Overview

Spark is an on-chain allocator of stablecoin liquidity and a lending protocol governed by its own SPK token — and, distinctively, it is a sub-DAO that operates inside a larger DAO. Spark is the first live "Star" in the Endgame architecture of Sky (formerly MakerDAO), so its scope, budgets, and risk parameters are defined by a governance artifact that lives within Sky's constitution rather than standing fully independent. That structure makes Spark the wiki's clearest worked example of the sub-DAO pattern: a semi-sovereign unit with its own token and voters, nested under a parent that retains ultimate control.

Functionally, Spark does three things — it lends (SparkLend), it pays a savings rate on stablecoins and ETH (Spark Savings), and it routes idle Sky-backed liquidity into external venues to earn yield (the Spark Liquidity Layer). Where standalone money markets such as Aave, Compound, Morpho and Euler answer to no one but their own token, Spark's mandate is set — and can be revised or revoked — by its parent DAO.

From a MakerDAO front-end to a Sky "Star"

Spark began in May 2023 as Spark Protocol / SparkLend, a lending market forked from Aave v3 and stood up by MakerDAO to lend out its own DAI at a rate the Maker governance voted directly — a native front-end for Maker's stablecoin rather than a third-party venue. It was, from day one, an arm of a DAO rather than a rival to one.

When MakerDAO rebranded to Sky and launched its Endgame plan, the design called for semi-independent "Stars" — sub-DAOs with their own tokens and governance that build on Sky's liquidity while remaining bound to the Sky Atlas. Spark became the first Star to go live, graduating from an internal product into a token-governed protocol in its own right while staying inside the parent's orbit. Its SPK token, launched in 2025, formalised that half-step: real governance for Spark, still framed by Sky.

What Spark builds

SparkLend is the lending core: users borrow USDC and USDS against ETH, staked-ETH (wstETH, rETH, weETH) and BTC collateral, at rates set through governance rather than a pure utilization curve. Spark Savings issues yield-bearing vault tokens (sUSDS, sUSDC) that pay a rate sourced from Spark's allocations, withdrawable back into USDC, USDT, PYUSD, USDS or ETH.

The Spark Liquidity Layer (SLL) is the piece that makes Spark an "allocator" and not just a money market: it deploys Sky-backed stablecoin reserves directly into external venues — Aave, Morpho, Ethena and tokenized real-world assets — across Ethereum, Base, Arbitrum, Optimism and Unichain. Because it can bootstrap products with its own reserves, Spark does not depend on mercenary third-party capital to seed a new market. Its 2025 $1B tokenized-treasuries "Grand Prix" — won by BlackRock's BUIDL, Superstate and Centrifuge — routed a large slice of that liquidity into RWAs.

The SPK token

SPK is Spark's governance and staking token, native to Ethereum. Its supply is fixed at genesis but its distribution is controlled by the parent: 10 billion SPK were minted, split 65% to a ten-year "Sky Farming" campaign (6.5B SPK, released to users through a Sky-controlled Pause Proxy), 23% to the Spark ecosystem & treasury (2.3B, including the concluded airdrop), and 12% to the team (1.2B, a 12-month cliff then three-year vest). Users still earn SPK today by depositing USDS into the SPK Farm.

SPK can be staked to earn Spark Points (and, via Symbiotic, external restaking points), with staked SPK earmarked to help secure future Spark products. One clause underlines the sub-DAO relationship: Sky retains the ability to mint additional SPK under extreme circumstances, governed by the Sky Atlas — the parent keeps a lever over its Star's own money.

Governance: a sub-DAO of a DAO

Spark's governance does not edit Spark's contracts directly. Instead it revises the Spark Agent artifact — a scoped document inside the Sky Atlas, Sky's machine-readable constitution — which defines Spark's budgets, risk settings, asset onboarding, Liquidity-Layer integrations and new-chain deployments. Changes take the form of Atlas Root Edit Proposals, so every Spark decision is expressed as an amendment to a sub-section of the parent's rulebook and must stay aligned with it.

SPK holders vote on and delegate over these proposals, with a whitelisted delegate set and an Operational Facilitator plus a risk-review mandate that can stop a malicious or high-risk proposal before it reaches a vote. The result is scoped sovereignty: Spark's community steers Spark, but only within a perimeter its parent drew and can redraw. This is the pattern the wiki covers in the abstract under sub-DAOs and working groups — Spark is what it looks like at a multi-billion-dollar scale.

Why the sub-DAO model matters

The sub-DAO answers a real governance problem: a large treasury-rich DAO wants to move fast in a new product area without either (a) subjecting every operational decision to a slow full-DAO vote, or (b) spinning out a fully independent protocol it can no longer align. A Star like Spark splits the difference — a dedicated token and voter base for speed and accountability, a constitutional tether for alignment and safety. It also lets the parent concentrate liquidity behind a focused team while keeping the option to wind the unit down.

The cost is the same tether: SPK governance is real but bounded, and the parent's residual powers — controlling the farming schedule, holding a mint switch, owning the Atlas the Agent lives in — mean Spark's holders never have the last word the way an Aave or Euler holder does over their own protocol. Whether that trade is worth it is the open question every DAO weighing a sub-DAO has to answer.

How Caper approaches this

Spark shows a parent DAO delegating scoped authority to a sub-unit and keeping residual control through tokens, budgets and a constitution its Star cannot rewrite. A caper works on a different unit of trust: rather than nesting governance inside a parent's discretion, it fixes each member's influence with a canonical vote weight w = (t · v) / (V · T) that folds together how much of the token they hold (t) and a soulbound, non-transferable proof-of-vote balance (v) earned by actually voting. The v factor cannot be bought, minted for insiders, or moved between wallets — so influence tracks participation, not a distribution schedule a parent controls.

The same weight drives the exit right: a member redeems a share of the treasury proportional to that weight and burns their vote balance on the way out. A large bag still matters — t is a multiplier, so holdings are not irrelevant — but a bag alone cannot capture control, and the treasury claim is a first-class protocol primitive rather than a budget a parent grants and can revoke. Where a sub-DAO borrows legitimacy from the DAO above it, a caper builds the accountability into the weight itself.