Mechanism tied to SHELL whereby users commit value to obtain computation and storage on the network.

Universal updatable zk SNARK protocol that supports arbitrary circuits with succinct proofs.

Fast zk proving system combining PLONK like constraints with FRI, optimized for recursion.

NFT badge that proves someone attended an event, often used for community engagement.

Scheme that lets a prover commit to a polynomial and later open evaluations with small proofs.

SNARK friendly hash function optimized for arithmetic circuits.

Method to inject token order information into transformer models.

Updated distribution over parameters after observing data, combines prior and likelihood.

Reinforcement learning algorithm that stabilizes updates with clipped objectives, used in RLHF.

Cryptographic schemes believed to be secure against quantum adversaries, such as lattices.

Built in contract address that exposes native cryptographic or system functions efficiently.

Hash property that makes it infeasible to find an input that maps to a given digest.

Data feed that publishes asset prices on chain for lending, liquidations, and derivatives.

Bidding war where bots raise gas fees to get transactions mined first, a source of MEV.

A secret number that controls a blockchain address, used to sign transactions. Must be kept offline and secure.

Pseudorandom number generator used for randomness in protocols and simulations.

Micropayment scheme where a sender issues lottery tickets with expected value equal to the price.

Consensus approach where validator selections and confirmations are probabilistic, enabling fast finality and parallelization on Acki Nacki.

Smart contract controlled privacy where viewing and spending rights are enforced by cryptography.

Crafting inputs and context to steer LLM outputs toward desired behavior and quality.

Attack where crafted text causes a model or agent to ignore instructions and exfiltrate data or perform actions.

PEFT method that learns a small set of continuous prompt vectors to condition a frozen model.

Consensus where a fixed set of approved validators produces blocks, trading openness for speed.

Consensus or allocation method where coins are provably destroyed to gain rights or bootstrap a chain.

Verifiable delay based sequence used by Solana to order events without relying on synchronized clocks.

Schemes that try to prove a unique human user without central authorities.

Proof that a storage provider has stored a unique encoding of data, not deduplicated copies.

Attestation that a custodian holds sufficient assets to cover liabilities, often with Merkle proofs.

Sybil resistance where miners commit disk space rather than compute power.

Proof that storage was provided over a period, used by decentralized storage networks.

Consensus where validators lock tokens to propose and attest blocks, earning rewards and risking slashing.

Consensus where miners expend energy to solve puzzles, securing the network and earning block rewards.

Liquidity controlled by a protocol treasury rather than rented from external LPs.

Upgradeable contract design where a proxy delegates calls to an implementation contract.

Cryptographic technique where a proxy can transform ciphertext from one key to another without learning the plaintext.

Removing parameters or neurons from a model or old states from a node to reduce size.

The cryptographic pair of a private key, shared publicly to derive addresses and verify signatures.

Publish subscribe protocol used to spread messages across a P2P network efficiently.

Payment pattern where recipients withdraw owed funds to avoid reentrancy and simplify refunds.

Payment pattern where the sender transfers funds directly to the recipient during execution.

Lightweight transformer module that learns query tokens to extract useful features from vision encoders for multimodal models.

Mapping from state‑action pairs to expected returns, often approximated with neural networks.

Model‑free reinforcement learning algorithm that learns action‑value functions to maximize cumulative reward.

Parameter‑efficient fine tuning approach that combines low‑rank adaptation with quantized base weights to cut VRAM use.

Tabular representation of Q‑values for discrete state and action spaces.

Expected return of taking an action in a state under a policy, central to Q‑learning and DQN.

Quantum Approximate Optimization Algorithm, hybrid method for solving combinatorial optimization problems.

Wallet UX pattern that encodes addresses or requests as QR codes for easy scan‑to‑pay on mobile.

Blockchain project using XMSS hash‑based signatures to provide quantum‑resistant security.

Lattice‑based digital signature scheme proposed for post‑quantum use, not selected in NIST’s final standardization.

Public goods funding mechanism where many small donations receive higher matching than a few large ones, encouraging broad support.

Voting mechanism where the cost of additional votes grows quadratically, balancing intensity and fairness.

Compressing model weights and activations to lower precision, for example 8‑bit or 4‑bit, to reduce memory and speed up inference.

Training procedure that simulates low‑precision arithmetic during training to preserve accuracy after quantization.

When a quantum algorithm or device performs a task faster or cheaper than classical methods.

Heuristic quantum technique for finding low‑energy solutions to optimization problems, typically via annealers.

Technique for establishing shared keys with security based on quantum physics, eavesdropping is detectable.

Entropy source that uses quantum phenomena to produce truly random bits, useful for cryptography.

Controversial milestone where a quantum device performs a specific task infeasible for any classical computer.

Also called post‑quantum, schemes designed to resist quantum attacks, for example lattice‑based signatures and KEMs.

High probability that a block will not be reverted after several confirmations, typical for Nakamoto consensus.

Optimization formulation targeted by quantum annealers and used to encode combinatorial problems.

Throughput metric for APIs, indexers, and RPC endpoints measuring how many queries are handled per second.

Execution strategy chosen by a database or indexer to answer a query efficiently.

Transformer mechanism computing attention weights from queries and keys to mix values across tokens.

UDP‑based transport used in HTTP/3 and libp2p that offers low latency, multiplexing, and better congestion control.

Minimum number of nodes that must agree to commit an operation, used in BFT and consensus protocols.

Aggregate signature or proof that a quorum of validators voted for a block, used in HotStuff‑style BFT.

Configuration of slices and thresholds that define which validator combinations form a quorum in FBAS.

Subset of validators sufficient to convince a node of agreement in federated Byzantine agreement systems like Stellar.

Platforms that provide turnkey tooling and infrastructure to deploy and operate custom rollups, similar to cloud for L2s.

Technique where a model retrieves relevant external documents at query time and conditions generation on them to improve accuracy and reduce hallucinations.

On‑chain randomness mechanism where participants commit to and reveal random values, combined to produce an unpredictable beacon for consensus or leader selection.

Controlling the number of requests a client can make in a period, protecting RPCs or APIs from abuse and helping ensure fair access.

Bitcoin mempool policy allowing an unconfirmed transaction to be replaced with a version that pays a higher fee, improving confirmation reliability under congestion.

Token that programmatically expands or contracts user balances to target a price or supply, keeping proportional ownership the same.

Systematic probing of an AI model with adversarial inputs to discover safety, security, and alignment weaknesses before release.

Smart contract vulnerability where an external call re‑enters the calling contract before state updates finalize, enabling double‑spend style exploits without proper guards.

Movement using web3 rails to fund public goods and ecological projects with transparent on‑chain incentives and impact tracking.

Practice where custodians or lenders reuse client assets as collateral for their own borrowing, a risk factor in centralized crypto finance if not transparently disclosed.

Learning paradigm where an agent interacts with an environment and optimizes actions to maximize cumulative reward over time.

In multi‑chain designs like Polkadot, the central chain that coordinates consensus and communication among connected parachains.

Service that submits transactions or messages on behalf of users or protocols, often paying gas and being reimbursed via meta‑transactions or fees.

Property of a network or application to tolerate short‑range chain reorganizations without loss or double‑spend, often via confirmations or finality gadgets.

When a node switches to a longer or heavier chain tip, replacing recent blocks. Causes temporary rollbacks of unfinalized transactions.

Attack where a valid transaction is captured and resent on the same or another chain to repeat effects, mitigated by chain IDs, nonces, and domain separation.

Mechanisms like chain IDs, EIP‑155 signing domains, and unique nonces that prevent transactions from being valid across different networks or contexts.

Methods that learn useful latent features from raw data, enabling downstream tasks with minimal task‑specific engineering.

Principles and practices for building and deploying AI systems that are safe, fair, transparent, privacy‑preserving, and accountable.

Reusing staked assets or their yield to secure additional services or networks, extending economic security across multiple protocols (e.g., via middleware frameworks).

Selecting the most relevant documents, chunks, or vectors from a corpus given a user query, typically via embeddings and similarity search.

Data structure that accelerates similarity search, for example HNSW graphs or IVF lists over embedding vectors.

Specification that maps states or actions to scalar rewards in reinforcement learning, defining the agent’s objective, often the hardest part to get right.

Model trained to score outputs according to preferences or policies, used to guide reinforcement learning or re‑ranking of generations.

Network rewards are split among Block Keepers, Mobile Verifiers, and Block Managers. Rewards are precomputed per epoch and paid at epoch end.

Privacy‑preserving signature scheme where a signer is indistinguishable within a group, used in some cryptocurrencies to hide the true signer.

Privacy technique combining ring signatures with confidential amounts to hide sender, receiver, and value in transactions (popularized by Monero).

Training approach where feedback labels are generated by AI systems instead of humans, often used to scale preference data for alignment.

Alignment method where a reward model is trained on human preference comparisons, then the base model is fine‑tuned with reinforcement learning to follow desired behavior.

Reverting a chain’s state to a prior block height. Generally avoided in public chains except during reorgs or critical incidents with community consensus.

A Layer 2 that executes transactions off‑chain and posts data or proofs to Layer 1 for security.

A Layer 2 that executes transactions off‑chain and posts data or proofs to a Layer 1 for security, commonly optimistic or zero‑knowledge designs.

Contract and off‑chain infra that allows deposits to and withdrawals from a rollup, often using canonical bridges with exit delays or validity proofs.

Component that generates validity proofs (e.g., zk‑SNARKs) attesting to correct state transitions in a rollup, verified on Layer 1.

Actor that orders transactions within a rollup and submits batches to Layer 1; can be centralized, decentralized, or shared among rollups.

How a rollup finalizes on its base chain, including data availability, proof verification, and the economic guarantees for bridging assets.

Top hash of a Merkle tree summarizing all leaves; enables compact proofs of inclusion for transactions or state.

Minimal set of entities or keys implicitly trusted by a system’s security model, such as genesis validators or hardware secure modules.

Endpoint that lets applications read chain data and send transactions to nodes, usually via JSON‑RPC over HTTP or WebSockets.

Public‑key cryptosystem based on the difficulty of factoring large integers; used for signatures, encryption, and randomness beacons in some protocols.

Systems programming language with strong safety guarantees and performance, widely used for blockchain clients, provers, and smart contracts on some chains.

Smart‑contract wallet enabling N‑of‑M approvals for transactions, widely used by DAOs and teams for treasury and permission management.

MEV exploit where an attacker places a buy before and a sell after a victim’s trade to profit from induced price movement and slippage.

Empirical regularities showing how model loss improves predictably with more parameters, data, and compute, guiding training budgets and model sizes.

Signature scheme enabling key aggregation and smaller proofs, used in Bitcoin Taproot and multi‑sig constructions to improve privacy and efficiency.

Actor that scans mempools and order flow to construct profitable transaction bundles, exploiting arbitrage, liquidations, or sandwich opportunities.

Elliptic curve used by Bitcoin and many EVM chains for public‑key cryptography, enabling compact keys and signatures with strong security.

Holding and controlling one’s own private keys rather than relying on a custodian; enables permissionless access but requires strong key hygiene.

Learning signals are derived from the data itself (e.g., next‑token prediction), enabling large‑scale pretraining without manual labels.

Training that mixes a small labeled set with a large unlabeled set, leveraging pseudo‑labels or consistency losses for better generalization.

Ephemeral keys with restricted permissions used for a limited time, enabling safer automated actions without exposing the main wallet key.

Fine‑tuning a pretrained model on instruction/response pairs to teach task formats and improve helpfulness before alignment steps like RLHF.

Splitting state or data across multiple partitions so different validators process different subsets, increasing parallelism and throughput.

Mechanism where SHELL can be bought at the System Pool price set by Block Keepers, or sold back for equal or less, anchoring price to real‑world costs.

Usage token for paying compute and storage. Sold from a System Pool for accepted currencies; unlimited supply with revenues sent to the Accumulator.

Rate set by Block Keepers per currency pair in the System Pool, acting as a collective signal for current conversion price.

Independent chain bridged to a mainnet, with its own validators and security model, offering different performance or features than the base chain.

Penalties applied to staked validators for misbehavior or downtime, burning or confiscating collateral to enforce protocol rules and security.

Difference between expected and executed price due to liquidity depth and price impact, controllable via slippage tolerance in swaps.

User‑set threshold for maximum acceptable price movement during a swap; trades revert if final price exceeds this bound.

Programmable code deployed to a blockchain that runs deterministically, enforcing rules without intermediaries.

Deterministic code deployed on a blockchain that enforces rules automatically when conditions are met, removing the need for intermediaries.

Off‑chain vote‑signing system that tallies token balances at a specific block height, enabling gasless on‑chain governance signaling.

Backward‑compatible protocol change that tightens rules so upgraded nodes accept a subset of previously valid blocks while legacy nodes remain functional.

Smart contract programming language for the EVM, with static typing, ABI compatibility, and a large ecosystem of tooling and libraries.

Non‑transferable token meant to represent credentials, reputation, or identity attributes bound to a wallet rather than transferable value.

Rollup that uses a base chain for data availability but validates and finalizes via its own consensus or light‑client rules rather than a smart contract on L1.

Attention mechanisms that compute over a subset of tokens or use structured patterns to reduce quadratic cost and extend context windows.

Decoding method where a small draft model proposes tokens and a larger model validates them, increasing generation throughput with minimal quality loss.

Latent diffusion model family for image generation that denoises from learned latent space, enabling high‑quality outputs on consumer GPUs.

Token designed to hold a steady value, typically pegged to fiat, collateralized on‑chain or off‑chain.

Locking tokens to participate in consensus or delegate to validators, earning rewards while contributing to network security on proof‑of‑stake chains.

Process of locking NACKL to participate in network roles. There is no fixed staking interest rate; rewards derive from network participation.

Token representing a claim on staked assets and rewards, providing liquidity while the underlying stake remains locked (e.g., liquid staking).

Off‑chain interaction where parties exchange signed updates and only settle the final state on‑chain, reducing fees and latency for repeated transactions.

Cryptographic proof (e.g., Merkle or proof‑of‑inclusion) that a given account or storage slot had a specific value in a block, verifiable by light clients or bridges.

Merkle root summarizing the entire chain state at a block, enabling light clients to verify account or storage proofs without full data.

EVM call type that prohibits state modification, used for read‑only contract queries to guarantee no writes during execution.

Technique that re‑renders content in the visual or textual style of a reference, via optimization or model conditioning techniques.

Framework for building blockchains in Rust with pluggable consensus, runtime pallets, and tooling, powering Polkadot parachains and standalone chains.

Training models on labeled examples (input, desired output) to learn mappings, common for classification and regression tasks.

On‑chain pool where SHELL is minted and sold for currencies chosen by Block Keepers; also the venue for selling unused SHELL back.

Hidden priming instructions that steer an assistant’s general behavior and safety policy, applied ahead of user messages.

Bitcoin upgrade adding Schnorr signatures and script path privacy via Tapscript and MAST, enabling more efficient and flexible spending conditions.

Script version introduced with Taproot that modernizes Bitcoin’s scripting with Schnorr support and Merkleized script paths for efficiency and privacy.

Training technique where the model is fed ground‑truth tokens for the next step instead of its own predictions, speeding convergence but risking exposure bias.

Threshold variant of ECDSA where multiple participants jointly create signatures compatible with standard ECDSA verification, used in bridges and MPC wallets.

Decoding parameter that sharpens or flattens the probability distribution over next tokens; lower values are more deterministic, higher values more diverse.

Byzantine fault tolerant consensus engine with fast finality used in Cosmos‑style chains; validator set votes in rounds to commit blocks deterministically.

Technique that applies multiple transformations at inference and aggregates outputs to improve robustness, common in vision and sometimes LLM prompts.

Public or private blockchain for experimentation with free tokens and relaxed economics before deploying to mainnet; often has faucets and faster reset cycles.

Service that distributes free test tokens to developers and users for deploying contracts and trying apps in non‑production networks.

Milestone when a new token is created and becomes transferable or claimable; distribution may follow a vesting schedule and regulatory constraints.

Decentralized indexing protocol for blockchain data, using subgraphs and query marketplaces.

Scheme where decryption requires cooperation of at least t of n key shares, enabling secure committees and time‑locked reveals in protocols.

Cryptographic method where a signature is produced jointly by multiple parties without reconstructing the private key, improving custody and MPC wallets security.

Rate at which a blockchain processes transactions; constrained by block size, latency, verification, and data availability bandwidth.

Mechanism that delays execution of privileged actions for a set period, giving token holders or users time to review and react to governance changes.

Recorded time associated with a block or transaction; may differ slightly from wall‑clock time and is bounded by protocol rules for validity.

Smallest unit of text a model processes, roughly a word piece; context limits are measured in tokens and billed per token in many APIs.

Common interfaces for tokens that define how balances, transfers, and metadata work across wallets and dapps, enabling composability and tooling support.

Maximum number of tokens a model can attend to in one request, constraining prompt size, tool outputs, and in‑context examples.

Process of splitting text into tokens using schemes like BPE or WordPiece; affects vocabulary, context use, and cost.

The supply, distribution, incentives, and utility design of a token that shape user and validator behavior.

Agentic pattern where models call external tools or APIs (search, code, DB) via function‑calling or adapters to extend capabilities beyond pure text.

Sampling restricted to the k most‑likely tokens at each step, trading off creativity and risk of nonsense for control and stability.

Sampling from the smallest set of tokens whose cumulative probability exceeds p (e.g., 0.9), balancing diversity with coherence in generation.

Metric for blockchain throughput, often compared across networks; raw TPS claims can be misleading without details on security and data availability.

A rough throughput metric for blockchains, often limited by block size, latency, and verification costs.

Telemetry that records model calls, prompts, tool invocations, and latencies for debugging, evals, and compliance in production AI apps.

Corpus used to fit model parameters; quality, diversity, and licensing shape capabilities and risks of memorization or bias.

Execution of model training with a specific dataset, objective, and hyperparameters, tracked for reproducibility and evaluation across checkpoints.

Signed message that changes on‑chain state, such as transfers or contract calls; includes nonce, gas, fee, and payload fields specific to each chain.

Amount paid to compensate validators/miners for computation and inclusion; may include base fee, tip, and burn depending on the fee market design.

Assurance that a confirmed transaction will not be reversed; can be probabilistic (PoW) or deterministic via BFT finality gadgets or checkpoints.

Monotonic counter per account that prevents replay and orders transactions; gaps can stall later txs until missing nonces are mined or replaced.

Pending transaction queue held by nodes before inclusion in a block; subject to fee competition, replacements, and local policy differences.

Reusing knowledge from a pretrained model and adapting it to a new task or domain via fine‑tuning or prompting strategies.

Neural network architecture built on self‑attention and feed‑forward layers that models long‑range dependencies efficiently; foundation of modern LLMs.

Reasoning strategy that explores multiple solution branches with reflection or scoring, guiding the model to better final answers than single‑pass decoding.

Cross‑chain bridge that reduces reliance on multisigs or custodians by using light clients, validity proofs, or on‑chain verification for safety.

Hardware‑based secure enclave that runs code with memory isolation and remote attestation, used for oracles, rollups, and privacy‑preserving apps.

Property of a system where users need not trust a specific party; correct behavior is enforced by cryptography, consensus, and open verification.

Aggregate value of assets deposited in a protocol or chain, commonly used to gauge usage and liquidity, but sensitive to price swings and double counting.

AMM design that executes large orders gradually over many virtual trades against a pool, approximating TWAP execution on‑chain with lower slippage.

Average price over time, used for on‑chain oracles and execution strategies to reduce market impact and manipulation risk relative to spot readings.

Design that separates network security (NACKL) from network usage (SHELL) to align incentives and costs.

Time delay between initiating unstake and when tokens become transferable, used to protect PoS networks from instant exit attacks or slashing evasion.

Quantifying model confidence or epistemic/aleatoric uncertainty in predictions; used for calibration, abstention, and risk‑aware decisions.

Ethereum term for a valid block not in the canonical chain due to a race; referenced by later blocks for partial rewards, now called “ommer”.