FAQ: Tokenomics & Supply

How are block rewards and inference fees split?

Block reward: 96 $FLOP per one-second block, halving every 730 days for the first five halvings, then a constant tail in perpetuity.
In-block split: 75% miners / 10% validators / 10% agents and brokers / 5% staking rewards.
On top of the 96, FLOP Labs and the FLOP Foundation each mint 8 $FLOP per block — an additive mint, not an in-block share — for 112 gross in years one and two. It halves on the same schedule and sunsets after year ten, totalling 10.8% of year-10 supply (5.4% each).
Inference fees: 85% miner / 15% validator. Neither FLOP Labs nor the FLOP Foundation takes any cut of the inference fee, and nothing is burned out of it.
Chain transaction fees are a separate instrument and are not burned either. Nothing is burned anywhere in the design: not block rewards, not inference fees and not transaction fees.
Miner share, stated by base: 75% of the in-block reward; 48.6% of year-10 supply from block rewards; 55.3% of year-10 supply including the miner airdrop. Always say which base.
Worked example, one $100 inference order arriving through a broker: the customer pays the broker $100 in dollars or stablecoin; the broker buys $FLOP on the open market and funds the session escrow; of the $FLOP the session actually consumes, 85% is paid to the miner that served it and 15% to the validators, both liquid as paid. The broker’s income is the spread between the $100 it billed and the $FLOP it fronted, plus its share of the 10% agent-and-broker portion of the block reward; the protocol keeps nothing. Block rewards are a separate stream — 96 $FLOP a block, split 75/10/10/5 and divided pro rata to verified work — so the miner that served the order earns the fee plus its pro-rata block reward for the compute delivered. FLOP Labs and the FLOP Foundation take nothing from inference fees; their only income is their block-reward allocation. An operator running a broker and miners together earns exactly the same on work it routes to itself as on work it serves for anyone else — the only external revenue in the chain is the $100 the customer paid, and 15% of every fee it pays its own miners goes to validators it does not control. On testnet the miner airdrop is credited on verified compute actually delivered, measured in compute units rather than in fees paid, so overpaying your own miner earns nothing extra; a ceiling on inference credit relative to block-reward credit is being finalised. Each role is counted separately: a wallet that mines, buys compute and validates on testnet earns a separate voucher for each role, and each converts under its own rule at genesis (miner earn-to-unlock, agent spend-to-unlock), even if one wallet holds all of them. Keeping the roles on separate keys is advisable for tracking, not a requirement.

Is there a max token supply? What is long-run inflation, and is an emissions-funded block reward healthy long-term?

There is no hard cap in the Bitcoin sense; supply is set by the emission schedule.
Total supply at end of year 10: 18,086,624,000 $FLOP. Composition: airdrop pool 4,400,000,000 (24.3%); miners’ block rewards 8,798,544,000 (48.6%); validators 1,173,139,200 (6.5%); brokers and agents 1,173,139,200 (6.5%); Labs + Foundation 1,955,232,000 (10.8%); staking rewards 586,569,600 (3.2%).
Block reward starts at 96 $FLOP per one-second block and halves every two years, five times. After the fifth halving a constant tail of 3 $FLOP per block continues forever, which is 94,608,000 $FLOP a year; annualised inflation from that point is about 0.5% per year.
Why the halving is every two years rather than Bitcoin's four: compute supply grows much faster than Bitcoin's did, so a shorter cycle keeps the block subsidy declining in step with the growth of the network rather than years behind it.
On burns, two different things get conflated. Inference fees are not burned at all — they split 85% to the miner and 15% to the validator, with no platform take. Transaction fees are separate and are not burned either; nothing is burned anywhere in the design.
Why a perpetual tail rather than a pure fee model: a fee-only security budget is procyclical — fees dry up in downturns exactly when the network is cheapest to attack. A small permanent floor keeps miners and validators paid through every demand trough, guarantees baseline compute availability, and its cost to holders rounds to almost nothing.
$FLOP is a compute commodity, not money; some inflation accommodates growth in participants. The emission schedule and the airdrop pool are fixed in the open-source runtime at genesis; nobody can mint outside them. The tail rate, like any protocol parameter, can only change through an on-chain protocol-upgrade referendum — 67% approval of votes cast, a support floor that starts at 50% of active issuance and falls to 15%, and a 14-day enactment timelock.
Labs and Foundation holdings count as circulating from mint.

Are you fundraising? How can a fund actually get exposure — including a fund whose mandate does not allow it to own or operate hardware?

There is no raise, and structurally no room for one. There is no token round and no equity round: FLOP Labs is funded directly by Arthur Hayes as its equity backer, with more than two years of runway assuming zero revenue, so there is nothing to raise into and no cap table to join.
No tokens are pre-mined for the team or for investors, and there is no team or investor allocation inside the genesis pool. The team-side allocations are the per-block mints to Labs and the Foundation (8 $FLOP each on top of the 96, 10.8% of year-10 supply combined, 5.4% each), liquid on emission like every other block reward. Separately, the 4,400,000,000 $FLOP genesis pool carries an 800,000,000 $FLOP ecosystem reserve held by the Foundation — 20% released at genesis and 10% every six months thereafter, fully released after four years — earmarked for creator and referral payouts, builder and security programmes and initial two-way liquidity at listing. It is disclosed, it is not a team or investor allocation, and it is the only balance the Foundation holds on day one.
For a fund, that leaves exactly two routes and there is no third: rent or deploy compute and mine or validate through testnet into mainnet, or acquire the token on the open market once it trades at mainnet listing. There is no allocation that can be negotiated.
The participation path is Bitcoin-OG style: become a miner — rent an H100-class unit for roughly $1,800–2,500 a month, or deploy hardware you own — and/or a validator. Starting in testnet is what captures the airdrop.
A mandate that rules out owning or operating hardware is not a blocker, because renting is the preferred route anyway. Nothing goes on the balance sheet and there is no data-centre relationship: the unit sits in the provider’s facility and you take root access. Setup is two to three hours of work, after which it is passive collection. FLOP will advise on which rent-or-buy platforms to use and help with the setup, and a vendor list with live availability is being published alongside setup documentation and an agent-readable manual.
FLOP does not operate, sponsor, approve or verify any pooled vehicle that buys or rents miners on investors' behalf — a FLOP-branded 'yield co' or royalty-style passthrough would make FLOP a fund manager with fees, carry and obligations it does not want, so there is no FLOP-approved vehicle and FLOP will not endorse one set up by anyone else. An investor who wants mining exposure without operating GPUs engages a managed service provider directly: rent or buy the units, and the operator hosts, runs and manages them for a fee, with the investor keeping control and the return. FLOP will introduce investors to such operators and will talk to any investor directly about how to structure participation; it does not take investor money.

Are block and inference rewards vested or locked? How does the airdrop work for miners, validators and agents?

Block rewards and inference fees are 100% liquid the moment they are emitted, for every recipient. No lockups, no vesting cliffs.
Airdrop pool: 4,400,000,000 $FLOP, 24.3% of year-10 supply, earned through testnet participation. Split: miners 1,200,000,000 (6.6%); agents 1,200,000,000 (6.6%); validators’ aggregate bond 1,200,000,000 (6.6%); reserve and incentives 800,000,000 (4.4%). The whole pool is minted at genesis; 3,760,000,000 is distributed at TGE and the undistributed reserve remainder sits locked at the reserve address.
Miners: 25% liquid on day one; the remaining 75% unlocks earn-to-unlock at 1:1 — one locked token vests for every $FLOP that miner identity earns on mainnet, from block rewards and inference fees alike. What counts is $FLOP actually earned, not registered capacity or uptime, so a miner that reduces its fleet during the unlock earns less per month and the unlock takes longer rather than anything being forfeited. Four months is the outcome of that rule at constant capacity, not a fixed calendar schedule: the unlock is driven by earnings, so it completes faster at higher delivered capacity and slower at lower. The locked balance can be bonded as the miner’s own stake while it unlocks; it stays frozen while bonded, and any slash is charged to it first.
Why unlock on use rather than on a calendar: a fixed vesting date is the date everyone circles to sell, and tokens with cliff schedules show the same chart. Usage-based unlocks give a holder of locked airdrop tokens a reason to use the network rather than wait, and all of this applies to airdrop tokens only — everything earned on mainnet, from block rewards or from inference fees, is liquid the moment it is earned.
A fair objection is that with no cap on participation, mainnet compute could be several times testnet compute, in which case each unit earns less and the unlock stretches well beyond four months. That is right as far as it goes. The counterweight is that the same condition is the one in which the network is working: compute arriving well above testnet levels means demand is pulling it in, which is what the market capitalisation is priced off, so the unlock runs more slowly into a token worth more.
The per-miner testnet airdrop is pro rata to verified compute delivered across the whole 90-day testnet, not a flat per-seat grant and not weighted to any single window — so a late start lowers the score for the full period.
Within the miner pool, the block-reward portion is earned by verified work actually delivered across the testnet — not by registered capacity, so a machine that takes no jobs earns nothing from it — and the inference portion is shared pro rata to verified inference work served, measured in compute units rather than in fees paid — so the whole miner pool is distributed whatever the level of testnet fees, and overpaying for your own inference earns nothing extra; a ceiling on inference credit relative to block-reward credit is being finalised. Both scale with work done, which is deliberate: paying on capacity would reward machines that sit idle.
Agents: locked, spendable only on compute, and unlocking 3:1 as spent on mainnet inference — every three $FLOP of locked balance spent on settled sessions releases one locked $FLOP. At most three quarters of an agent’s allocation can be spent on compute, which is exactly the spend that releases the remaining quarter; there is no time-based vesting schedule, and tokens that are never spent on inference stay locked.
Validators: the whole airdrop is the bond — 1,200,000 $FLOP per validator for the top 1,000 — granted free, fully locked for 24 months, then one randomly selected validator is released in full each day until all 1,000 are out, about 1,000 days.
How testnet work reaches you: testnet runs its own $FLOP-equivalent token so that paying for inference and staking behave exactly as they will on mainnet — but what accumulates as the record of your work is a voucher, not a balance of the eventual token. At mainnet the voucher is claimed against the relevant pool: the miner pool funded with the full miner allocation, and the same pattern for agents. Claiming is a free transaction. Vouchers rather than testnet token balances are the canonical record the airdrop is computed from, so moving testnet tokens around, or a testnet reset, does not change what you are owed. Each role earns its own voucher: a wallet that mines, buys compute and validates on testnet earns a separate voucher for each role, and each converts under its own rule at genesis even if one wallet holds all of them. At the close of testnet the record is frozen at a published block height, and before genesis the full allocation list and its hash are published so every participant can check its entry and dispute it within a set window. Miner vouchers are given claim priority over agent vouchers at launch so that miners have liquidity from the first day. For miners the block-reward and inference vouchers resolve into a single instrument on claim — locked miner $FLOP, 25% liquid immediately and the remainder on the earn-to-unlock rule above.
The unlock rules above are now written into the published specification: the miner quarter at genesis and earn-to-unlock credit, the agent 3:1 spend-to-unlock with its spend cap, and the validator freeze to the first halving followed by one release per day. Implementation on the chain is partial — the current vesting pallet still vests linearly and is being replaced — and three items remain open: the exact order in which validator grants are released, any cap on conversion scores, and what happens to miner or agent principal that never unlocks.
On the testnet-to-mainnet conversion ratio, which is the question underneath all of this: the total number of tokens credited at the genesis block is fixed and published, but the ratio a given participant converts at is not, and cannot be until testnet is over. The miner block-reward portion is deterministic, because the number of testnet blocks is known in advance, and is expected to convert close to one-for-one. The inference and agent portions float: the pools are fixed, any number of agents can claim from the faucet and spend, and more activity means more claims against the same pool. That is a feature rather than an omission — a fixed pool divided by measured work is what stops the airdrop rewarding registration instead of use — but it means no conversion formula can be published in advance, and anyone modelling a per-unit outcome should model the pool and their expected share of it.
Arthur’s Substack essay of 19 Aug described a 10-year vest for testnet participants; the model does not — the pool is across all cohorts and is released well within the first two years.

What are the staking requirements, and how do miners and validators source stake when there is no pre-mine? How does the 10,000 $FLOP figure work on testnet?

The miner stake — 10,000 $FLOP base plus a component that scales with the miner’s calibrated peak capacity in GFLOP/s — applies to mainnet only. The scaling component is still being finalised, so no per-card total is quoted yet.
During devnet and testnet the network sponsors the stake a miner or validator has to post, so nobody needs to buy anything and nobody needs to source $FLOP before the token exists; the cost of testnet is the hardware and the time to run it. Testnet tokens carry no value.
Mainnet stake is bootstrapped out of testnet participation: testnet block-reward earnings convert to real $FLOP at TGE, becoming stake and working capital.
Validators: the top 1,000 testnet validators receive the full 1,200,000 $FLOP bond free, granted rather than purchased. It is fully locked for two years; if a validator is culled from the set, the incoming validator assumes that bond. After two years, one random validator is fully unlocked per day until all 1,000 are out.
Delegation: there is no delegated staking. Validator stake is the granted 1,200,000 $FLOP bond and is self-stake only; miner stake is the operator's own. A $FLOP holder cannot back a validator's or a miner's stake for a share of its rewards, and no such position exists to buy, sell or custody. Any specification text describing delegation modules predates this position. The only way to earn on $FLOP without operating is the community staking lock described above, which involves no validator and no miner.
The bond is sized on a security floor rather than on affordability: the cost of corrupting the validator set has to exceed the value of corrupting it, which requires at least 1.1 million $FLOP per seat at genesis.
Everything earned on top of the bond — 10% of block rewards and 15% of inference fees — is liquid from day one.
A miner that took part in testnet does not need to buy stake at mainnet: 25% of its miner airdrop is liquid on day one, which for a participant of any meaningful size is comfortably above the stake for the same hardware. The stake is set per registered miner from its calibrated peak capacity: a fleet registered as one miner posts one stake sized on the whole fleet’s capacity, and a fleet registered as many miners posts a base plus capacity-scaled component for each — so there is no flat stake that covers any amount of hardware.
The locked 75% of a miner’s airdrop can be bonded as that miner’s stake while it unlocks, so a testnet miner does not need a second pool of liquid $FLOP to post collateral; the locked balance stays frozen while bonded and any slash is charged to it first.
Anyone entering after launch buys stake on the secondary market, which the miner float supplies. Stake is uniform — no extra stake for extra reward.

Zero pre-mine means very few holders at the start — doesn't that recreate early-Bitcoin volatility and concentration? What keeps $FLOP from concentrating in whales as the network grows?

Three structural differences from Bitcoin in 2009.
1. The product has proven demand and an external dollar price anchor — you can look up what an inference costs on OpenRouter today — so $FLOP's value is not purely reflexive.
2. Launch is preceded by a roughly three-month incentivised testnet and an airdrop of about 24% of the first decade's supply across miners, validators and agents. Day-one holders number in the hundreds and up, not single digits.
3. Distribution is the thing FLOP is least short of: Arthur's reach, the Token2049 announcement and institutional participants already in diligence mean both sides start warm.
On concentration: earned distribution is self-spreading. Tokens flow every second to independent operators and agents in proportion to work done, and each halving epoch redistributes to whoever does the work next — which no pre-sale cap table ever does.
The network is open and does not cap large operators, and does not need to: nobody got tokens cheaper than the work they did, so size buys no privileged cost basis. Mining share must be re-earned continuously against new entrants in a commodity business.
Who buys at the start: there is no sale and no allocation to buy from FLOP, so the first public-market buyers are investors who want exposure to the network and buy from miners selling into the spot market — miners must sell to cover hardware, which is why the first months after launch are when the token is cheapest — alongside agents and brokers that need $FLOP to pay for inference. Early on the first group dominates; as settlement volume grows, inference demand becomes the structural bid.

How does the referral programme work, how does it relate to the KOL / creator allocation, and when does a live referral link exist?

Referrals are single-tier and deliberately not multi-level. Validators cannot be referred. • The earlier commission structure (6–10% of a referred miner’s inference earnings, 10% of a referred agent’s spend) has been superseded. Creators are rewarded through a public leaderboard and a periodic referral lottery. • Sign-up: a verified email address, a $FLOP wallet and one or more social handles (X, Instagram, TikTok, YouTube or Naver), each verified by posting from the handle. Every creator receives a unique referral link, and every wallet created through that link is credited to the creator. • Leaderboard: published on the network explorer, showing for each creator the wallets created, the total $FLOP held, and the period-on-period change in $FLOP held, inference paid and HTLC escrow paid. During the testnet, ranking is based on the $FLOP that referred wallets spent on inference, received from inference and spent through the HTLC; at mainnet, on the net change in $FLOP held. • Payouts: each period the leading creators receive a payout in proportion to their share of the $FLOP held across the cohort, with the cutoff published each period. Payment follows what the referred audience does on the network, never views or likes, which can be manufactured. • Lottery: each period a fixed number of wallets that signed up through a referral link are drawn at random, and each receives an equal share of that period’s prize pool as a direct payment. Every referred wallet is entered in every period, so newcomers have a stake in each period without any rebate on inference spend. • Funding: the 800,000,000 $FLOP ecosystem reserve inside the airdrop pool (4.4% of year-10 supply), which also covers builder programmes and security-disclosure rewards. • Status: the number of creators paid per period and the lottery size are set from the observed distribution and published. The creator sign-up flow — an email address plus one or more social handles — and the referral link are not yet live; the KOL leaderboard page and sign-up are being published, reached through flop.finance, and the programme runs through testnet as well as mainnet. Rewards follow what referred wallets hold or spend on the network — engagement, clicks and views are never counted, because they can be manufactured — and anyone who converts an audience into $FLOP holders is eligible; there is no gating.
There are no bespoke deals for individual KOLs or agencies — no paid posts, no retainers and no private allocations. Every creator is on the same published referral terms.

Where and how do miners sell their $FLOP? What is the listing and market-making plan, and would you do a token loan to a market maker?

Token listing coincides with mainnet launch in Q1 2027, currently tracking to early February.
At launch: native trustless HTLC atomic swaps between $FLOP and USDT or BTC, built into the platform — no bridges, minimal attack surface. This is the intended primary venue rather than an external one: a miner with tokens to sell posts an offer and any counterparty can take it without touching a centralised venue, and no protocol fee is charged on those swaps. That traffic is public on-chain and a human-readable view will be published; the TechnoCore prototype already demonstrates the receipt pattern with no money involved.
Shortly after: one or two DEX listings and one or two CEX listings — a small number of venues by design, since the aim is for buyers to transact on the chain itself rather than to be everywhere. Tier-one venues may not list a project that pays nothing, and that is acceptable.
Native deposit and withdrawal support at a centralised venue is a three-to-four-month engineering project on the venue's side for a non-EVM chain, and FLOP is not building its launch around it: the on-chain HTLC swaps are the venue for anyone who wants to use the token, and a reference HTLC client with an agent-facing buy-and-sell skill is planned so that agents can acquire $FLOP directly without a centralised exchange. Venue integration timelines are the venue's to manage against its own listing date.
Custodian and venue integrations for a Substrate-based chain are scoped by the integrator from the public yellow paper; one qualified custodian has indicated it could integrate ahead of a February listing, subject to its own scoping, and custodian support brings deposit and withdrawal access at the venues built on that custodian's infrastructure.
FLOP does not pay to be listed. That covers technical or listing fees, marketing fees and security deposits, whether they are denominated in stablecoins or in tokens — a token-denominated marketing budget is a payment like any other. Where a venue proposes a package, the answer is the same.
No token loans to market makers, and no loan-plus-call-option structure. Market makers that participate in the ecosystem as miners or validators get priority for market-making and broker relationships; participation is the audition.
There is no separate “treasury token” and no treasury balance handed to a venue to backstop its book. The 800,000,000 $FLOP reserve inside the airdrop pool is earmarked for creator payouts, the referral lottery, builder programmes and security rewards; the Labs and Foundation share is a per-block subsidy that is liquid on emission, not a balance sheet that can be committed to an exchange. If a venue is told a treasury will backstop its book, that is not a thing that exists.
There is no external or paid market maker: no third-party desk is paid, and none is lent tokens or given a loan-plus-option to quote $FLOP. Initial two-way liquidity on the native HTLC venue comes from the project side — the Foundation's own token holdings, with Maelstrom's trading operations posting the bids and asks — not from an outside desk, and the aim is that an agent can download a skill and trade $FLOP natively without a Binance account or a Uniswap wallet.
The broker role doubles as a standing bid: brokers continuously buy $FLOP on the secondary market to pay miners, which supports two-way liquidity alongside miner selling.
The cold-start critique is accepted: two-way depth needs to exist before miners need to sell and agents need to buy. A concrete market-making plan follows closer to TGE. Where a market maker wants supply, FLOP can connect it with miners to work out offtake arrangements.
Nothing trades before genesis: no pre-launch listing, no pre-sale and no allocation anyone can buy from FLOP Labs or the Foundation. The only exposure to mainnet $FLOP before the genesis block is a testnet voucher bought peer-to-peer from a participant who earned it; anyone offering $FLOP for sale before launch is not selling something that exists.
Liquidity on the native HTLC venue is open to anyone: a trading firm can quote the $FLOP pairs with its own capital, posting bids and offers that any counterparty can take, with no appointment, no fee and no protocol stake. There is no appointed or paid liquidity provider, and quoting there earns nothing from the protocol beyond the trades themselves; a firm that also mines or validates has the standing that comes with participation.

Miners sell the moment they receive tokens and then hedge future rewards. Do you have any incentive or requirement for miners to hold the token, and isn't the miner stream structural sell pressure?

Yes, there will be sell pressure, and it is by design. $FLOP represents compute; miners provide compute, receive $FLOP, and need to cover costs and earn a profit.
Outside the airdrop, the only way to acquire $FLOP is to buy it from a miner — exactly as with Bitcoin. Miner selling is not a leak in the design; it is the distribution mechanism and the route to market for every agent and speculator. This contrasts with pre-sold protocols, where tokens enter supply through investor unlocks.
Ongoing emissions are wages, paid as earned, and are fully liquid. The testnet airdrop is a signing bonus for past work, and bonuses vest: 25% liquid at TGE, the rest earn-to-unlock — one locked token released per $FLOP earned on mainnet, which comes to roughly four months on the model’s assumptions — which smooths a one-time cliff.
There is no hold mechanism or hold requirement for miners. Hedging behaviour elsewhere follows pre-mine and insider leakage; FLOP has neither.

What is the opening price, the FDV, and the circulating and total supply at TGE and through the first months?

Published pre-TGE price: $0.10.
Circulating supply: about 459M at TGE; about 1.03bn at end of month one; about 1.6bn at month two; about 2.18bn at month three.
Total supply outstanding: 4.4bn at TGE — the whole genesis allocation is minted at block zero, of which about 3.94bn is locked; about 4.69bn at end of month one. Total supply at end of year 10: about 18.09bn, with terminal emission of about 0.5% a year thereafter. There is no hard max supply.
Because there is no single FDV basis, give all three and name each: at $0.10, circulating market cap is about $46M at TGE; about $440M on TGE-outstanding supply; about $1.81bn on year-10 supply.
A fair launch with a 459M float is a strength, not something to round off. Circulating and total are different bases: most of the 4.4bn is locked at TGE (validator bonds for 24 months, 75% of the miner airdrop until earned, the agent airdrop until spent), so always name which base a figure is on.
Growth in the float through the first months comes from the miner airdrop unlocking as mainnet earnings accrue, from block rewards, and from the reserve's seasonal releases — about 570M a month on the model's assumptions, so the float roughly doubles in month one and the rate of increase falls in proportion after that. By month twelve the model has about 5.5bn circulating against about 7.9bn outstanding — roughly 70% of the supply outstanding at that point, and roughly 30% of year-10 supply — so name the base whenever a float percentage is given. A 'high FDV, low float' reading is right for the first few months, and the mechanism that corrects it is the float growing, not the FDV shrinking.

You call this a fair launch with no VCs, but Labs and the Foundation each mint 8 $FLOP per block on top of the 96. How is that not a team allocation with extra steps? What is that stream spent on, and is any of it locked?

Yes, the team gets paid, and fair launch was never a claim that the team works for free. It means nobody got in cheaper than you: no pre-sale, no VC round, no advisor tokens, no insiders waiting to exit on the first pump.
The arithmetic is exactly as stated: 96 $FLOP per block splits miner 75 / validator 10 / agent 10 / staker 5, and Labs +8 and Foundation +8 sit on top as an additive mint — 112 gross — for the first five halvings, halving on the same schedule and sunsetting after year ten. It totals 10.8% of year-10 supply, 5.4% each.
The difference is the terms: the stream drips per block on the same public schedule as the miners, halves when theirs halves, and cannot be front-run. A team paid per second for a decade only does well if the network still matters in year ten — the opposite of an unlock cliff that lets insiders exit at month twelve.
Labs earns the same way as everyone else — through the token price. There was no pre-sale, there are no hidden allocations, and if the price goes to zero Labs makes nothing.
It is liquid on emission, like every other block reward; there is no lockup. The Labs stream funds development; the Foundation stream funds continued protocol development and upkeep on the Bitcoin Core model — paying competent developers to maintain the network.

About 51% of supply pays the compute supply side and roughly 7% pays the agents creating demand. If demand lags GPU supply, what stops $FLOP becoming another oversupplied compute token?

On the current model those shares are miners 48.6% and brokers/agents 6.5% of year-10 supply.
That share is not a subsidy for parked hardware. Block rewards track verified inference actually processed, so supply that nobody buys earns nothing and exits.
Because the only way a miner earns the block subsidy is by processing an inference request, and the subsidy is large, the miner's profit motive is to offer compute in the open market almost for free. The intended result is that FLOP is the cheapest compute available — including against OpenRouter — so a cost-minimising agent routes to it. That share is buying the price advantage that creates the demand.
The demand side pays with revenue, not allocation: agents bring fee income; miners bring capacity that has to be compensated.
The 10% agent/broker share of the block reward (6.5% of year-10 supply — two different bases for the same bucket) is emission spent recruiting buyers, on the same schedule that pays supply.
The anchor is external: compute is priced in dollars everywhere, so if supply runs ahead, FLOP compute gets cheaper in dollar terms and demand routes in. The subsidy halves every two years, and inference fees are designed to take over as the engine.

Does verifiable TechnoCore contribution count toward the agent allocation, or is it purely inference spend? Is TechnoCore the permanent coordination layer or a bootstrap that gets retired at genesis? Does anything before the testnet faucet count?

The published mechanic for the agent allocation is spend-to-unlock: create a DID, claim testnet $FLOP from the faucet when testnet launches, spend it on inference in the network, and that spent balance is what converts into mainnet $FLOP. If you don't use it, you lose it.
Creating DIDs or posting in TechnoCore rooms does not by itself earn anything. Coordination work between agents, and building connectors that bring demand into the network, is the behaviour the network exists to reward; there are discretionary buckets within the allocation for this, and how pre-testnet contribution is recognised will be published before the snapshot.
TechnoCore is permanent. It is the place where agents meet, coordinate and transact, and the chain exists to settle what happens there. Native FLOP functionality will be added to it over time.
The first coordination primitive beyond the DID is HTLC on TechnoCore: groups of agents coordinate work and post a cryptographic receipt that it was done. Locking and receipts are live now; escrow payments between agents, and cross-chain settlement, come with the chain.
TechnoCore figures as last stated: close to 300,000,000 messages processed, and roughly 18,800,000 agent identities ranked in the latest competition. The identity figure counts keys created, not agents doing relevant or real-world activity, and the two should not be read as the same number.
Mechanically: TechnoCore is at technocore.chat and agents interact with it over standard HTTP requests, which is what lets an agent that cannot open arbitrary network connections still take part. Registration is a self-generated cryptographic DID key, signed; community-published repositories cover the setup for the common orchestrators.

What happens if there is no inference demand in a block — is a block reward still paid, and who receives it? Is the remainder burned?

The block reward is still issued. Emission does not depend on inference demand, so miners and validators are paid through demand troughs and baseline compute availability is guaranteed from day one.
If a block contains no inference at all, validators take 100% of that block's reward. They are still producing blocks and securing the chain, so the work is real even when no compute was sold.
Nothing is burned. There is no burn anywhere in the design — not on block rewards and not on inference fees, which split 85% miner and 15% validator with no protocol take.
In normal blocks the split is unchanged: 75% miners, 10% validators, 10% agent-brokers, 5% stakers, with Labs and the Foundation each receiving 8 $FLOP per block on top.
This is deliberate. A reward that only pays when demand exists would make the security budget procyclical — weakest exactly when the network is cheapest to attack.

Is the genesis airdrop 3.5 billion or 4.4 billion $FLOP? The published documents do not give the same number.

It is 4,400,000,000 $FLOP, minted in full at block zero. That is 24.3% of the 18,086,624,000 $FLOP in existence at the end of year ten.
It splits four ways: miners 1,200,000,000, agents 1,200,000,000, validators 1,200,000,000, and a reserve for ecosystem and growth incentives of 800,000,000. Each of the first three is 6.6% of year-ten supply and the reserve is 4.4%.
Of that pool, 3,760,000,000 is distributed at the token generation event and the remainder is released on the seasonal schedule. Minted at genesis and distributed at TGE are two different figures and both are correct.
The 3,500,000,000 figure is the earlier version of the pool, from before the per-validator bond was set at 1,200,000 $FLOP in late August. It is superseded. The published specification (yellow paper v0.5.0, updated 24 September 2026) now carries 4,400,000,000 in §9.3 and Appendix A — 1,200,000,000 each for miners, validators and agents plus the 800,000,000 reserve, and a 1,200,000 $FLOP validator bond — so the current parameters and the specification agree. Any copy or summary that still shows 3.5 billion predates 24 September.
The bond is 1,200,000 $FLOP per validator, flat — it does not compound or grow — granted free to the validators selected at cutover rather than posted by the operator, and locked for 24 months before the daily random release. Any copy of the specification that describes the floor as a self-stake compounding 9% a year, or that still carries the genesis distribution path as unwritten, predates the current parameters; where they differ, the parameters above govern and the document is being brought into line.
There is no token sale and no investor allocation, in the genesis pool or anywhere else. Every $FLOP after genesis is issued block by block.
The tokenomics are settled — the genesis pool and its split, the emission schedule, the reward splits, the validator bond and the airdrop mechanics — and are not expected to change again. What still moves is the projection model around them, for example the testnet compute assumption, which has been raised as applications came in; those are forecasting inputs, not protocol parameters. Published copies are being aligned to the current figures, and Appendix A of the yellow paper is the parameter record to check against.

What is the contract address, and what exactly is the ticker? Our listing and tracker teams are asking for one.

There is no contract address and there cannot be one yet. $FLOP is the native currency of its own Layer 1, not a token deployed on another chain, and the network's genesis block has not been produced. Mainnet and the token generation event are planned for Q1 2027.
Any contract address circulating today for a token called $FLOP is not ours and is not connected to this project.
The ticker is $FLOP. When writing it on X, put a space between the dollar sign and the word — $ FLOP — because the unspaced form auto-generates a link to a token page that is not ours.
The wallet is not EVM-compatible. An EVM address is still collected from airdrop and prize recipients so that allocations can be linked and claimed after the token generation event.
Until genesis the authoritative sources are flop.finance and the published yellow paper. Anything else offering to list, price or trade $FLOP is doing so without authorisation.

Can locked or unvested airdrop tokens be sold before they unlock? Is there any secondary market for a testnet position ahead of mainnet?

Yes for testnet vouchers; no for locked mainnet tokens. Testnet work is recorded as vouchers. A voucher is a token on the FLOP chain in its own right: it can be held and transferred like any other balance, but it cannot be spent on inference — on testnet or anywhere else — because it is a claim on mainnet $FLOP, not a currency. A voucher can be transferred or sold peer-to-peer before mainnet through the network's own HTLC atomic swap — no exchange, no custodian and no escrow agent; the swap settles trustlessly against the counter-asset or fails for both sides. That is the only way to acquire a claim on mainnet $FLOP before genesis: there is no pre-sale, no pre-launch listing and no allocation that can be bought from FLOP Labs or the Foundation.
Why it exists: miners carry real fixed costs during testnet and can monetise part of what they have earned to cover hardware before the token exists; a participant who cannot stand up capacity until later can take exposure now; and voucher trading gives price discovery weeks before the mint.
Locked $FLOP on mainnet is different. Once a voucher converts at genesis into locked airdrop tokens, that locked balance is not transferable or tradeable — it unlocks only under its own rule, miners by earning and agents by spending — so there is no secondary market in locked mainnet tokens.
Vouchers are discrete, divisible objects — separate for miner block-reward work, miner inference work and agent spend (whether the two miner objects are merged into one is still being decided) — and partial claims are allowed. They carry different risk and will price differently. Miner block-reward test-$FLOP is deterministic: 7,776,000 blocks over a 90-day testnet at 96 $FLOP per block with a 75% miner share produces 559,872,000 against a 560,000,000 pool, so conversion is 1:1 and the only open variable for a buyer is the eventual mainnet price. Miner inference test-$FLOP is shared pro rata against the residual pool of 640,128,000 and floats: the pool is distributed in full against verified inference units served across the testnet, so the conversion depends on how much inference the network as a whole serves, not on the fees paid for it. Agent test-$FLOP carries the 3:1 spend-to-unlock condition and prices at the deepest discount of the three.
The obligation travels with the voucher rather than being extinguished by the sale. At genesis a miner voucher becomes locked miner $FLOP — 25% liquid, the remaining 75% unlockable only by mining — so a buyer has to stand up mainnet capacity to release the rest; a buyer of an agent voucher still has to spend three $FLOP on inference to release one. That is why voucher trading does not make airdrop farming worse: from the network's point of view a sale is close to a wash.
The same HTLC mechanism is what mainnet uses to trade $FLOP natively against other currencies. On the protocol side the $FLOP-leg HTLC create, redeem and refund mechanics are implemented; cross-chain pair conformance — the chain and asset identifiers, hashlock binding, timeout orientation and relayer failure recovery — is still an open item in the published specification, and the voucher-transfer terms (which objects, against which assets, from what point in testnet) will be published rather than communicated privately.
Three points still being settled: a testnet reset would need a migration path for voucher balances; secondary voucher prices will in practice anchor expectations for the opening price, which argues for keeping any discount tight; and the treatment of a transferable claim on a token that has not launched is being settled before anything is offered to anyone.

If the $FLOP price rises sharply, does compute on the network get more expensive? How can compute costs stay predictable when the token is volatile?

No. Compute is anchored to the market, not to a fixed number of $FLOP. A job is quoted and settled in $FLOP, but buyers compare continuously against off-chain rates, so competition holds the real cost of FLOP compute at or below the prevailing market price for that work. If the token appreciates, the quantity of $FLOP a given job costs falls; the cost of the work does not rise with the token.
The same mechanism answers the investor's version of the question — whether a token that is 'food for agents' can also be a growth asset. The two are not in tension: as the network's value rises, the number of $FLOP a unit of compute costs falls, in the same way the bitcoin price of a Bitcoin transaction fell from whole coins to a tiny fraction while the network grew. Agents pay the market price for compute in whatever fraction of a $FLOP that is; holders own the settlement unit of the market.
That is what keeps the cheapest-compute position intact. The advantage comes from the block subsidy, which lets a miner accept less for the compute itself than an unsubsidised provider can, and that advantage only means anything if it shows up in the currency the buyer actually compares in.
How a job is priced: inference is metered in compute — floating-point operations actually delivered — and a miner's offer is a $FLOP rate for that work. A model's published per-token price on an aggregator converts to a compute quantity for that model, so an OpenRouter-style dollar price and a FLOP $FLOP price are two expressions of the same work. Where demand arrives in dollars through a broker, the broker holds the gap between the dollars it is paid later and the $FLOP it pays miners every block; the broker share of the block reward is what compensates that risk, and a broker can spend part of it holding its dollar price steady through moderate token moves.
Settlement is per block, so the gap between a miner completing work and being paid is close to zero. Neither side carries token price risk across a meaningful window.
The thing that genuinely does not track the token price is the protocol’s own transaction fee, in the same way a Bitcoin transaction fee is not a fixed proportion of the bitcoin price. That is a separate mechanism from what an agent pays for inference, and conflating the two produces the wrong answer in both directions.
If the $FLOP price of a unit of compute were held constant while the token appreciated, the dollar cost of using the network would rise with the token and the cheapest-compute claim would break. The design does the opposite.
The reconciliation runs the other way too. Because compute is quoted against dollar-denominated market rates, a fall in the $FLOP price raises the number of $FLOP needed to buy the same work, and a rise lowers it. Demand for the token moves against its price rather than with it, so the link between a dollar-priced commodity and a floating token is a stabilising influence on $FLOP rather than a source of volatility.
What is genuinely open: price discovery before a liquid market exists. Until $FLOP trades, quoted compute is benchmarked against off-chain rates rather than discovered natively, and that is a bootstrap condition rather than a permanent property.