Inspiration
Listed options only come in round lots of 100 shares. An advisor with a client who holds ~44 shares of an S&P ETF can buy puts for 0 or 100 — never exactly 44. Mini options failed for volume; we thought the fix was aggregation: sell odd lots on-chain, then net them back into round lots to hedge.
That story lined up perfectly with UBC 2026 — Solana for the vault, Pyth for the underlying, Gemini’s S&P year-end market for crowd pricing and residual hedges, and QuickNode Streams to wake the hedger on every trade.
What it does
Oddlot sells capped call and put spreads on VOO (S&P 500 ETF) in one-share lots, fully backed by USDC on Solana.
Puts — enter a dollar amount or share count; buy exact-quantity put protection in one transaction (the “protect 44 shares” demo) Calls — buy cheap one-share upside for retail-sized premiums Vault — LPs deposit USDC; the dashboard shows solvency (≥ 100% coverage), a netting meter (odd lots → round lots), and a live trade tape Every token locks its max payout in USDC at mint. Holders never depend on the hedger. Quotes take the higher of Black–Scholes (Wall Street IV) and Gemini’s crowd distribution, plus a small spread.
How we built it
Four workstreams sharing frozen contracts after M0:
Anchor program (Rust) — vault, series, buy, settle; Token-2022 option mints with permanent delegate for auto-exercise; Pyth pull oracle + market-hours guard; integer money only on-chain Pricer (Python / FastAPI) — Black–Scholes + Gemini SPXEND26 crowd pricing; hedger that nets every 100 tokens into one listed spread and papers residual Gemini bucket hedges Keeper (TypeScript) — posts quotes ~every 15s, cranks settlement, runs the test-USDC faucet and demo seeder Web app (Next.js) — Phantom wallet adapter, Puts / Calls / Vault screens, live Pyth updates in the buy path QuickNode Streams watches our program logs and webhooks Bought events into the pricer so hedging is event-driven, not poll-only.
Challenges we ran into
Hermes auth + feed entitlements — Pyth trial keys required Bearer tokens and entitled VOO rather than SPY; we cut over feeds and filtered leftover series so buys wouldn’t hit InvalidOracle SBPF / LiteSVM / Anchor versioning — platform-tools emitted ELFs that older LiteSVM rejected; we pinned versions and deployed with anchor build --arch v2 for devnet QuickNode Stream fragility — wrong dataset terminated silently; missing Content-Type caused FastAPI 422s and stream teardown — we parse the webhook body ourselves RPC quota burn — open judge tabs hammered QuickNode; we added visibility-paused polls, dual read/send RPCs, and batch account fetches Wallet buy UX — year-end buys needed Pyth ephemeral signers before the wallet and ATA setup folded into one batch so Phantom simulation wouldn’t fail with extra signatures
Accomplishments that we're proud of
A working Protect flow: exact-share puts in one tx, tokens in the wallet Backed by construction — max payout locked at mint; unsafe buys rejected; Vault coverage stays ≥ 100% Odd lots in, round lots out — live netting meter that tips at 100 shares Crowd vs Wall Street pricing wired to Gemini, with residual hedges on the year-end distribution Market-hours guard + auto-exercise via Token-2022 permanent delegate (demo shock settlement without holder clicks) Hosted demo: oddlot-vault.vercel.app + pricer on Vercel + program live on Solana devnet
What we learned
Fractional, fully collateralized options force you to design the vault first and the hedger second — off-chain can lose edge, never holder payouts. Prediction-market distributions are a real pricing and hedging input when payoffs are capped spreads. And on Solana, “it works locally” is only half the battle: oracle keys, stream filters, wallet simulation, and RPC budgets decide whether judges can actually click Buy.
What's next for Oddlot
More underlyings via xStocks / Pyth, hedge-aware collateral so LPs free up capital faster, on-chain delta hedges beyond paper fills, and a path to a licensed launch outside the US. The program was built so new tickers don’t require a redesign — just feeds, series, and risk limits.