Risk Disclosure

1. Fixed risk does not mean no risk

1.1 Loss of what you paid. You pay for a position in full when you open it, and that amount is the most the position itself can lose. A long loses all of it if the index settles at or below the low end of its range; a short loses all of it if the index settles at or above the high end. You can lose the full amount even when the index moved in the direction you expected, if it did not move far enough.

1.2 Capped gain. A position's payout is capped by its range. A long gains nothing more once the index settles above the high end; a short gains nothing more once it settles below the low end. The maximum payout is gross settlement proceeds, not maximum profit.

1.3 No liquidation. Every position is fully funded when it opens, so the markets have no margin calls, liquidations, or funding payments. That does not guarantee an early exit, prevent a decline in a position's market value, or stop a pause, software failure, or other event from blocking access.

1.4 Limits of a payoff description. The amount a ticket shows as "You pay" describes the intended position economics. It does not cap every consequence of a stolen wallet, malicious approval, counterfeit asset, contract exploit, or unauthorized transaction involving other assets.

2. What the benchmarks represent

2.1 RAMX. For months expiring after November 1, 2026, RAMX tracks contract prices of new, standard-production, single-die DDR4 and DDR5 memory chips in four buckets: DDR5 16Gb (40%), DDR5 24Gb (20%), DDR4 16Gb (25%), and DDR4 8Gb (15%). DDR5 32Gb chips are a research candidate at 0%: their prices are evidence for a later decision, not a settlement input. Memory modules, stacked or multi-die packages, LPDDR, GDDR, HBM, and reclaimed, downgraded, or eTT/uTT material are outside it. It is not a comprehensive index of all DRAM, HBM, GPU hardware, manufacturers, or transactions in the memory industry. Read the applicable manifest rather than inferring coverage from a product name or a market illustration.

2.2 NANDX. For the same months, NANDX tracks contract prices of new, standard-production, single-die 3D NAND flash chips in four buckets: 1Tb TLC (40%), 1Tb QLC (30%), 512Gb TLC (20%), and 256Gb TLC (10%). NAND wafers, bare dies, stacked or multi-die packages, SSDs, eMMC, UFS, memory cards, SLC, MLC and planar NAND, and downgraded, white-label, untested, or reclaimed material are outside it. An assessment measures its publisher's defined market, which can mix manufacturers and layer generations; it is not a constant-quality transaction price.

2.3 BATX. From its first month, BATX Nov 26 (expiring December 1, 2026), BATX tracks China domestic industrial prices of new manufactured lithium-ion cells, in Chinese yuan, in four buckets: LFP prismatic cells for stationary storage (40%), LFP prismatic cells for electric vehicles (30%), nickel-based ternary prismatic cells for electric vehicles (20%), and ternary pouch cells for electric vehicles (10%). Assembled packs, complete storage systems, raw materials such as lithium carbonate or cathode powders, used or second-life cells, retail prices, and cylindrical cells are outside it. It measures price changes in yuan: settling in USDC adds no currency return, so a cost you pay in another currency can move differently.

2.4 Launch months and weights. The launch months, RAMX Oct 26 and NANDX Oct 26 (expiring November 1, 2026), keep their original definitions: 13 RAMX memory-module categories and 22 NANDX categories covering SSDs, eMMC, UFS, and raw NAND wafers, each priced by one memorymarket.com category page. Bucket weights are product-definition weights, not a statement of market share and not cash-purchased source influence. A later product version requires its own identified methodology.

2.5 Synthetic exposure. These contracts settle in USDC against a defined benchmark. They do not reserve components, guarantee a supplier's price, provide title to inventory, fund an insured production order, or entitle a holder to physical delivery. Proposed future industrial or inventory products are not rights attached to a current position.

3. Coverage, basis, and concentration risk

3.1 Basis risk. Your actual purchasing or selling price may move differently from RAMX, NANDX, or BATX because of specifications, quantity, geography, currency, supplier relationships, condition, taxes, shipping, credit, delivery timing, or negotiated terms. A benchmark position is not a guarantee of procurement cost or an exact hedge of a specific invoice.

3.2 Concentration. A limited basket can be sensitive to particular technologies, product rows, channels, or source publishers. Multiple pages can depend on the same supplier or underlying assessment and therefore provide less independent information than their number suggests. Published contract-price coverage is thin for some chip buckets, especially DDR5 24Gb, and few publishers cover BATX's electric-vehicle cell buckets, the pouch bucket especially, so a bucket may rest on few sources.

3.3 Data is not necessarily an executable price. Listings and assessments can be stale, unavailable for the displayed quantity, or unrelated to a completed transaction. A normalized index level or a futures entry price is not the cash price of a physical chip, wafer, module, drive, or battery cell.

3.4 Coverage gates. Incomplete or disputed source coverage can delay the oracle month that a market settles on. A market can be traded only while its oracle period is live, and a late listing does not move the fixed expiry, so a delay can shorten the period available for trading.

4. Oracle and settlement risk

4.1 Source-local measurement. The oracle follows accepted changes relative to each source's own opening state. A wrong opening value, wrong field, SKU mismatch, or incorrectly identified source can distort subsequent movement even when later arithmetic is correct.

4.2 Median aggregation. The current oracle uses deterministic medians within governed SKU buckets rather than weights proportional to source-backing cash. Fixed product weights apply at their defined basket level. A median can resist some isolated outliers but does not guarantee truth when sources share errors, are manipulated together, or fail to represent the intended market.

4.3 Settlement window. Final settlement takes each source's temporal median over the last five business days before expiry, relative to its opening, and then the bucket medians and product weights. The final value can differ from the live index shown on the site, the last trade, a chart, or a value you observed outside that window. The live index is an indicative reference, not the settlement value.

4.4 Unchanged sources. A source with no accepted observation inside the settlement window counts at its latest accepted print before the window. A carried-forward observation can be economically stale.

4.5 Disputes. Source challenges, opening and update challenges, and council decisions on unresolved disputes can affect timing and the eventual accepted value. The council has five seats and three must agree. Evidence may be inaccessible or ambiguous, a valid challenge may be missed, and a council decision can be wrong.

4.6 Publication and payment. The oracle finalizes a month only after expiry and a grace period, and settlement requires the applicable authorized publication and on-chain checks. Each market then settles on the value its oracle period publishes. The current deployment runs this automatically, but it does not promise payment at expiry or at any particular time. A settlement payout is credited to your cash in that market; moving it to your wallet is a separate withdrawal.

4.7 Fallback settlement. If a market's oracle period fails, or is still unsettled after expiry, the governance council may set the market's settlement value instead of an index result, at any time after expiry. The market stays unsettled until the oracle or the council settles it. The council's value is its decision, not a published index result, and can differ from where the index stood. Positions then pay as if the index had settled at that value.

4.8 Security assumptions. Settlement signers can withhold a settlement but cannot change the value the program computes. That does not protect against a wrong input, a manipulated source, or a defect in the program. Source independence, bond economics, dispute participation, and the actual deployed configuration remain important.

5. Trading, the order book, and liquidity

5.1 One shared book. Each market has one order book shared by every range and leverage. Your order is filled from whatever combination of resting orders and market-maker quotes gives the best total price, oldest orders first at equal prices. Your counterparty can be another participant or Amoeba's treasury market maker.

5.2 The market maker is not a guarantee. Amoeba's treasury market maker quotes from a limited, fully funded budget, like any other participant. It is not obliged to quote, its losses shrink what it can quote next, and it can stop or run out of budget. Its quotes do not guarantee liquidity, a price, or an exit.

5.3 Price and execution. Quotes can change before execution, and a displayed entry, price line, or estimated position value may not be executable for your size. Thin depth or a price gap can cause slippage or rejection. A slippage limit or limit price can prevent an unacceptable fill but cannot create liquidity, and a limit order can rest unfilled.

5.4 Exits. Before expiry you can close a position only by selling it into the book at the price available then. That price can be far below the position's value at the current index, especially for wide or deep in-the-money ranges, and nothing may be available at all. You may have to hold a position through settlement even when you would prefer to sell it.

5.5 Trading can stop. A market trades only while its oracle period is live and before expiry. A pause, governance freeze, stale transaction preparation, infrastructure failure, or a market's finite capacity for orders, traders, and positions can prevent trading. New expiries are listed separately, and a listing is not guaranteed.

5.6 No standing buyer. There is no assurance that Amoeba, its market maker, or any participant will buy your position. A displayed book, price line, or P&L figure is not a redemption promise.

6. Contribution, bond, and reward risk

6.1 Assets at risk. Source support, claims, and challenges can require assets to be committed. Incorrect work, a failed challenge, missed deadlines, an invalid evidence format, or an adverse dispute result can result in losses or forfeited rewards. Being sincere is not the same as satisfying the protocol's validity rules.

6.2 Rewards are conditional. Reward budgets are finite and subject to funding, eligibility, and finalization. Ordinary cash challenges are participant-funded contests, not risk-free tasks. An advertised reward is not a salary, investment yield, or guaranteed payment, and an unchanged-source observation is not automatically a rewarded update.

6.3 Evidence risk. Web pages can change, archives can fail, dynamic prices can be misread, and source publishers can block access. Do not assume that a live page remains available throughout a challenge or that a screenshot proves the required product, field, and time.

6.4 Public exposure. Evidence, wallet activity, disputes, and outcomes may become public and durable. Do not submit private credentials, unnecessary personal information, confidential correspondence, or information you are not permitted to disclose.

7. Wallet, infrastructure, and settlement-asset risk

7.1 Wallet and identity security. A compromised device, email account, embedded-wallet login, private key, or recovery mechanism can cause loss. A convincing website or support message may be fraudulent. Confirm the origin and inspect each signature and approval.

7.2 Software and dependencies. Smart contracts, transaction builders, clients, databases, APIs, RPC services, and wallets can contain errors or be compromised. Tests and reproducible builds cover particular properties and artifacts, not every source of operational or economic risk. Amoeba's programs have not been audited.

7.3 Transaction lifecycle. Preparation, signing, submission, confirmation, and finality can fail at different stages. Separate operations in a multi-step workflow are not all reversed merely because a later operation fails. Verify existing status before retrying, and retain the relevant transaction references.

7.4 Stablecoins and networks. A USDC-denominated amount is a claim involving the specified token and network, not cash held in your bank account. Token value, issuer restrictions, transfer freezes, redemption access, and network disruption can affect usability and value. A token with the same name and a different mint address is not interchangeable with USDC (mint EPjFWdd5AufqSSqeM2qN1xzybapC8G4wEGGkZwyTDt1v).

7.5 No assumed insurance. Do not assume that a position, cash held in a market, Earn deposit, or blockchain wallet is protected by deposit insurance, investor compensation, or an Amoeba guarantee. Any actual protection would need to be expressly identified and applicable to your situation.

8. Understanding payout, cost, and sizing

8.1 Gross payoff. A long position on a range from L to H pays, per lot, min(max(S − L, 0), H − L) index points at settlement value S. A short on the same range pays the rest, (H − L) − min(max(S − L, 0), H − L). On the current markets one lot is worth $0.01 per index point. The specific market controls its units and rounding.

8.2 Cost and net outcome. A long pays its entry minus L per lot; a short pays H minus its entry per lot. Held to settlement, the net result is lots × $0.01 × payout points − what you paid. A chosen cash amount is not an underlying notional exposure of the same amount.

8.3 Example, not a quote. A long of 100 lots on the range 98 to 120, entered at 100, costs 2 points per lot: $2.00. Settlement at 106 pays 8 points per lot, $8.00, a net gain of $6.00; settlement at or above 120 pays $22.00, the most it can pay; settlement at or below 98 pays nothing and loses the $2.00. A short of 100 lots on the same range at the same entry costs $20.00, receives $22.00 at or below 98, and loses the whole $20.00 at or above 120. Actual markets and prices differ.

8.4 Leverage. The leverage you choose sets the range's losing end: the higher the leverage, the closer that end is to your entry. In the example the long is 50x and the short 5x. At high leverage a small move against you loses everything you paid.

8.5 Capped hedges. A long stops gaining at the top of its range even if your physical procurement cost keeps rising, and a short stops gaining at the bottom of its range even if the value of your inventory keeps falling. Holding a range position is not equivalent to holding the underlying or an uncapped forward.

8.6 Models and estimates. Suggested ranges, the live index, the price line, P&L at the index, and leaderboard figures depend on assumptions and data. None of them is a forecast or a guaranteed outcome.

9. Earn and liquidity pools

9.1 Providing liquidity can lose money. An Earn deposit funds liquidity in a market. Its value depends on how that liquidity trades and how the market settles, and it can fall below what you deposited, including to zero.

9.2 No promised return. A pool is not a savings account. A displayed value or past result is not a promised return, and a withdrawal may not be available at the time or for the amount you want. Review the pool's rules shown at deposit and withdrawal.

10. Governance, upgrades, and operational control

10.1 Council control. On Solana mainnet the oracle and futures programs are upgraded through a governance controller whose five-seat council needs three approvals. The controller and its target programs retain upgrade mechanisms. This is not a claim that the entire system is immutable, governed by all participants, or free of concentrated control.

10.2 Freezes and changes. Governance or safety gates can block operations, and an upgrade can change executable behavior. Published checks, delays, and separated activation steps reduce selected risks but cannot guarantee that an authorized decision is beneficial or free of defects.

10.3 Frozen terms versus software. A versioned product definition or frozen oracle recipe is not the same thing as immutable application code, immutable governance, or permanent availability of an API. Review the actual deployment and applicable authority model. Do not assume that an old document or test receipt describes the current operational state.

11. Legal, tax, and participation restrictions

11.1 Legal treatment can differ by instrument, user, jurisdiction, and role. Futures positions, liquidity-pool deposits, and paid oracle participation need not have the same classification or requirements. The words "decentralized," "fully funded," "no liquidation," or "software" do not themselves establish permission to offer or use a product.

11.2 Access restrictions, provider decisions, regulatory action, or a legal dispute can interrupt services or affect transfers, withdrawals, and continued operation. A wallet connection is not a completed eligibility review. Do not bypass restrictions or treat an enabled button as legal authorization.

11.3 You are responsible for your own reporting and other applicable obligations. Transactions, rewards, token conversions, and dispositions may have tax or accounting consequences. Interface summaries are not a substitute for complete records or qualified advice.

12. Before committing assets

12.1 Confirm that you are eligible and that the environment is appropriate for the intended activity. Verify the network, program, token mint, index and its version, UTC expiry, direction, range, lots, entry, and what you pay.

12.2 Check the current market, oracle, and governance status; the depth available for your size and for a later exit; and any challenge or transaction deadline. For Earn, review the withdrawal rules rather than assuming a deposit can be withdrawn on demand.

12.3 Use only assets you can afford to put at risk. If you cannot verify the product, understand the signed operation, or tolerate a loss or an extended inability to exit, do not enter the commitment. Contact [email protected] for a service issue without sending credentials; contacting support does not stop a protocol clock.