Specialized network for publishing and sampling data, used by rollups to reduce L1 costs.

Guarantee that transaction data is published for verification so that state transitions can be proven.

Graph structure without cycles used by some ledgers for parallel transaction confirmation.

Planned Ethereum upgrade to scale data availability by introducing many blob lanes with single proposer.

On‑chain governance structure where token or reputation holders propose and vote to manage resources.

A token-governed group coordinating via smart contracts and on-chain voting, with transparent rules and treasuries.

Application that runs on a blockchain or decentralized network using smart contracts.

An application that runs on a blockchain or similar network, using smart contracts for core logic.

Trading venue where orders are hidden until execution to reduce front‑running; hard to decentralize safely.

Technique where light nodes randomly sample encoded shares to verify that full data was published.

Evidence produced by data availability sampling that a block’s data is accessible with high probability.

Adversarial manipulation of training data to degrade or control model behavior.

Selecting, deduplicating, and filtering data to improve quality and reduce bias in training sets.

When the distribution of real‑world data differs from the training distribution, causing degraded model performance.

Investment method of buying a fixed amount at regular intervals regardless of price, reducing timing risk.

Automated strategy that splits orders into timed slices, often implemented via keeper networks or smart vaults.

Build process that produces identical binaries from the same source to enable reproducible verification.

Interactive verification process used by optimistic rollups to prove fraud via bisection and on‑chain resolution.

Attack that floods a service or network with traffic to overwhelm and degrade availability.

Mempool design that mitigates spam and flooding using fees, per‑peer limits, and reputation.

Distribution of control across many parties to reduce single points of failure and capture.

Surface‑level decentralization signals without real dispersion of power or risk, common in governance washing.

Surface in feature space that separates classes according to a model’s predictions.

Sequence‑to‑sequence architecture with bidirectional encoders and autoregressive decoders, used in translation.

Architecture used in LLMs that autoregressively predicts the next token with causal attention.

Combines neural networks with reinforcement learning for agents that act in environments to maximize reward.

Synthetic media generated by AI that swaps or fabricates faces, voices, or scenes.

Financial services built on smart contracts, including lending, exchanges, derivatives, and asset management.

Financial services built on public blockchains, open to anyone with a wallet, programmable and composable.

Design choice that discourages stake delegation through economic and security constraints, favoring direct participation.

Path notation like m/44’/60’/0’/0/0 that derives child keys deterministically from a master seed.

Contracts whose value derives from an underlying reference, such as perpetuals, futures, and options.

Finality model where once a block is finalized it cannot be reverted unless safety assumptions break.

Wallet where all addresses are derived from a single seed using a standardized path scheme.

On‑chain exchange using AMMs or order books where users keep custody of funds.

Peer‑to‑peer key‑value lookup system used for locating content in networks like IPFS and BitTorrent.

Globally unique ID controlled by the subject, resolved to a DID Document containing keys and endpoints.

JSON‑LD document describing public keys, verification methods, and service endpoints for a DID.

Specific scheme such as did:ethr or did:key that defines how DIDs are created and resolved.

Software that takes a DID and returns its DID Document using a method‑specific driver.

Privacy guarantee that limits the impact of any single record on aggregated outputs by adding calibrated noise.

Measure of how hard it is to find a valid PoW hash; adjusts to target block time.

Mechanism that exponentially increases mining difficulty to encourage protocol upgrades.

Generative model that learns to denoise data from noise through a reverse diffusion process.

Cryptographic proof that a message was authorized by the holder of a private key, e.g., ECDSA, EdDSA, BLS.

Reduction of ownership percentage as supply increases through issuance, vesting, or inflation.

Time during which parties can challenge a rollup batch or bridge message before finalization.

Technique where a small student model learns to mimic a larger teacher model’s outputs.

Protocol where multiple parties jointly generate a keypair without any single party knowing the full secret.

Email authentication method using domain‑signed headers; relevant for oracle feeds and governance emails.

Bitcoin smart contract approach where an oracle’s signature determines contract outcomes without revealing terms.

Umbrella term for distributed databases maintained by a network of participants, including blockchains.

Use of distinct prefixes or tags in hashing/signing to avoid cross‑protocol signature reuse or collision.

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.