The OGONG L1

OGONG runs its own sovereign Layer 1. It is a Substrate chain whose consensus resource is not staked capital but verified inference work: the thing a validator must produce to get a seat is the same thing the network exists to produce. This page is about that chain. Where the rest of these docs mention staking, they are describing the EVM lane, which is what the client settles against today.

Zero bond, everywhere

There is no staking pallet, and no bond of any kind, for any role. That is not an omission to be filled in later; it is the design:

• A seat is earned by having work verified, and it is lost by ejection. A validator that issues false verdicts forfeits its seat and every future emission that seat was worth.
• Nothing is ever confiscated, because nothing was ever posted. The penalty in a zero-bond protocol is always a forfeited earning, never a seized balance. A slash has no referent here: there is no account to take it from.
• That makes the barrier to securing the chain a GPU, not a treasury. Capital cannot buy consensus weight on this chain at any price, because there is no mechanism that sells it.

The reason this works is the verification result: because checking a reply costs a few percent of producing it, nearly every reply is checked, so a cheat is caught essentially every time and the forfeited fee is deterrent enough on its own. See Why there is no correctness bond.

Consensus

BABE for authorship (VRF-elected slots, 2-second blocks) and GRANDPA for finality. The one change from a stock Substrate chain is where authority weight comes from: a validator’s GRANDPA weight is its recently verified work, computed at a scale of 2^32 and refreshed every session.

That weight is deliberately linear in work. A cap or a floor would both be concave, and any concave weighting pays an operator to split one machine into several identities, so the honest big operator subsidises the sybil. Linearity is the only shape under which splitting an identity is exactly neutral.

Work also decays. Each identity’s recent work halves every 24 eras (one day at hourly eras), computed on read rather than by sweeping storage. A validator that stops contributing slides out of the set on its own, so a seat has to keep being earned rather than being held on the strength of past performance.

The committee is bounded; mining is not

GRANDPA messaging is quadratic in the committee size, so the set that finalises blocks has to be bounded. It is bounded at 300 seats, and the ranking is by decayed recent work.

Mining is not bounded. Anyone may earn and claim emission without ever holding a seat, and the great majority of earners never will. The 300 is a limit on who finalises blocks, not on who may do work and be paid for it. Contenders for the set are tracked in a bounded index of 2,048 so that the per-era recompute stays cheap regardless of how many identities are mining.

Emission

A hard cap of 5,000,000,000 OGONG, all of it earned, on a Bitcoin-style halving schedule with an 8-year period (70,080 hourly epochs). There is no premine and no genesis allocation of any kind: supply at genesis is zero. The opening ceiling is derived from the cap rather than chosen (a halving series sums to twice its first period, so the curve exhausts the pool exactly instead of overshooting or undershooting it), which puts it at about 35,674 OGONG per epoch: at most 312.5M in year one, half the cap at the year-8 halving.

Emission is priced per epoch, and this is the part worth understanding. Each epoch the chain divides that epoch’s reward by the work done in that epoch, and records the resulting rate. It does not divide by the chain’s all-time work total. The difference is the difference between paying for work being done and paying a perpetual dividend on work having once been done: under the second, an identity that contributed early would keep drawing from every future epoch forever. Rates are kept for 2,160 epochs (90 days), which is therefore the deadline for claiming what an epoch owes you.

A bootstrap floor pays provably ready capacity before real demand exists, so miners earn from day one rather than waiting for inference volume to arrive. It starts at half the opening ceiling and tapers linearly to zero across the first year (8,760 epochs). It is drawn from the same finite tranche as everything else, so it front-loads emission rather than adding to it.

The capacity challenge

The floor pays provably ready capacity, and the proof is a capacity challenge: a piece of synthetic inference the chain asks an identity to perform each epoch. Three properties make it worth anything.

• Beacon-derived. The challenge is computed from randomness settled one epoch earlier, so nobody can know it before the epoch opens, precompute answers, or rent a GPU only for windows it can foresee. Renting continuously costs the same as owning, which is the point.
• Identity-bound. Two identities are never asked the same question, so one answer never satisfies two of them. An operator running N identities owes N different answers inside the same deadline, which takes N machines, not N times the patience. This is the leg of sybil resistance that weighting alone cannot provide: linear weight makes splitting neutral, and the challenge makes it expensive.
• Deadline-bound. Fresh each epoch and each index, so nothing replays forward.

The chain does not take a submitted challenge id on trust; it recomputes it and refuses anything else. Anyone can compute anyone’s challenge, and a verifier has to in order to know what it is scoring. The secret was never the question, it is the capacity to answer it in time.

Nor does the chain take anyone’s word that a challenge was answered. A pass is attested by a work-weighted supermajority of seated validators, the same quorum that resolves a dispute, accumulated one vote at a time on chain. There is no privileged origin on this path and root cannot record a pass: a readiness credit mints emission, so a single trusted issuer would have been a single party able to mint, and able to hand a passing grade to every identity a sybil operator cared to register.

A passed challenge credits work, in the same unit audited inference is credited in, so it lands in the same per-epoch denominator and the floor divides across it. That is how an operator earns in week one with no consumer in sight. The credit is scaled by the floor’s own taper, so it is worth a full challenge at genesis and nothing once the bootstrap year is over: after that, answering challenges still holds your seat but no longer pays, because a credit that never expired would be a permanent subsidy for being switched on, funded by diluting everyone doing work someone actually paid for.

Status: the derivation, the crediting, and the taper are implemented. The gate that refuses a seat to an identity that has not answered is built but switched off, because the off-chain responder that answers challenges does not exist yet, and enforcing it before then would empty the validator set.

Settlement

Settlement on the L1 accumulates on chain rather than gathering signatures off it. A settlement is opened, each validator casts its vote as its own extrinsic, and the settlement finalises when the accumulated weight clears a work-weighted supermajority (more than two-thirds of the era’s total active work; a quorum below that floor is rejected outright, and a quorum of zero is rejected outright).

Two consequences of that shape:

• Voting is free. cast_vote pays no transaction fee. A validator voting on settlements is doing the network’s work, not consuming its capacity, and charging it per vote would make the marginal seat cost more than it earns.
• Spending is authorised by allowance. A consumer admits an allowance once, and settlements draw against it. The allowance is additive and is not revocable mid-flight, so a settlement cannot be opened against funds that vanish before it finalises.

Pallets

registry (identity and discovery), pouw (work accounting, decay, and the validator set), readiness (the capacity challenge), calibration (the drift thresholds a verdict is measured against), settlement (escrow and the quorum), disputes (challenges to a verdict), and emission (the schedule and claims). There is no staking pallet.

The two lanes

The L1 carries an EVM lane of its own at chain id 2030, with AccountId20 addresses and Ethereum signatures, so Solidity tooling works against it directly.

The older deployments are superseded. OGONG was built twice before this: a Solidity contract set on Robinhood Chain, and a Solana program set. Both worked and neither is the direction. The sovereign L1 replaces them, and because it carries its own EVM lane there is nothing the Robinhood deployment offered that the L1 does not: the same Solidity, the same tooling, on a chain whose consensus is verified work rather than rented capital.

Those older contracts are why staking still appears in these docs. They do have a staking pool, because they were written before the L1 existed and stake is how they weight a quorum. They are also still what the client settles against today, so the operator pages under Run a node describe them and are accurate about what you would run right now. Migrating the client onto the L1 is roadmap work; until it lands, those pages describe a lane on its way out rather than a second design being maintained alongside this one.

See also Network roles and Tokenomics.