I have been specifying a new flagship this week: a pooled, autonomous, market-cap-weighted index vault on Ethereum that takes deposits and redemptions in a single asset, runs ERC-7540 async settlement, and holds up to 100 constituents. The index computes its own target weights onchain directly from market-cap data with no manager or committee. The engine runs the real index construction rules in Solidity, including iterative capping with overflow redistributed to convergence (preventing collapse into 70% Bitcoin and Ethereum), free-float adjustments excluding locked, treasury, and vesting balances, and a reconstitution buffer to avoid churning at the rank-100 boundary. The binding constraint is the circulating-supply number feeding the weights. Whoever controls that input can shift target weights and front-run the subsequent rebalance, making supply the soft underbelly of the autonomous-weighting thesis. The reflexive answer of decentralizing the oracle is wrong, as circulating supply for most tokens is an offchain fact that does not fully exist onchain and cannot be made trustless by force. The reframe is that supply moves slowly, unlike price, since emissions and unlocks are gradual and largely known ahead of time. This allows buying security with latency, which defeats front-running that relies on surprise. The design thus has three layers: minimize, then secure, then contain. 1. Derive circulating supply from totalSupply minus a timelocked, publicly auditable list of excluded addresses. 2. Secure the small residual with a bonded, challenge-windowed value where the window matches the delay that defeats front-running. 3. Contain slippage with a hard cap on how far any weight can move per epoch. The per-asset cap also does security work, as a pinned constituent ignores supply precision, and the oracle needs accuracy only where dollar amounts are small.