ATERA Protocol · Documentation

How Atera turns trading fees into inference.

$ATERA is a BEP-20 token launched on four.meme — the tax that funds this system is set once at creation on four.meme, written on-chain as an immutable rule, and converted into OpenRouter credit distributed to holders every hour. This document explains the full mechanism — where the money comes from, how a share is computed, and how a holder spends it. Every third-party figure quoted here is cited in References.

Chain

BNB Chain56

Transaction tax

1.00%

To holders

0.30% of volume

Cadence

60min

In one paragraph

$ATERA launches on four.meme. While it is on the bonding curve no tax is charged. Once the curve fills and liquidity migrates to PancakeSwap, a 1% tax applies to every buy and sell and routes to the collector, which converts it to USDT. Once an hour — an orbit — the collector's realised USDT is split 30/70: thirty percent buys OpenRouter credit at face value and is written to holders' ledgers pro-rata by time-weighted balance; the remainder is retained for liquidity, infrastructure, growth and the credit float. Net, 0.30% of every dollar traded arrives as spendable credit. Payouts are ledger rows, not on-chain transfers, so there is no gas and no minimum economic payout size.

01

The fee-stream problem

It is a common misconception that BNB Chain lacks fee infrastructure. It does not. The machinery is mature and, by fee revenue, competitive with Solana's. What is missing is a destination — every stream terminates somewhere the holder cannot spend.

Why a launchpad fee cannot do this job

four.meme's creator fee is real and it pays — while the token sits on the bonding curve. When the curve fills, liquidity migrates to PancakeSwap ^[3]. From that block onward the AMM pays the deployer nothing; its 0.25% is split entirely between liquidity providers, the PancakeSwap treasury and CAKE buyback-and-burn ^[2]. A launchpad fee is a launch-window fee, and a protocol that has to keep paying holders in year two cannot depend on one.

The plumbing exists. The destination does not.

four.meme's tax mode already lets a project charge a tax after graduation and route it wherever it likes — a wallet, a burn, automatic liquidity, or dividends paid straight to holders ^[1]. The rules are written on-chain when the token is created and cannot be modified afterwards. None of that is the hard part, and Atera invents none of it.

The hard part is what the money turns into. A dividend pays BNB, or more of the token you already hold — an asset you must still sell before it does anything for you. A burn pays nothing and asks you to believe scarcity will. Both hand you back the same bet in smaller pieces.

02

Fee mechanics

The tax is 1%, charged on every buy and sell once the token has graduated from four.meme's bonding curve to PancakeSwap. Nothing is charged during the curve phase, and plain wallet-to-wallet transfers are not taxed. The rate is one of four.meme's fixed options and is immutable once the token is created.

Per $100 of volume

03

The collector

Tax is collected in ATERA, not in dollars. It must be swapped before it means anything, and that swap is the most delicate operation in the system.

Swap-back

• The collector accumulates until its ATERA balance crosses a threshold, then swaps ATERA → WBNB → USDT on PancakeSwap V2.
• The swap is guarded by a reentrancy lock. A transfer that triggers swap-back must not re-enter the tax path mid-swap — this is the classic failure mode of fee-on-transfer tokens.
• Swap-back never fires on a buy, only on a sell or a plain transfer, so it cannot be used to sandwich an incoming buyer.
• Slippage is bounded. A swap that would exceed the bound reverts and the balance is carried into the next attempt rather than dumped into thin liquidity.

04

Orbits & time-weighting

An orbit is a sixty-minute settlement window. At its close, the pool is divided by time-weighted average balance (TWAB) rather than by balance at a single instant. A snapshot would be trivially gamed by buying one block before the boundary and selling one block after.

The algorithm

Maintained per address, updated only when that address's balance changes:

// on every Transfer touching address a, at block timestamp t
cum[a] += bal[a] * (t - last[a]);
last[a] =  t;
bal[a]  =  newBalance;

// at the close of orbit [T0, T1], after flushing every touched address to T1
TWAB(a) = (cum@T1[a] - cum@T0[a]) / (T1 - T0);

This is O(transfers), not O(holders × ticks) — the indexer does no work for an address whose balance did not move. At BNB Chain's current 0.45-second block time ^[4] an orbit spans roughly 8,000 blocks, so the distinction matters.

The split

Q       = { a : TWAB(a) >= 100,000 and a is not excluded }
share(a)= TWAB(a) / Σ TWAB(q) for q in Q
credit(a) = pool_usdt * share(a)

Because the denominator sums only qualifying addresses, weight held below the floor or by excluded addresses is redistributed automatically to holders who do qualify. There is no separate redistribution step.

Finality

An orbit is never settled before its closing block is final. BNB Chain provides deterministic fast finality within a small number of blocks, so the delay between an orbit closing and its ledger rows appearing is seconds rather than the multi-minute or multi-day windows an optimistic rollup would impose.

05

Eligibility & exclusions

The floor is 100,000 $ATERA on time-weighted average across the orbit. Falling below it mid-orbit forfeits that orbit; topping up just before the close does not buy in, because the average already reflects the hours spent below.

Excluded addresses

Smart accounts

A wallet delegated under EIP-7702 — live on BNB Chain since the Pascal upgrade ^[5] — still has a private key and a person behind it, and is treated as an ordinary holder. Because such an account returns non-empty EXTCODE, a naive "has code → is a contract → exclude" check would wrongly exclude it. The correct test is the delegation's exact shape:

// EIP-7702 delegation indicator: exactly 23 bytes
code = eth_getCode(addr)
isDelegatedEOA = len(code) == 23 and code[0:3] == 0xef0100
// anything else with code is a real contract → exclude

Signature verification follows the same split: a plain EOA is checked with ecrecover over a SIWE message, and a smart account falls back to EIP-1271 isValidSignature.

06

Transponders

A transponder is an ordinary OpenRouter API key with a spend ceiling, provisioned through OpenRouter's key-management API ^[6]. There is no ATERA SDK and nothing to install — anything that already works with OpenRouter works unchanged.

Claim flow

1. Prove ownership of the wallet by signing a SIWE message (EIP-1271 fallback for smart accounts).
2. The service reads accrued − granted from the ledger to get the claimable balance.
3. It raises the existing key's ceiling: limit = spent + min(balance, 200).
4. Anything above that tranche stays in the ledger, still owed, rather than idling on a credential.

Reconciliation

A holder's claimable balance is always derived from the ledger, never from the key. The key's reported spend is used only to know when a tranche is exhausted. Balance is Σ credited − Σ granted; if those two ever disagree with OpenRouter's reported totals, the ledger is authoritative and the discrepancy is alerted rather than silently reconciled.

Privacy

Requests go directly from the holder's machine to OpenRouter. Atera provisions keys and reads their aggregate spend totals; it does not proxy traffic and has no access to prompt or completion content.

07

Architecture

Distribution deliberately does not happen on-chain. An on-chain loop over thousands of holders costs gas proportional to holder count, prices small holders out of their own payout, and is a denial-of-service surface. A ledger row costs nothing and settles in the same second the orbit closes.

08

The token and the tax

The token

• Fixed supply. 1,000,000,000 $ATERA. No mint function exists.
• A 1% tax on trades, and nothing else. No blacklist, no pause, no max-wallet. Plain wallet-to-wallet transfers are untaxed.
• No owner key over the tax. The rate and its destination were written on-chain at creation by four.meme and cannot be modified by any address, ours included.
• Not upgradeable. No proxy, no delegatecall to a mutable implementation.

Where the tax goes

• The tax routes to the collector, which converts it to USDT and splits it 30/70 at each orbit boundary.
• Liquidity sits in a PancakeSwap V2 pool created automatically at graduation, with the LP tokens burned so it cannot be withdrawn ^[3].
• Because the token is untaxed, this is an ordinary concentrated-liquidity pool. Nothing about $ATERA trips the fee-on-transfer incompatibility that forces taxed tokens onto legacy V2 routing ^[7].
• Exclusion of pool, locker, collector, burn and exchange addresses happens in the indexer, off-chain. It affects payout eligibility only and can never affect anyone's ability to transfer or sell.

09

Risks & limits

• Accrual is volume-dependent and can go to zero. 0.30% of nothing is nothing. At $1M of daily volume the entire holder pool is $4,500/day across all holders; a wallet with 0.1% of qualifying weight earns about $4.50. Launch volume decays, and the model should be judged on sustained volume, not on day one.
• Launchpad dependency. Collection is four.meme's contract, not ours. That removes our own audit surface but makes the fee stream dependent on their infrastructure continuing to operate as documented.
• Nothing accrues pre-graduation. See The token and the tax.
• Scanner labelling. Token scanners will correctly flag $ATERA as carrying a 1% tax. This is accurate and should be expected rather than argued with.
• Concentration risk on the credit provider. All credit sits in a single OpenRouter account. An action against that account affects every holder's balance simultaneously.
• Exchange-held balances earn nothing. Covered in Eligibility; it is a product limitation, not a bug.

10

References

Third-party figures quoted in this document, with sources.

1. four.meme fee structure — 1% trade fee; creator-fee mode scaling 0.02%–1% with market cap; on-chain royalty layer. four-meme.gitbook.io/four.meme/guide/how-it-works
2. PancakeSwap V2 fee split — 0.25% total: 0.17% to LPs, 0.0225% to treasury, 0.0575% to CAKE buyback and burn. docs.pancakeswap.finance/trade/pancakeswap-exchange/fees-and-routes
3. four.meme bonding-curve migration to PancakeSwap at approximately 18 BNB. four-meme.gitbook.io/four.meme/guide/how-it-works
4. BNB Chain block time reduced to 0.45s (Fermi upgrade) with strengthened fast-finality rules. bnbchain.org — Fermi hard fork
5. EIP-7702 enabled on BNB Chain via the Pascal hard fork. bnbchain.org — Pascal hard fork
6. OpenRouter key management API — /api/v1/keys, per-key limit and limit_remaining, usage reporting. openrouter.ai/docs/features/provisioning-api-keys
7. PancakeSwap V3 does not support fee-on-transfer tokens; swaps revert with IIA, guidance is to use V2. developer.pancakeswap.finance/contracts/v3/faq
8. four.meme Tax Token Mode — on-chain transaction tax, rates fixed at creation, proceeds routable to wallets, dividends, burn or liquidity; rules immutable after creation. Four.meme introduces Tax Token Mode
9. four.meme graduation deploys a PancakeSwap V2 pool automatically and burns the LP tokens. chaincatcher.com — liquidity pool adjustment