The index layer for tokenized stocks

01Background: equities become composable

On July 1, 2026, Robinhood launched the public mainnet of Robinhood Chain, an Ethereum Layer 2 built on Arbitrum technology with 100ms block times, purpose-built for tokenized real-world assets. Its flagship asset class: stock tokens, Robinhood-issued instruments tracking the performance of more than 200 US-listed stocks and ETFs, available in over 120 jurisdictions and trading around the clock.

This is a genuinely new substrate. For the first time, instruments tracking Apple, Nvidia and the S&P 500 exist as permissionless ERC-20 tokens with real distribution behind them: composable with AMMs, lending markets and smart contracts, and tradeable during the 128 hours per week when the New York Stock Exchange is closed.

The launch ecosystem reflects where incentives naturally pointed first: venues. AMMs, aggregation, perpetuals and lending all shipped on day one. Trading infrastructure is well served.

02The missing index layer

Most of the world does not hold equities as individual tickers. It holds them through index products: more than half of US fund assets are passive. The reasons are structural, and they apply onchain with at least equal force: diversification is the only free lunch in finance, picking single names underperforms on average, and wrappers compress decisions and costs.

Yet the onchain equity ecosystem launched with no index layer at all. The gap is not an oversight; it reflects who built first. Exchanges list instruments. Someone else has to package them.

Traditional index products carry three structural constraints that a token-native design removes:

• Issuance is gated. Launching an ETF takes an issuer, an exchange listing and months of process. Launching a basket takes one transaction.
• Discretion is hidden. Fund managers retain latitude that investors cannot audit. A basket's entire rule set is public bytecode.
• Access is fragmented. An index product listed in one market is unavailable in another. A basket is a token; it goes where the chain goes.

03Protocol design

3.1 Baskets and vaults

A basket is a standard ERC-20 backed 1:1 by a weighted set of 3 to 20 stock tokens held in a dedicated vault contract. The vault exposes exactly four capabilities: mint, redeem, rebalance and collectFees. Its rule set (holdings, target weights, rebalance schedule, drift bands, management fee) is fixed at deployment. Vaults have no admin keys and no upgrade path for holdings logic, and no role in the system can pause redemption.

3.2 Minting and in-kind redemption

Minting deposits the underlying tokens (directly, or via a router that swaps from a single asset) and issues basket tokens at oracle-priced NAV. Redemption is the inverse and is always available in-kind: burn basket tokens, receive the underlyings pro rata, in a single call with no external dependencies. This is the creation/redemption mechanic that anchors ETF prices, implemented without an authorized-participant gate: on Skep, everyone is an authorized participant.

3.3 Rules-based rebalancing

On the first trading day of each period (monthly or quarterly, fixed at creation), a rebalance window opens and anyone can execute rebalance(). The contract restores target weights for any holding outside its drift band, executing via TWAP through Uniswap to limit price impact. Outside the window the function reverts. The consequence is central to the design: there is no path, for any party, to trade vault assets outside the locked rules.

3.4 The AI research layer

Skep's research agent turns a plain-language thesis into a proposed composition: it screens the full universe of listed stock tokens, proposes holdings and weights within protocol constraints, and writes a complete research memo covering screening logic, per-holding rationale and risks. Two properties keep this layer honest:

• The agent has no onchain authority. It proposes; the creator decides; the contract enforces. A basket cannot be deployed, modified or rebalanced by the agent.
• Research is tamper-evident. Every memo is stored on IPFS and its hash registered onchain at deployment and at each rebalance, producing a research trail that cannot be quietly rewritten after performance is known.

3.5 Pricing

NAV is computed from Chainlink feeds per underlying. Secondary-market basket prices may deviate from NAV; in-kind redemption bounds the deviation to the cost of arbitrage. During hours when US reference markets are closed, stock token prices are set by onchain trading; baskets inherit that price discovery, and their NAV reflects the same 24/7 marks as the underlyings.

04The creator economy

Index composition becomes an open, credentialed market. Any address can deploy a basket and attach a management fee between 0% and 1% per year, accruing per block via supply dilution. Fees split 70% to the creator and 30% to the protocol. There are no performance fees, lockups or exit penalties anywhere in the system.

The interesting asset in this market is not the fee stream but the track record. A basket's performance exists onchain from block one, produced by locked rules, with its research trail hash-registered before outcomes were known. It is a performance record that cannot be backfilled, cherry-picked or survivorship-biased. We expect this to become the native credential for investment research: analysts publish baskets the way developers publish repositories.

05Design principles

• Non-custodial, always. User assets sit in per-basket vaults that no party, including Skep Labs, can access outside the four locked functions.
• Rules over discretion. Anything requiring ongoing human judgment over pooled assets is excluded from the design space by construction.
• Exit is sacred. In-kind redemption is unpausable and dependency-free. Every other component (router, front-end, agent, oracles for secondary pricing) is bypassable.
• Transparency before trust. Rule sets are bytecode, research is hash-registered, rebalances emit full reports. Nothing about a basket's behavior requires trusting its creator.
• Composability as distribution. Baskets are plain ERC-20s so that DEXs, lending markets and future protocols become distribution channels rather than integrations.

06Regulatory posture

Skep is designed as infrastructure for rules-based wrappers, not as a fund platform. The properties that matter to that distinction are structural rather than cosmetic: no party exercises discretion over pooled assets after deployment; holders retain a continuous, unconditional right to exit in-kind; economic rules are fixed, public and enforced by contract; and the protocol takes no custody at any point.

We are engaging counsel in the jurisdictions where the underlying stock tokens are distributed and will publish our legal framework before mainnet launch. Access restrictions follow the underlyings: stock tokens are not available to US persons, and neither are baskets built on them. Front-end access is geofenced accordingly.

Skep is an independent project. It is not affiliated with, endorsed by, or sponsored by Robinhood Markets, Inc.

07Risks

An honest account of what can go wrong:

• Smart contract risk. Audits and invariant testing reduce, and cannot eliminate, the risk of defects in vault or rebalancing logic.
• Issuer risk. Stock tokens are Robinhood-issued instruments that track equity performance. They are not direct share ownership, and baskets inherit their issuer, structural and legal properties.
• Market risk. Baskets hold equities. Equities fall. Diversification narrows the distribution of outcomes; it does not truncate it.
• Liquidity risk. Robinhood Chain is weeks old. Thin early liquidity affects single-asset mint/redeem routing and rebalance execution quality, particularly outside US market hours.
• Oracle risk. NAV depends on Chainlink feeds. Feed staleness or failure degrades mint/redeem pricing; in-kind redemption remains available regardless.
• Regulatory risk. The treatment of tokenized equities and wrappers built on them is evolving in most jurisdictions and may change adversely.

08Roadmap