model training.
Sparse Merkle tree variant optimized for key value state with efficient updates and proofs, used in Diem and Aptos style blockchains.
Symmetrized and smoothed version of KL divergence bounded between 0 and 1; used to compare probability distributions in GANs and topic models.
Lightning wallet pattern where a channel is opened on demand to receive a payment, shifting on chain cost to the moment of first use and improving UX for new users.
Just in time compilation that generates optimized machine code at runtime from high level graphs, accelerating ML workloads and smart contract VMs.
Ecosystem around Solana block production that introduces MEV aware clients, bundles, and auctions to reduce spam and share tips with validators and stakers.
Variation in network latency over time; affects block propagation, mempool synchronization, and user perceived responsiveness in dapps and wallets.
Bitcoin CoinJoin marketplace that coordinates collaborative transactions between makers and takers to improve privacy while paying liquidity providers.
Transaction component that proves shielded inputs and outputs balance without revealing addresses or amounts, enabling private transfers in Zcash.
Representation space learned for multiple modalities or objects so semantically related items lie close together, used in retrieval and cross modal search.
Probability model over multiple random variables, capturing their dependence structure, foundation for graphical models and Bayesian inference.
Training a model on several related tasks at once to share representations and improve generalization, balancing losses with schedulers or task weighting.