Where the money comes from

2.1 The user earns option premium

The revenue is not a yield rate, an emission, or a share of someone else’s borrowing. It is the price a buyer pays for an option, received in cash at the moment of the sale. It arrives whether or not the option is ever exercised.

Here is a worked cycle. The premium and the fee are measured, the cycle is not: no option has ever been sold on this platform. The two numbers come from a live Derive quote captured on 2026-08-13 — a 7-day ETH call at delta +0.174, bid 9.40 — priced at the size the product would actually sell. Everything below the fee is what the ledger would record for an out-of-the-money expiry.

ETH call, 0.5 contracts, 7 DTE, delta +0.174     quote measured 2026-08-13

  gross premium                 +$4.70      bid 9.40 × 0.5
  Derive fee                    −$0.78      TAKER: $0.50 base + 0.03% of $944 notional
  ────────────────────────────────────
  net premium                   +$3.92      → the user

  settlement (OTM)               $0.00      option expires worthless, collateral untouched
  swap slippage                  $0.00      no conversion — OTM starts nothing
  interest                       $0.00      an ETH-side cycle holds no idle USDC to earn on
  ────────────────────────────────────
  user result                   +$3.92

  platform gas                  −$0.00      no swap this cycle
  platform fee                   $0.00      rate is 0, and the mechanism is not wired

Each line corresponds to a real account in the double-entry ledger — but no endpoint produces this decomposition today. The only reporting endpoint, GET /portfolio/statement, includes a row only when an event posts to the user’s transit accounts, which by construction excludes premium, the exchange fee, settlement, interest and gas. Building the per-cycle statement is outstanding work (§4.3), and the layout above is what it is specified to produce.

Had the same order filled as a maker rather than a taker — the engine tries maker first for 20 minutes — the fee would be $0.09 instead of $0.78, and the net premium $4.61. That single line is the difference between keeping 83% of the premium and keeping 98% of it, which is why the selection threshold spends the maker window on it.

2.2 The cost stack, measured rather than assumed

This is the part that most option-selling pitches omit, and it is where the product either works or does not. Two costs sit between the theoretical premium and the money that reaches the user.

Execution cost — roughly 13.7% of mark, about 16% of the premium received. Selling into a bid rather than at mark loses the bid/mark gap. The percentages below are shares of mark, which is the larger denominator; expressed against the premium actually received they are correspondingly bigger. Measured across the |delta| 0.13–0.30 band on 2026-08-12:

Exchange fee — largely flat, and therefore brutal on small positions. The variable part is min(rate × notional, cap × premium), so on ordinary premiums the notional term binds and the fee barely moves with the premium; on top of that a taker pays a flat $0.50 base fee. In practice that is about $0.78 per taker trade at a 0.5 ETH notional, whether the premium is $3 or $30. That fixed cost sets a hard economic floor:

for the fee to be ≤ 10% of premium   →   gross premium ≥ $7.80
                    ≤ 15%            →   gross premium ≥ $5.20
                    ≤ 20%            →   gross premium ≥ $3.90

The consequence, stated bluntly: far strikes on short expiries are economically dead. Delta 0.03–0.07 on a daily expiry loses money before the market moves at all. The composite threshold in the selection threshold rejects these automatically.

Together, roughly 20–25% of gross premium is consumed by execution. This figure matches what was independently measured by hand, though it arrives from a different source than expected — the order book, not the conversion.

Two structural mitigations follow directly from these numbers, and both are already in the code: the maker-first policy (turning $0.78 into $0.09 when it fills), and RFQ alongside the book (the bid/mark gap is a property of a thin book, and market makers quoting directly often beat it).

2.3 A second, smaller income line

Idle USDC earns interest. Derive distributes interest paid by negative balances to holders of positive ones, net of roughly 20% routed to the Security Module. On the USDC side of the wheel the full collateral sits idle for the life of every option, so this is not a rounding error — it is a small structural yield underneath the premium. It shows as +$0.11 in the cycle above.

2.4 Who bears which cost

The rule is one sentence: the platform pays gas; anything the market or the exchange charges falls to the user.

Costs are attributed per user regardless of who pays them. Gas is tagged to the user and the action that caused it but accumulates on a platform account — otherwise the unit cost of servicing one user would be unknowable, and that number is precisely what sets the eventual fee.

A useful structural finding: because Derive itself pays the gas for subaccount creation, deposits and withdrawals, platform gas reduces to sweeps on Derive Chain, the withdrawal relay fee and swaps on Base. This is materially cheaper than originally assumed.

2.5 How Arkada earns

Today the platform fee rate is zero, and — stated plainly — the mechanism is written but not yet wired into any step. The gross → fee → net path exists as a component with its own tests and its own account in the chart of accounts; no cycle calls it. The only fee actually booked today is the exchange’s.

That is a smaller thing than it sounds, because the accounting shape it needs is already in place: premium, exchange fee and net are separate postings rather than one collapsed number, so introducing a platform fee is adding a step, not restructuring the ledger. But it is not “already flowing”, and this document should not imply it is.

The form of the fee has not been chosen, and choosing it prematurely would be guessing. Three forms are on the table; only the percentage one is implemented — a performance fee would need a high-water mark, and there is none in the code:

The reason for not choosing yet is that the honest inputs do not exist. The entire purpose of the first live run is to produce two numbers:

• (b) what fraction of premium survives all costs — fees, spread, slippage, ITM losses;
• (c) what it costs the platform to service one user for one month.

(c) sets the floor under any fee. (b) sets the ceiling. Naming a fee before measuring both would be a number invented to fill a slide.

Three reports are specified against the ledger to produce exactly this: a per-cycle breakdown, a cumulative “why not more” decomposition from the theoretical maximum down to the actual result, and a per-user platform cost report. The ledger already carries every field they need — the reporting module itself is not yet written.

2.6 What we do not claim

There is no projected yield in this document, and there is not one in the product. The 20% annualised threshold in the selection threshold is a selection filter — the minimum at which an instrument is worth selling at all — not a target, a promise, or an expectation. It rejects; it does not forecast.

On a live board, typically two or three instruments out of 586 pass the threshold. That is the normal state of the market, not a malfunction.

Any realistic annual return also depends on a number nobody currently has: the fraction of days actually spent in a position. A cycle that finds nothing worth selling earns nothing, and also costs nothing. Until the live run measures how often that happens, an annualised return figure would be arithmetic performed on an unknown.