## Making synthetic tokens perpetual
Nice framing. The maturity in this design is doing two jobs at once: it lets the oracle resolve slowly, and it forces the holder to roll. I think one can keep the payoff and get rid of the maturity, at the cost of needing a continuous oracle. This might solve the roll-over cutoffs and make these markets more liquid.
### Value tranching
Locking 1 ETH and splitting it at a cutoff `S` could be called a **junior / senior tranche** of the collateral. With ETH/USD and a cutoff `S = $1500`:
- **Senior** = `min(x, S)` — capped at the cutoff. The “stable” leg.
- **Junior** = `max(0, x − S)` — the residual / call-like leg. The leveraged leg.
- **Senior + Junior = x** = the value of 1 ETH. Always fully collateralized; undercollateralization isn’t representable. Equivalent to `P + N`.
Worth keeping the option decomposition in view, because it tells you who pays whom: the junior `max(0, x − S)` is a **long call** (long convexity), and the senior `min(x, S) = S − put(S)` is **cash minus a short put** (short convexity).
At an oracle price of **$3000**, the 1 ETH of collateral splits into **$1500 senior + $1500 junior**. Because we’re well above the cutoff, the senior is flat there and **the junior absorbs the local moves**: ETH ticks up or down, the junior responds to it first. Same as your `N`.
### Making it perpetual + self-balancing
Two mechanisms, both borrowed from perpetual futures rather than from dated options:
**Perpetual: junior funds senior, which is just the option premium made continuous.** Note that even with synthetics with maturity, there’s *already* an implicit junior to senior payment. The junior is a long call, so it **pays** time decay; the senior is short a put, so it **collects** it. The long-convexity leg pays theta to the short-convexity leg; it’s simply baked into the prices the legs mint and trade at, and settled once, at maturity.
Drop the maturity and you can no longer bake that premium into a terminal settlement. You pay it **continuously**, as a marked funding accrual that transfers NAV from the junior tranche to the senior. **Funding is the perpetualized option premium.** This is exactly the dated-future to perpetual-future move: a dated future has no funding (its basis just converges to zero at expiry); a perp needs periodic funding *because* it never expires. Same here, the funding rate is what replaces convergence-at-maturity, and the continuous accrual *is* the roll.
**Self-balancing — the market sets the funding rate.** Let anyone permissionlessly mint or redeem either tranche against the underlying ETH. The **ratio of junior NAV to senior NAV** sets the funding rate the junior pays the senior.
- Higher Junior vs Lower Senior NAV → Higher funding rate (to incentivize more Senior vs Junior)
- Lower Junior vs Higher Senior NAV → Lower funding rate (to incentivize less Senior vs Junior)
This balances the market continuously.
### What this gives you
**Perpetual risk tranching at NAV.** The junior is **never liquidated as a one-time cutoff event**, there’s no strike date and no trigger. The exposure simply re-prices continuously. What it *does* still cost is **volatility decay**: holding a fixed exposure target on a volatile underlying via continuous rebalancing bleeds value ∝ σ².
**Liquid stablecoin.** The senior tranche in this example behaves like an ETH backed, yield bearing stablecoin (dependent on the oracle price). CDP designs have the same fragility, but only one sided liquidity. Only borrowers can mint or redeem the stablecoin (unless they are being liquidated). In the design above, senior tranche holders can always mint and redeem against ETH at 100% capital efficiency. The cost being borne by leverage token holders (by ceding control of the leverage in their ETH exposure).
### Risks
- **Oracle risk.** A continuous oracle can have a price that’s off vs the true underlying. Someone then mints/redeems the cheap tranche to harvest the mispricing. This is the *same* failure mode as oracle risk in lending, but now affects both sides (not just the borrower via liquidations)
- **Rebalancing risk.** This is the quadratic-decay channel: users exiting the senior or entering the junior *at an inconvenient time* accelerate the junior’s volatility decay. The decay is structural, but its rate is path-dependent on when others rebalance.
### Potential protections
- **Faster, more trustworthy oracles**, the kind used in perps markets
- **Edge-case gates.** If collateral value falls hard, disallow junior redemption (a max-LTV-style protection, exactly as in lending) so the senior buffer can’t be drained when value is falling.
- **Let a secondary market trade away from the oracle.** allow the tranche tokens to trade at a **premium or discount to the oracle price** via a bonding curve (slippage on entry/exit). This is the continuous analog of your prediction-market pre-filtering: if the market thinks the oracle is wrong, the senior’s discount widens and *prices the disagreement in* before the oracle catches up. **This is the answer to the “isn’t a continuous oracle fragile?” worry** — the oracle anchors NAV, but the market is allowed to overrule it on price.
### What you lose vs the dated-option version
You give up the **prediction-market secondary that lives *between* rollover events**. The original design is a secondary (prediction like) market interspersed with occasional primary (oracle) settlement events; this design flips it: a continuous primary mark with the secondary living *on top of* it at all times.
### What you gain
- **Full exit liquidity, always.** A dated/prediction market needs market makers and has no full exit liquidity until maturity. A perpetual NAV-redeemable structure can be exited at any time (except when gated as per above)
- **Continuously priced rollover risk.** The roll cost isn’t a discrete event you eat at maturity; it’s marked continuously into the funding rate and the secondary price.
final thoughts: the dated-option version buys a **slow oracle** at the cost of the **roll**; the perpetual version buys away the **roll** at the cost of a **continuous oracle**; but a continuous oracle whose errors are softened (no liquidation cliff) and overridable (market premium/discount).