A liquid staking token is supposed to trade at the value of the SOL behind it. When it slips, the discount is a number nobody can hedge: lending positions liquidate, withdrawals queue, and the only instrument is patience.
Unpeg turns that discount into a seven-day contract. One lot is 1 SOL of notional, funded in full with 0.03 SOL and split into a BREAK note and a HOLD note. Every 300 seconds, time spent below the peg moves collateral from HOLD to BREAK. Nothing is borrowed, nothing is levered, and the two sides always sum to the vault.
- Mint a pair before the series starts: 0.03 SOL in, one BREAK and one HOLD out
- Keep the side you believe. Sell the other at a fixed premium from escrow
- Every 300 seconds, two of three reporters co-sign the pool discount
- Discount above 25 bps accrues to BREAK, capped at 300 bps per interval
- After 2,016 intervals, burn your note for its share of the vault
Drag a scenario. The number is the real v1 formula, floored once per lot. A 325 bps discount held for 48 hours pays the full vault; 100 bps for the whole week pays 87.5%. This is a payoff preview, not a market observation.
SCENARIO
SPLIT OF THE 30,000,000-LAMPORT VAULT
- hSOL4
- JitoSOL16
- mSOL13
Superscript is the number of direct LST/WSOL concentrated pools Raydium's API indexed on 2026-09-20. Every market stays MONITOR_ONLY until two approved price-forming pools show sustained sampled coverage and measured manipulation cost. No series is listed.
| Term | 604,800 s |
| Observation interval | 300 s |
| Intervals per series | 2,016 |
| Discount deductible | 25 bps |
| Maximum excess per interval | 300 bps |
| Normalization duration | 172,800 s |
| Collateral and maximum payout per lot | 30,000,000 lamports |
| Note decimals | 0 |
| Collateral mint | WSOL |
discount_i = floor(10000 · max(0, reference_i − market_i) / reference_i)
excess_i = min(max(discount_i − 25, 0), 300)
area = Σ excess_i · 300
break_per_lot = min(30,000,000, floor(1,000,000,000 · area / (10000 · 172800)))
hold_per_lot = 30,000,000 − break_per_lot
Missing data favors HOLD. If an interval is not reported by its deadline, anyone may skip it permanently with zero accrual. There is no backfill, no governance override and no invented fallback price. The holder of BREAK accepts data-availability risk during exactly the period they care about.
Reporters are trusted, not proven. The program checks signatures, bounds and sequence. It does not verify that an RPC provider returned the true state of a pool. Three keys in one backend are not decentralization.
BREAK is not slashing insurance. It settles the market discount of an LST against its reported SOL value. NAV and price can fall together without a payout. HOLD is not yield, and a premium is not a probability.
All three markets are monitor-only.
| Pool · 2026-09-20 18:50–18:55 UTC | Samples | Max gap | Replayed 300 s window |
|---|---|---|---|
| JitoSOL 2uoK… | 21 | 31 s | Rejected · two gaps over 30 s |
| JitoSOL 5cnf… | 23 | 29 s | Coverage passes |
| mSOL 8Ezb… | 24 | 18 s | Coverage passes |
| mSOL HZf7… | 24 | 18 s | Coverage passes |
| hSOL 51V5… | 24 | 17 s | Coverage passes · single source |
Real finalized public-RPC snapshots replayed from PostgreSQL. A working collection path is not an approved market: listing a series requires two approved price-forming pools with sustained coverage, manipulation cost measured against series exposure, and operationally independent reporters. The program address in the IDL is a development address.