Application Binary Interface described in JSON that defines a contract’s functions, events, and types so clients can encode calls and decode logs.

Design that lets smart contracts act as user accounts, enabling custom validation, sponsored fees, and better UX (e.g., EIP‑4337).

Ledger model where balances live in accounts and transactions update balances directly, contrasted with the UTXO model.

Counter stored with an account to prevent replay and enforce transaction ordering. Each new transaction increments the nonce.

Contract that locks all payments collected from SHELL sales. NACKL holders can burn tokens to redeem a proportional share of its contents.

A Layer 1 blockchain focused on practical UX and game‑ready throughput in the Acki Nacki ecosystem.

Nonlinear function like ReLU or GELU that lets neural networks approximate complex relationships.

Reinforcement learning architecture with a policy network (actor) and a value estimator (critic) that learn together.

Scam where attackers send tiny transfers from look‑alike addresses to trick users into copying the wrong address.

Privileged key that can upgrade or pause a protocol. Useful for safety, but adds centralization risk.

Input crafted to fool a model into misclassification or unsafe behavior without obvious changes to humans.

LLM‑driven systems that autonomously decompose tasks, call tools, and coordinate with other agents to reach goals.

Designing AI systems whose goals and behavior remain reliably beneficial to humans, even as capabilities scale.

Wallet kept completely offline, signing transactions without a live network connection to reduce attack surface.

Free distribution of tokens or NFTs to wallets, usually to bootstrap communities, reward early users, or promote a launch.

Strategically using a protocol to maximize eligibility for future token airdrops, often with many wallets.

Stablecoin that targets a peg using market incentives or supply rules instead of fully backed collateral. Risky during stress.

Permission that lets a contract spend a user’s tokens up to a set amount, approved via the ERC‑20 standard.

List of addresses granted early access or special mint/claim rights, often used in NFT launches.

Information edge that can generate outsized returns, also used for early product previews in software.

Any cryptocurrency that is not Bitcoin. Broad term that spans utility, governance, meme, and stable tokens.

On-chain trading model that uses liquidity pools and a pricing formula instead of traditional order books.

Techniques to spot unusual patterns in data, used for fraud detection, ops alerts, and model safety.

Number of indistinguishable participants in a privacy system. Larger sets usually mean stronger privacy.

API

Application Programming Interface, a contract for software to talk to other software, including blockchain RPCs.

Secret credential used to authenticate API requests. Treat like a password, rotate if exposed.

Profiting from price differences for the same asset across markets or pools, often with bots and MEV techniques.

An Ethereum Layer 2 rollup family focused on lower fees and higher throughput while inheriting L1 security.

Node that stores full historical state for all blocks, enabling queries at any past block height.

Specialized hardware optimized to compute a specific proof‑of‑work algorithm at very high efficiency.

Cryptography using public and private keys, enabling signatures and secure key exchange over open networks.

Cross‑chain trade that either completes for both parties or reverts for both, often using HTLCs.

Technique that lets models weight different parts of input sequences, central to transformers and modern LLMs.

Verifiable statement about an identity, event, or state. Used in staking, oracles, and decentralized identity.

Metric for binary classifiers representing area under the ROC curve. Higher is generally better.

Mechanism that periodically claims and reinvests rewards back into a position to increase APY.

Neural network trained to compress and reconstruct inputs, useful for denoising, anomaly detection, and latent features.

AutoML

Automation of ML tasks like feature engineering, model selection, and hyperparameter tuning.

AI system that can plan, act, and iterate toward goals with minimal human input, often chaining tools and memory.

Family of probabilistic consensus protocols based on repeated random subsampling and metastability.

Privacy‑focused Layer 2 leveraging zero‑knowledge proofs for shielded transactions and programmable privacy.