# Tokenomics — Utility Token & DePIN Model This document is the full analysis behind the README [Tokenomics — Utility Token & DePIN Model](../README.md#tokenomics--utility-token--depin-model) section. It covers the economic design of SN90 (KubeTEE) Alpha: recycle vs burn, securities posture, TAO on Base, the cross-subnet consumption loop, the vertically-split corporate structure, and the DePIN subsidy trajectory. > **Not legal advice.** The securities analysis below is engineering design rationale, not legal advice. --- ## Overview SN90 (KubeTEE) Alpha is a **utility token consumed to access confidential compute**, not a security. The design follows a DePIN subsidy model: - External inference demand buys Alpha on the **open market** and spends it to consume compute. TAO — the asset swapped for Alpha — is now also **live on Base** as a Chainlink CCIP-bridged ERC-20 (2026-08-21; see [TAO on Base](#tao-on-base)). - Spent Alpha is **recycled** to unissued supply and re-emitted through the protocol's fixed emission split — a self-sustaining security budget for the compute network (the Bitcoin-fee model applied to Alpha). - No entity sells tokens against promises, holds customer balances, or accumulates a treasury — all unused emissions are recycled. Value reaches each entity only through protocol safe-harbor channels (owner emissions, scored miner emissions). The goal is a token whose value narrative is **compute access**, verifiable on-chain by anyone — the cleanest fact pattern for a utility-token / DePIN classification. --- ## Recycle vs Burn ### Mechanics In Bittensor, burned tokens are removed from circulation **irreversibly**, while recycled tokens are removed from circulation but **can be issued again** — they return to unissued supply and flow back out through future emissions. Two operational consequences: - **Burning does not reduce `SubnetAlphaOut`**, unlike recycling. - **Recycling shifts halving timing** — halvings trigger on supply thresholds, and the date moves based on the amount recycled each day. ### Where "recycle instead of burn" does not buy anything Switching between recycle and burn is **not an emission-share optimization** — it does not change a subnet's emission share, only how supply moves. Recycle and burn differ on supply dynamics (below), not on how much emission a subnet captures. ### Where it does matter - **Halving runway.** Halvings trigger on supply thresholds; the date shifts with how much TAO is recycled daily. Recycling pushes the threshold out and preserves the future emission stream. Burning does not. If the revenue model is emission-denominated over a multi-year horizon, recycle is strictly better for the operator and worse for near-term holders. - **Price support.** Burn is irreversible and removes float permanently. Recycled supply comes back as future emission — the sell pressure is **deferred, not eliminated**. Burn is the stronger commitment signal precisely because it is costly and unrecoverable. The choice is a **duration trade**: recycle optimizes the operator's long-run emission entitlement; burn optimizes current-holder scarcity. ### Why recycle for a compute subnet For a compute subnet whose product is **ongoing work** (inference, training), recycle is the right economics: consumption funds future miner emissions — a self-sustaining security budget. The honest framing of the tradeoff: burn optimizes holder value and legal defensibility; recycle optimizes network sustainability and the operator's own future emission entitlement. A subnet can legitimately choose recycle and defend it as the more "commodity-like" behavior — reinvesting fees into the programmatic production of the commodity itself. ### Registration collateral can change where the registration fee goes (Phase 1) Subtensor v437 adds a third possible destination for the registration price. With a nonzero `CollateralLockShare` (p), only the `(1 − p)` share is destroyed; the `p` share is **staked to the registering hotkey and locked**, then released back to withdrawable stake at `CollateralDrainRatio` (k) Alpha per Alpha of emission that hotkey earns. **SN90 sets p = 0 in Phase 0**, so the registration price is burned and recycled exactly as described above — nothing in this section changes today's supply mechanics. What exists now is the **100 TAO deposit** miners post directly onto their own hotkeys, which is a validator-enforced requirement rather than a registration split (see [Miner deposit](../README.md#miner-deposit-registration-collateral)). Bonding a share of the registration price is a [Phase 1](../README.md#phase-1--expansion) decision, because choosing `k` requires per-miner emission data the subnet does not have yet. Three consequences worth stating plainly, because collateral is easy to mistake for a fee or a bond a platform holds: - **It is not revenue and it is not a treasury.** The locked Alpha sits on the *miner's own* hotkey. KubeTEE never holds it, no KubeTEE key can move it, and it never appears on anyone's balance sheet. It does not touch the no-treasury posture above. - **It is not slashable.** Bittensor has no confiscation extrinsic. A miner that breaches the SLA stops being scored, which stops its emission, which stops the lock draining — the deposit **freezes** on the miner's hotkey rather than transferring to anyone. Nobody is enriched by another miner's failure, so the mechanism creates no incentive to get rivals disqualified. - **It would reduce the recycled share of registration.** This is the one that matters for the model above, and the reason it is a deliberate Phase 1 decision rather than a default. A locked Alpha is not a recycled one, so a nonzero `p` means proportionally less of each registration returning to unissued supply. Set against that: a miner's only route to recovering the lock is to earn emission, so collateral converts part of the registration fee from an immediate supply event into a **claim contingent on doing the work** — closer to, not further from, the commodity framing above. --- ## Securities Posture — No Treasury SN90 has **no treasury**. All unused emissions are **recycled** to unissued supply with zero discretion anywhere — nobody holds them, nobody times their disposal, no customer-balance liability exists. This removes the entire treasury category: no accumulation, no discretion, nothing to mischaracterize. The SEC/CFTC interpretive release (Release No. 33-11412, March 17, 2026) applies the Howey test to **transactions** rather than assets themselves, and addresses how a non-security crypto asset may become subject to, and cease to be subject to, an investment contract. The question is not "is Alpha a security" but "do any of our practices create a scheme where purchasers reasonably expect profits from our managerial efforts." Eliminating the treasury eliminates the most attackable set of those practices up front. ### Fact patterns the no-treasury design avoids 1. **Receiving emissions (safe).** The release includes explicit safe harbors for protocol mining, staking, and airdrops — tokens received programmatically per protocol rules are not the problem. The subnet owner's 18% arrives the same way a miner's 41% does. Passive accrual to a wallet is hard to attack. 2. **Accumulating and holding a discretionary position (avoided).** A large discretionary insider position is a Howey factor (common enterprise, reliance on a promoter). A dormant treasury sits as ammunition for the argument that value depends on what you will do with it. With no treasury, there is no such position to characterize. 3. **Selling treasury to fund operations (avoided).** Emissions → sell → servers and salaries is functionally a continuous primary distribution funding the enterprise; if buyers reasonably expect the funded development to drive token value, those sales transactions start to look like investment contracts — even while secondary trading of the same token stays commodity-like. The token can be a digital commodity and treasury sales can still be securities transactions. With no treasury, there is nothing to sell this way. 4. **Deploying treasury for holder benefit (avoided entirely).** Buybacks, price support, liquidity backstops, yield programs, "treasury works for the community" messaging — anything that positions discretionary decisions as the value driver walks directly into the "expectation of profits from the essential managerial efforts of others" prong. With no treasury, there is no discretionary deployment to make. ### Why recycle fulfills the burn-adjacent posture Durbin's programmatic purchase-and-burn model is extreme precisely because it removes the entire category: no accumulation, no discretion, nothing to mischaracterize. Burn makes consumption **terminal**: no beneficiary, no residual claim anywhere. Recycling all unused emissions reaches the same posture by a different route: there is still no accumulation, no discretion, and no residual claim — spent Alpha returns to unissued supply and is re-emitted only through the protocol's fixed 41/41/18 split. Nobody chooses who receives it; the protocol does. The load-bearing property is **no discretion anywhere in the loop**, not the choice of burn over recycle. For a compute subnet whose product is ongoing work, recycle preserves the emission runway that funds future miner compensation while keeping the same "nothing to mischaracterize" posture. ### Mitigants - Token is already functional (used per its programmatic utility on a functional system). - No fundraising framing, no roadmap promises tied to sales. - Arm's-length OTC disposal of any emission the owner entity does liquidate. - Disclosure — especially important for a related-party structure. - No staking of accumulated balances for yield (would add a passive-income feature on top). ### Owner conviction auto-locked to perpetuity The subnet owner — **KubeTEE AI LTD** (owns the mechanism, the $198k (≈1,003 TAO) subnet registration, and the 18% owner emission stream; no token sales against promises, no customer balances, no treasury — all unused emissions are recycled; see [Corporate Structure](#corporate-structure-vertically-split)) — commits to **auto-locking 100% of its conviction to perpetuity**: the owner's dTAO conviction — built by locking TAO into SN90's pool — is continuously re-locked on-chain so that it **never decays and is never withdrawn to liquidate**. This is a permanent, programmatic commitment, not a discretionary promise — the protocol enforces it, so it does not depend on the owner's ongoing goodwill. This complements the no-treasury posture on the owner side. The owner does not hold a discretionary liquid position to sell: value reaches the owner only through the 18% owner-emission safe-harbor stream, and the owner's own conviction is permanently locked out of the float. There is no large discretionary insider position to characterize as a Howey common-enterprise reliance factor — the owner has unilaterally removed its own exit. Mechanically, a perpetually locked owner conviction also reduces circulating supply and signals permanent alignment: the owner's economic interest is bound to SN90's long-run success, not to a liquidation event. This is the owner-side analog of "recycle, don't accumulate": where the subnet recycles spent Alpha instead of holding a treasury, the owner locks conviction instead of holding a liquid position. --- ## TAO on Base **TAO is live on Base** as of 2026-08-21 — announced by [ForeverMoney (SN98)](https://x.com/forevermoney_ai/status/2090469070248235027). It is a Chainlink **CCIP-bridged ERC-20** of native Finney TAO, not a new token and not liquid-staked xTAO (Project Rubicon). Same asset, two rails. | | Finney TAO | TAO-on-BASE | |---|---|---| | Chain | Bittensor (Subtensor) | Base (Coinbase L2) | | Form | native `TAO` | ERC-20 `TAO` | | Contract | — (native) | [`0xf3081494b87e8d5fb7960f066e931d1d0e6e3d67`](https://basescan.org/token/0xf3081494b87e8d5fb7960f066e931d1d0e6e3d67) | | Bridge | — | Chainlink CCIP via [forevermoney.ai](https://forevermoney.ai/) | | Spot market | SN dTAO pools (TAO ↔ Alpha) | Aerodrome TAO/USDC | What this changes for SN90: - **Acquisition is no longer Bittensor-wallet-only.** A consumer (or a consuming subnet's treasury) can buy TAO from USDC or ETH in any Base wallet, then CCIP-bridge to Finney and swap into SN90 Alpha on the open pool. The "must download a Bittensor wallet first" friction that kept EVM capital off-subnet is gone. - **Phase 2 billing gets a native Base settlement asset.** The Early Access path is still Alpha / Finney TAO at a resources price per hour. Phase 2 layers **USDC-on-BASE and TAO-on-BASE** pull-based billing on top. The USDC→TAO hop is now a Base DEX swap (Aerodrome), not a custom KubeTEE bridge. Recycle still happens on Finney: USDC → TAO-on-BASE → CCIP → Finney TAO → SN90 Alpha → spend → recycle. - **The no-treasury / no-discount posture does not change.** TAO-on-BASE is another on-ramp onto the same open-pool acquisition leg. There is still no preferential Alpha placement, no volume discount, and no KubeTEE-held customer balance. A Base wallet buying TAO on Aerodrome and a Finney wallet swapping TAO for Alpha are the same economic act. What this does **not** change: SN90 Alpha is still the access ticket; spent Alpha is still recycled on Finney; emissions still split 41/41/18. Bridged TAO is a representation of Finney TAO, so circulating supply on Base is locked on Finney — it does not mint extra TAO. --- ## Cross-Subnet Consumption Loop ### The protocol-native (no-invoice) model The cleanest structure is the **protocol-native model**: SN90's Alpha is the access ticket to the compute; any other subnet or AI service (e.g. an inference subnet like SN64 / Chutes) acquires it on the open market and spends it to consume. No bilateral paper anywhere. **Securities-wise this is the strongest possible fact pattern for SN90.** The absence of a contract is not a gap — it is the feature. There is no bilateral scheme to characterize as an investment contract, just spot acquisition of a token followed by its consumptive use, which lands squarely in the release's carve-out that securities laws generally do not apply to items purchased for use or consumption. Better still, it upgrades SN90's own token classification: an Alpha that **must be spent to obtain compute** is a token with genuine programmatic utility on a functional system — precisely the digital-commodity definition — rather than a token whose only story is speculation. Demand from subnets and AI services gives organic, verifiable consumption. ### The flywheel ```mermaid flowchart LR Cust["External customers
pay fiat for AI services"] --> Cons["Subnets & AI services
(e.g. SN64 / Chutes inference)"] Base["BASE L2
USDC / ETH"] -->|"Aerodrome swap"| BaseTAO["TAO-on-BASE
CCIP ERC-20"] BaseTAO -->|CCIP bridge| Cons Cons -->|swap TAO for Alpha
on open pool| Pool["SN90 Alpha pool
(open market, no discounts)"] Pool -->|Alpha| Cons Cons -->|spend Alpha
for compute| SN90["SN90 (KubeTEE)
confidential compute"] SN90 -->|spent Alpha recycled| Unissued["Unissued supply
(zero discretion)"] Unissued -->|re-emit via 41/41/18| Proto["Bittensor protocol"] Proto -->|miners / validators / owner| SN90 Cons -.->|run validator on SN90| Val["Consumer-aligned validator
scores miner output = SLA"] Val -.->|Yuma Consensus| Proto ``` The acquisition leg is where the value transfer happens: a consuming subnet or AI service swaps TAO into SN90's pool to get Alpha — that swap is the real payment. Every purchase is TAO inflow into SN90's reserve and upward pressure on the Alpha price. TAO itself can now be bought on **Base** (USDC/ETH → Aerodrome → TAO-on-BASE → CCIP → Finney) as well as on Finney, so the buyer set is any EVM wallet, not only a Bittensor-native one. Under the current emission model, each subnet's share of block emissions is proportional to its EMA token price normalized across all subnets, so sustained consumer buying directly raises SN90's emission share. The flywheel: external enterprise AI-service revenue → consumers buy SN90 Alpha → price and emissions rise → miner incentive grows → more compute capacity → more services served. Because consumer demand is funded by outside customers, this is genuinely **external demand one hop removed** — not the circular emissions-recycling pattern that gets subnets dismissed as hot-potato economics. ### The spend leg (design decision) When a consumer spends Alpha for compute, SN90's mechanism decides what happens to it: - **Burn** — permanent supply reduction, does not touch `SubnetAlphaOut`, maximum scarcity signal. - **Recycle** (chosen) — returns to unissued supply, reduces `SubnetAlphaOut`, extends the Alpha emission runway and pushes halving thresholds out; effectively refills the budget that pays miners. - **Route to miners** as supplemental incentive on top of emissions — directly couples miner pay to real usage (tightest incentive alignment) but reintroduces sell pressure since miners liquidate. Recycle is the mechanically interesting choice for a compute subnet: recycling consumption back into future miner emissions is a self-sustaining security budget, the Bitcoin-fee model applied to Alpha. ### Reflexivity risks (the flywheel spins both ways) - **Concentration**: if any single subnet or AI service becomes the dominant source of TAO inflow, SN90's emission share is a derivative of one customer's purchasing schedule. The day they pause, your price EMA decays, emission share follows, miner revenue drops, and capacity exits. Single-customer alpha demand is how a subnet deregisters fast — diversify the consumer base across multiple subnets and services. - **Predictability**: if a consumer buys on a fixed cadence sized to revenue, that flow is visible on-chain and will be front-run. Use continuous small swaps (TWAP-style) rather than monthly lumps — this also smooths the EMA rather than spiking it. - **Inventory exposure**: a consumer holds SN90 Alpha between purchase and spend, carrying the token's volatility as working-capital risk. The deeper the pool, the smaller that risk; pool depth is partly the subnet owner's problem to solve. ### Quality enforcement without contracts No SLA means SN90's validators **are** the SLA: scoring must measure exactly what consumers experience — latency, throughput, uptime, correctness of the compute delivered — or miners will optimize for the metric while consumers experience something worse. An elegant protocol-native option: a consuming subnet or AI service (or an entity aligned with them) runs a validator on SN90, so the customer's own quality observations feed Yuma Consensus directly. The customer scores the vendor's workers, and the chain settles it. ### Keep the loop at the market layer The clean version is entirely arm's-length at the market layer: every consumer swaps on the open pool like anyone else, spends like anyone else, and SN90's edge is just being the subnet whose token has real consumers (subnets and AI services, not just one named partner). That is also the version where the flywheel narrative — usage-backed Alpha demand — is verifiable by anyone reading the chain. --- ## Corporate Structure (vertically split) ```mermaid flowchart LR Proto["Bittensor protocol
41 / 41 / 18 emission split"] Proto -->|"18% owner stream"| Kube["KubeTEE AI LTD
subnet owner
mechanism + IP"] Proto -->|"41% miner stream
(one pool)"| Pool["Miner emission pool
(scored, competed for)"] Pool -->|"scored share
shrinks as network grows"| Hori["1-HORIZON LTD
miner operator
GPU/TEE capex"] Pool -->|"scored share
grows as network grows"| Ext["External miners
(permissionless, competitive)"] ``` - **KubeTEE AI LTD — subnet owner**: owns the subnet layer (mechanism, the $198k (≈1,003 TAO) registration, the 18% owner stream — 100% conviction auto-locked at perpetuity). Receives value only through the owner-emission safe-harbor path. - **1-HORIZON LTD — miner operator**: a miner competing inside SN90 for a slice of the 41%. Funds GPU, TEE hardware, and rack commitments. Registers, competes, and is deregistered under identical rules as every other miner — same registration cost, same immunity period, no reserved UIDs. ### What the split gets right Separating the subnet-owner entity from the miner entity means each receives value **exclusively through protocol safe-harbor channels**: KubeTEE's 18% arrives as owner emissions; 1-HORIZON's share arrives as scored miner emissions. Neither entity needs to sell tokens against promises, hold customer balances, or accumulate a treasury — all unused emissions are recycled. Capex lands in the right place — GPUs, TEE hardware, rack commitments sit in 1-HORIZON, insulated from the subnet-layer entity, which owns only the mechanism and IP. ### Concentration risk (classification risk is concentration, not existence) A related-party miner on the subnet is unremarkable — most owner teams bootstrap capacity this way. What degrades both the DePIN story and the digital-commodity posture is **1-HORIZON becoming the dominant miner**: at that point the network is not meaningfully permissionless, the compute consumers pay for is mostly produced by the owner's sister company, and token value starts depending on one director's operational decisions across both entities — the managerial-efforts prong creeping back in. This is the "own a DC, fully self-mine, limit access" anti-pattern. ### Load-bearing mitigations (in order) 1. The incentive mechanism must be **objective and published**, scoring measurable compute properties so no validator judgment call can favor 1-HORIZON. 2. The validator set should include parties KubeTEE does not control (a consumer-aligned validator from a consuming subnet or AI service doubles as both quality signal and independence proof). 3. 1-HORIZON registers, competes, and is deregistered under identical rules as every other miner — same registration cost, same immunity period, no reserved UIDs. 4. **Want external miners to out-compete 1-HORIZON over time** — a declining related-party share is the on-chain evidence the network is real. Target state: a taostats chart where the related-party (purple) share shrinks as external (gray) miners grow. --- ## DePIN Subsidy Trajectory ```mermaid --- config: themeVariables: xyChart: plotColorPalette: '#D97706, #16A34A' --- xychart-beta title "DePIN Subsidy Trajectory" x-axis "Time (horizon unknown)" 0 --> 100 y-axis "Value" 0 --> 100 line "Emission subsidy" [88, 83, 77, 70, 62, 53, 43, 32, 22, 12, 5] line "Consumption revenue" [10, 11, 13, 17, 22, 29, 37, 46, 57, 70, 84] ``` The amber **emission subsidy** line decays as emissions taper over an unknown horizon; the green **consumption revenue** line rises as consumer spend (from subnets and AI services) grows. They cross at the **crossover** — the point where net Alpha issuance ≈ 0 and consumers (not emissions) fund the miner budget through the pool. The exact date is unknown (recycle shifts halving thresholds), so the x-axis is an undated horizon, not a halving schedule. ### The DePIN subsidy thesis Consumers (subnets and AI services) get compute at below-market price because SN90 emissions subsidize the cost of compute. This is the goal of SN90: help subsidize the cost of compute, with a more efficient tech stack that monetizes compute better than a VM. ### Mechanics of who is paying A miner's all-in compensation is **emissions plus consumption spend**. When emissions cover most of the cost base, miners can price delivered compute below their cash cost, and the consumer pockets the gap. That gap is not free — it is paid by **Alpha dilution**, i.e., by stakers and holders absorbing issuance. The honest description: token holders are collectively funding customer acquisition for the compute network. That is legitimate DePIN bootstrapping (the Helium/Filecoin/io.net playbook), and stated that way it is also classification-safe — the discount is protocol-programmatic, available to anyone who buys Alpha on the open pool, not a managerial pricing decision. ### Net issuance and the crossover Net Alpha issuance = emissions out − consumption recycled back. - **Pre-crossover (amber)**: low utilization; inflation-funding the subsidy. - **Crossover**: as consumer spend grows, recycling offsets more of the emission; when consumption spend equals emissions, net issuance is roughly zero, and the "subsidy" has become a closed loop where consumers fund the miner budget through the pool. - **Post-crossover**: net-deflationary while still paying miners fully. This gives a single on-chain KPI worth publishing from day one — the **subsidy ratio** (emission value ÷ total miner compensation) — and its required trajectory is **monotonically down**. That number declining is simultaneously the economic health metric, the DePIN credibility proof, and the answer to anyone claiming the token only exists to farm emissions. ### Defenses against the reflexive spiral Failure mode: Alpha price falls → subsidy value shrinks → below-market pricing evaporates → consumers route compute elsewhere → consumption falls → price falls further. 1. **Subsidy tapers by policy expectation, not by surprise.** If a consumer builds unit economics assuming permanent 40%-below-market compute, the first halving breaks their margin model and the demand simultaneously. Consumers should price in a published glide path. 2. **Stack efficiency is the moat.** If the KubeTEE stack genuinely extracts more revenue per GPU-hour than VM rental — Kubernetes bin-packing multiple workloads per card instead of dedicated-VM idle time, TEE attestation commanding a confidential-compute premium, higher effective utilization — then part of the discount is structural and survives emission decay. Emissions buy time; stack efficiency is the moat. A subnet whose price advantage is 70% subsidy / 30% efficiency at launch needs to be 30/70 by crossover. ### Wash consumption (attack surface specific to subsidized-compute subnets) If the incentive mechanism scores miners on utilization, a miner can spend Alpha consuming its own compute — the spend recycles, but the score farms emissions worth more than the spend. Defenses: - Score **verifiable properties** instead: delivered capacity, attested TEE execution, latency/correctness on validator-issued challenges. - Treat real consumption as a demand signal rather than the direct emission driver, **or** make self-consumption economically neutral (spend ≥ emission value gained). Given 1-HORIZON mines on the subnet its sister company owns, this must be closed before anyone asks — a related-party miner on a utilization-scored subnet is exactly where an observer would go looking first. SN90 takes the first defense: miners are scored on provable available capacity, so self-consumption buys a related-party miner no score and the spend is pure loss. The residual is on the demand side — the planned competitive-pricing formula derives `demand_pressure` from Armada queue depth, and a deeper queue lifts the target price every miner is scored against. That must be closed before the term bears weight: see [Wash consumption (open question)](./COMPETITIVE-PRICING.md#wash-consumption-open-question). --- ## Boundary Conditions (what breaks the model) 1. **Owner-hotkey withholding** of miner emission — penalized regardless of recycle/burn, and re-centralizes the flow. 2. **Discretionary accumulation** inserted between spend and recycle (no treasury — any accumulation is the poison; all unused emissions must be recycled). 3. **Public messaging** that frames Alpha appreciation — rather than compute access — as the reason to hold. 4. **Preferential Alpha placement**, side-letters, or volume discounts to large consumers — contaminates the loop into a primary distribution with investment characteristics. 5. **Owner unlocking conviction** — auto-lock to perpetuity is the commitment; unlocking would re-introduce a discretionary liquid insider position and the common-enterprise reliance factor the design removes. The mechanism is defensible only as long as the story told about it matches the arrows. --- ## References - [Bittensor dTAO Whitepaper](https://bittensor.com/dtao-whitepaper) - [Emission in Dynamic TAO](https://docs.bittensor.com/dynamic-tao/emission) - [Recycling — Learn Bittensor](https://learnbittensor.org/concepts/tokenomics/recycling) - [Recycling — TaoStats](https://docs.taostats.io/docs/recycling) - [Tao Emission — TaoStats](https://docs.taostats.io/docs/tao-emission) - [TAO live on Base](https://x.com/forevermoney_ai/status/2090469070248235027) — ForeverMoney (SN98), 2026-08-21 - [ForeverMoney bridge](https://forevermoney.ai/) — Finney TAO ↔ TAO-on-BASE via Chainlink CCIP - [TAO-on-BASE (Basescan)](https://basescan.org/token/0xf3081494b87e8d5fb7960f066e931d1d0e6e3d67) — ERC-20 `0xf3081494b87e8d5fb7960f066e931d1d0e6e3d67` - [Bittensor (CCIP-Bridged) — CoinGecko](https://www.coingecko.com/en/coins/bittensor-ccip-bridged)