ABYSS
Field Manual

Two systems, deliberately separate

ABYSS is two machines that share an aesthetic and almost nothing else.

The curve issues 3,333 permanent creature identities on a linear ETH price ladder and buys them back on the same ladder. It works with zero assumptions about player behaviour and carries no debt it cannot pay.

The dive is a press-your-luck game played with those creatures. A creature descends on its own once started; the owner only chooses the moment to stop. Surfacing successfully mints FATHOM. Surfacing unsuccessfully mints nothing.

The curve

Price is a straight line from the first entry to the last: P(n) = 0.0005 + (n / 3332) × 0.0045 ETH. A batch sums each marginal step rather than multiplying the first quote, and an exit is quoted from the same steps in reverse. Twenty per transaction, 9.166 ETH if every identity is claimed.

Where the ETH goes

Sell-side release totals 85% — exactly what the buy side funded at the same step — so the reserve is solvent by construction. The last exit is as covered as the first. There is no creator-seeded DEX pool and no liquidity anyone can pull.

Creatures are never burned

An exit moves the token into the vault. The next buy pops it back out: same ID, same seed, same traits. A collection of 3,333 identities stays a collection of 3,333 identities forever. Selling also clears what the previous owner did to it — any dive in progress is forfeited, and callsign and palette reset.

The creatures

Art seeds are committed at mint and revealed from a later block. This is not decoration. If the seed came from state that already existed while the buy was executing, a contract could read the traits it was about to receive inside onERC721Received and revert the whole purchase until it got a roll it liked, paying only gas. That attack is in the test suite; it now fails at the first step.

reveal() is permissionless — anyone can reveal anyone's token, and there is nothing to gain by holding one back. Past the 256-block window the fallback mixes a rolling accumulator with the previous block hash, so reveal promptly. Royalties are ERC-2981 at 5%.

The dive

A creature reaches the hadal floor 24 hours after it starts diving. No transaction pushes it deeper and nothing needs poking: depth is a pure function of elapsed time.

The two curves

Harvest is the integral of the rate curve over depth, normalised so a full descent yields 100 FATHOM in the first emission era.

Why the bottom is not the answer

Riding to the floor returns 79% of what stopping at the peak returns. The bottom is a gamble, not a reward. There is a second layer to it: under repeated play, expected value per hour peaks shallower still, around 2,500–4,000 m. Absolute maximisers go deep; throughput maximisers farm the middle.

Every depth carries a roll. There is no depth at which extraction is free, which is what stops any rig from becoming a risk-free printer.

Rigs

A rig is FATHOM burned before the descent. Each multiplies harvest and raises survival to a power, so the penalty bites hardest exactly where survival is already thin.

The drill costs little survival where survival is already high, so it wins the middle water. The seal buys survival back where it is scarce, and is the only sane way to work the hadal.

Settlement

Surfacing is two steps. requestSurface(tokenId, minDepth) locks the depth, harvest and survival, and escrows a resolver tip; minDepth stops a delayed transaction surfacing you somewhere you did not choose. Then resolveSurface(tokenId) reads the next block's hash and settles. Anyone may call it and collects the tip for doing so.

The tip exists because the outcome is public before settlement. Without it, a player who read a losing roll would simply walk away, leaving the token stranded and the crush unrecorded. With it, keepers clear the queue within a block or two.

FATHOM

Minted in exactly one place: a surviving dive. Nothing else — no presale, no team allocation, no admin mint. Of every harvest, 88% goes to the diver and 12% to the project treasury. Burned in three places: redemption, rig costs, and palette mutations.

Emission halves, then stops

Eras are measured in FATHOM actually minted, not in time, so a quiet month does not burn through an era and a busy one does not outrun the schedule.

After the eighth era emission stops and startDive refuses. A halving is neutral for the player — half the FATHOM, each worth about twice as much. Its point is that fees keep arriving while supply growth slows, so each halving doubles what one expedition fee buys per FATHOM minted: measured at 4,167 gwei in era 0 and 8,335 gwei in era 1. The floor rises with the schedule instead of drifting sideways.

The pool

There is no market maker and no seeded pair. Every FATHOM is a pro-rata claim on one ETH pool: floor = poolBalance / totalSupply.

redeem() burns before it pays, so the ratio can only improve for whoever stays, and a 5% exit fee is left behind for them. The pool cannot be insolvent because it only ever pays a fraction of what it holds.

Only expedition fees fund it

This one is worth spelling out, because getting it wrong is expensive. ETH that arrives before any FATHOM exists belongs, pro rata, to whoever mints first. With NFT trading fees flowing into the pool, a single early diver could redeem most of a pot that hundreds of buyers had filled.

The ratio is now self-consistent from the very first dive, with no first-mover ramp. And a dive that fails pays its fee while minting nothing — that ETH stays in the pool and lifts the floor for everyone who did surface. That, not emission control, is what holds the token up.

The money map

Gameplay is not a revenue centre. The project takes nothing in ETH from a dive. It is paid out of the harvest instead, in FATHOM, so its revenue rises with the floor and with participation rather than with how often a player loses.

That FATHOM inventory is also what pairs with the liquidity treasury's ETH if a Uniswap pool is ever opened — no presale, no genesis allocation, no founder capital.

What the numbers mean

For a player

A dive costs 0.0001 ETH. Ninety per cent lands in the pool the harvest is redeemed against and ten per cent pays whoever settles the roll — which can be the player themselves. Across many dives a player recovers roughly 83% of what they spend; the remainder is the tip and the exit fee, both of which stay inside the player economy. There is no house edge taken in ETH. What separates a good player from a bad one is where they stop.

For a holder

The floor is the pool over the supply. It rises when dives fail, when FATHOM is burned, and with every halving. It does not rise from new buyers paying old ones, because there is no mechanism by which they could.

For the project

Two legs, deliberately balanced: ETH from the NFT curve, FATHOM from the harvest tax. Neither depends on players losing.

Build status