Skip to content
For the complete documentation index, see llms.txt. This page is also available as Markdown.

📖Glossary and sources

Glossary

Term
Meaning

TPA

Tokenized Processing Architecture: an owned AI endpoint with an address of three letters and three digits (ABC-042) that points to an AI model, agent or API, either on the GPU pool or an outside API. The only gateway to AI on TPAI Chain. A TPA is not a wallet. Originally, and still among real estate and hospitality clients, Teleport Plaque Address. Also called an AI domain address.

DeAI TPA

The platform: Decentralized AI, Tokenized Processing Architecture.

TPC, TPCoin

Tokenized Processing Coin, the platform's settlement coin. Fixed supply of 17,576,000,000, never created and never burned.

TPAI Chain

A next-generation blockchain with built-in AI compute: Stage's layer 1 of original design, with a datacenter GPU pool under the protocol and a blockchain layer on top that decides, proves and pays. Mainnet on 11 October 2027.

TPE

Tokenized Processing Engine: TPAI Chain's own virtual machine. It has no EVM and no token-creation or mint function; everything deploys inside a TPA and uses AI.

AI contract

The rules for an AI model or agent inside a TPA: who can call it, its price, its permissions and where its income goes. TPAI Chain has AI contracts, not smart contracts.

GPU pool

The NVIDIA datacenter GPUs under TPAI Chain: Stage's at launch, approved partners in year one, open bonded datacenters later. Every GPU is attested before it takes jobs.

Session balance

TPC a user or agent places on a TPA once, like a prepaid card, so AI calls run without waiting for the chain. Calls are metered and settled on chain in batches; unused balance returns to the user.

Validator node

One of exactly 1,000 licensed, KYC-verified nodes that run TPAI Chain. One node, one vote.

Committee

The group of validator nodes chosen at random, with equal chance, to decide each block. It changes every round; every node still verifies everything.

Node tiers

Micro, standard, super and super max: tiers defined by a node's duties. Tier sets income and duties, never voting power.

Chain of nodes

Validator nodes working together as a pipeline that shares proposing blocks, relaying transactions and spot-checking AI compute. Every voting node still verifies for itself.

Super node

A higher-tier validator node that spot-checks AI jobs by re-running them.

Review council

Seven members elected by validators for one-year terms, with Stage holding at most two seats. It approves emergency fixes, hears appeals and handles force majeure.

Hardware attestation

Proof from NVIDIA confidential computing that a GPU is genuine and running approved drivers and serving software. It proves the machine, not the answer; the receipt records which model and adapter ran.

Receipt

The signed record of an AI job: fingerprints of the model, adapter, request and answer, plus metered tokens and GPU time, signed inside an attested GPU (or by the owner, for external API TPAs).

Spot check

A deterministic re-run of a sampled AI job, using its recorded model version, settings and seed, to confirm the result.

Challenge window

The period after each settled batch in which any user or validator can challenge a job with evidence. Proven cheating is slashed and refunded.

TPAI-21

TPAI Chain's built-in token standard, used only by TPC and by approved coins bridged in 1:1, with post-quantum signatures, confidential balances with an auditor key and configurable transfers.

TPAI21- address

A TPAI Chain wallet address: about 120–130 characters, built on a 512-bit hash with post-quantum keys and a checksum, controlled by two or more private keys.

Post-quantum signature

A digital signature designed to resist attack by quantum computers, such as NIST's ML-DSA (FIPS 204) and SLH-DSA (FIPS 205).

Auditor key

A key that lets an authorized auditor or regulator read confidential balances and transfer amounts.

TPC migration

The 1:1 swap of ERC-20 TPC for native TPC on TPAI Chain, live from mainnet day.

Coin bridge

The bridge for approved outside coins such as BTC and USDT, which arrive as TPAI-21 tokens through a TPA wallet after 48 hours. It opens only after the chain has run safely for a period.

System TPA

A TPA in the reserved range (YVH-216 to ZZZ-999) for hosted base models and network needs.

Meta Energy, PING

A direct, single-level referral: when a new TPA is minted it can ping an existing TPA, which earns TPC rewards from Community & Ecosystem and stronger compute priority. Self-ping is allowed.

Meta Energy Level

The share of the TPA price that sets the base reward α. It starts at 17.576% and only falls toward zero.

α ALPHA

The base Meta Energy unit: Meta Energy Level × TPA price, paid on every ping.

δ DELTA

0.1α, paid to the pinging owner on their first purchase.

β BETA, γ GAMMA

2α on a TPA's first ping received; 3α on its 5th ping.

λ LAMBDA, ω OMEGA

10α at each 10th ping up to the 90th; 100α at each 100th ping up to the 900th.

ω MAX

1,000α at the 1,000th ping. After that, each ping earns 2.9741α, with no end date, at the current Meta Energy Level and within the reward cap.

Trusted minter

Status that recognizes the biggest sellers at no extra cost: ω MAX (1,000 pings), Gold Minter (10,000) and Diamond Minter (100,000).

Minter branch

A broker or company that resells TPAs on its own website through Stage's API, like a domain registrar.

Community & Ecosystem

The 20% TPC allocation that pays staking rewards, Meta Energy rewards and early validator income, and receives half of every TPA subscription. Before vesting starts, those rewards can never exceed 49% of it.

SAFT

Simple Agreement for Future Tokens: the early adopters' purchase contract.

Stage model

Stage's open source, chain-hosted AI model, funded by the AI Computing Power allocation. Its code, methods, data and research are public; its weights run only on TPAI Chain. Part of Stage's long-term path toward ASI.

Frontier model

A leading closed model reached through its provider's API, such as GPT or Claude.

Base model

An open-weight model Stage hosts itself, such as Llama 3.2.

LoRA

Low-Rank Adaptation: a small add-on layer that fine-tunes a base model.

S-LoRA

A serving method that runs thousands of LoRA adapters on shared GPUs. Used on the GPU pool from day one.

Kubernetes pod

An isolated, addressable unit of running software. Each TPA works like one (preferred design).

/8 address space

A private block of 16,777,216 IP addresses, such as 10.0.0.0/8.

/8 overlay

TPAI's private network in which each TPA gets an IP from its address by a fixed formula, so finding a TPA needs no name server.

VXLAN

A network technology with up to 16,777,216 isolated segments: by default every TPA has its own, owners can share one among their TPAs, and validators can quarantine any TPA (preferred design).

ERC-20 / ERC-1155

Ethereum token standards: fungible tokens (TPC) and multi-token collections (TPAs).

Sepolia

Ethereum's public test network.

TGE

Token generation event: the token's launch.

VARA

Dubai's Virtual Assets Regulatory Authority.

Sources

Pages opened or records queried between 28 October and 16 November 2026.