The economics of ownable compute.
Three units, each with one job, built around a fleet GHOSTMETAL intends to own and operate: credits pay for usage, $GM owns the network, $CMP turns compute into a claim you hold directly.
Three layers, never confused
Credits — usage
$1 = 100 credits, flat. Pay-as-you-go for inference and the API. Not a token — just prepaid usage.
$GM — the network
Capital token. Stake it to own a pro-rata share of daily fleet capacity and a share of declining emissions. Revenue buy-and-burn is designed to offset emissions — net-deflationary only once buybacks outpace them.
$CMP — compute
Claimed daily by $GM stakers, pro-rata by stake weight. 1 $CMP targets ≈$1/day of inference while the network operates. Non-transferable on the MVP — a claim on capacity, held by you (see risks).
$GM supply & distribution
100,000,000 $GM100M genesis supply. A declining, staker-directed emission (≈12M over 3 years, below) is offset by revenue buy-and-burn. Insider & treasury allocations are subject to lockup / vesting (TBD).
Where all 100M ends up once everything has unlocked — the destination, not what exists at launch. 45% is liquid on day one (the amounts marked below); the remaining 55% arrives on the curve above, starting after a three-month cliff.
- 25% Stakers · 25,000,000 $GM10,000,000 $GM liquid at TGE · unlocks on the curve; raises the sGM share price
- 23.75% Community & airdrop · 23,750,000 $GM15,000,000 $GM liquid at TGE · earned via usage + referrals → TGE airdrop
- 15% Treasury · 15,000,000 $GM10,000,000 $GM liquid at TGE · development and growth (lockup/vesting TBD)
- 8.5% Development · 8,500,000 $GM5,000,000 $GM liquid at TGE · unlocks on the curve; engineering and infrastructure
- 7.75% Liquidity · 7,750,000 $GM5,000,000 $GM liquid at TGE · 5% liquid at TGE for DEX / CEX depth; the rest unlocks on the curve
- 20% Early investors & team · 20,000,000 $GMown vesting contract, not the curve: nothing for 6 months, then straight-line over the next 12
Unlock schedule
Circulating $GM over time. The grey base is what is already liquid at launch; the coloured bands are the locked remainder arriving on the curve, stacked by destination wallet. Hover anywhere for that epoch's figures.
The grey base is the launch float — liquidity plus the first airdrop tranche. It was never locked, so it never unlocks; the coloured bands are the locked remainder arriving on the curve.
Why a hyperbola, and why k = 238
We do not start at zero. 45,000,000 $GM — 45% of supply — is liquid the day the token launches, spread across five buckets so nobody waits on a cliff to stake, trade or be paid. The other 55,000,000 is locked, and nothing at all unlocks for the first three months. After that cliff the remainder releases on a curve. The obvious form E(e) = E₀/(1+k·e) is the harmonic series: it diverges, so cumulative unlock would be unbounded, which is impossible against a fixed supply. The schedule is therefore defined on the cumulative side as a rectangular hyperbola C(e) = B·x/(x+k), where x counts epochs since the cliff. Per-epoch unlock is its difference, decaying as ~B·k/x², while C(e) approaches B and can never exceed it. Boundedness is structural, not a check someone has to remember.
C(cliff + k) = B/2 exactly, so k is the half-unlock point measured from the cliff. B is 35,000,000 — the locked remainder MINUS the 20,000,000 investor and team allocation, which is not on this curve at all but in its own vesting contract. k = 238 epochs: half of B is out at epoch 328, about eleven months after TGE. Year one is only 275 epochs of curve because of the 90-epoch cliff, so it releases 35,000,000 × 275/(275+238) = 18,762,183 $GM; stakers take 42.86% of every unlock, which is 8,041,472 $GM in year one. The coefficient is not shown on the chart and is not selectable — the page plots the one schedule the protocol implements, not a menu of options.
Release per epoch
The hyperbola itself: K(e) falls about 115× over ten years and never reaches zero.
How a release happens
Every N blocks
One epoch = 43,200 Base blocks ≈ 24h at 2s per block. The same clock the $CMP epoch runs on, so both schedules tick together.
K coins unlock
K is not fixed — it is the difference between two points on the cumulative curve, so it falls every epoch. It is exactly zero for the first 90 epochs, ~75,200 $GM the day the cliff ends, ~39,800 a year in and ~13,300 at year three.
Split across Z wallets
Z = 5. The split happens at release time in one transaction, so no wallet can front-run its own allocation.
Anyone can poke it
drip() is permissionless and the amount is pinned to the curve, so calling it early, late or twice cannot change the total. A keeper calls it on a cron.
Where each release goes
Applied to every epoch's K(e), in one transaction at release time.
- 42.86% Stakersraises the sGM share price for everyone staked
- 25% Communitygrants, bounties, ecosystem incentives
- 14.28% Treasuryrunway and protocol-owned liquidity
- 10% Developmentcore engineering and infrastructure
- 7.86% Liquiditydeepens protocol-owned DEX / CEX liquidity over time
Governance
Voting power is staked $GM, not held $GM. sGM is already checkpointed for the $CMP claim, so the same balances that earn compute also carry the vote — you cannot buy a vote for one block.
the Z split percentages · the daily $CMP pool · treasury spending · pausing · which contracts hold roles
total supply (immutable, no mint function exists) · the 12M emission budget · the curve coefficient k — these are fixed at deploy so no vote can inflate the token
Model and page are implemented; the on-chain EmissionController still runs the older piecewise-linear schedule and the Governor is not deployed. Both are tracked before any value-bearing deploy.
Declining emissions → stakers
Emissions are 100% staker-directed and step down each year. Plain staking earns the full share; stake locked to mint $CMP keeps 80% (20% is protocol revenue). Not interest or a guaranteed return.
- TGE float45,000,000 $GM · 45%
- B (curve)35,000,000 $GM · 35%
- vesting20,000,000 $GM · 20% · 180+365 epoch
- cliff90 epoch ≈ 3 mo
- k238 epoch ≈ 8 mo
- N43,200 blocks ≈ 24h
- Z5
Buy-and-burn → deflation
We route 20% of platform revenue into buying $GM on the open market and burning it — permanently removing supply.
The value loop
Acquire & stake $GM
Own a share of the fleet + earn yield.
Lock to mint $CMP
Later minters lock more $GM (illustrative curve).
Draw compute
≈$1/day each while the network operates — spend or hold.
Burn to exit
Burn $CMP to recover your $GM principal.
The mint cost rises along an illustrative curve as $CMP supply grows, so each new unit of tokenized compute locks more $GM out of circulation while demand is high. This is a supply mechanism, not a prediction about token price.
Why it holds together
Designed for owned-GPU backing
Each $CMP is a claim on inference from hardware GHOSTMETAL operates, and today every request is served on rigs we own — no third-party provider sits in the path. What remains a roadmap target is the SIZE of that fleet: the capacity needed to back $CMP at scale, not whose metal runs it.
Revenue → buybacks
Usage is designed to fund buy-and-burn, linking buybacks to platform revenue rather than emissions alone (illustrative — revenue is not live yet).
Supply locks with demand
Staking and the exponential mint curve pull $GM out of circulation exactly as compute demand rises.
Oversell guard
A capacity governor is designed to cap total minted $CMP at a fraction of sustainable throughput, so daily allowances are intended to clear under normal conditions.
Capacity governor: total $CMP daily draw is designed to be capped at ~35% of sustainable fleet throughput so allowances can clear — a design intent, not a guarantee.
Risks — read them
- — This is a speculative, growth-dependent design. If the platform stalls, the real value of daily inference per staked $GM falls.
- — Token markets are volatile; nothing here promises price, yield, or liquidity. The exponential $CMP mint curve favours early participants.
- — Everything on this page is pre-launch and illustrative — parameters can change before (and after) any token exists.
Not financial advice. Not an offer or solicitation to buy any security or token.