Hartii · Quai Network · Edition 0.5 · 16 November 2026
The Hartii
Eco Thesis
A payment network on Quai where every bank runs the same code, checked against the chain around the clock.
A living document. Proposed, not live. Everything marked proven ran on the Orchard testnet; everything else is design.
Payments this round
SAME CODE
QUAI BLOCK —
or click the network · hold a bankor tap the network · hold a bank
01
The thesis, in one breath
A card network does five jobs: it authorizes a payment, clears it with others, settles the money, funds the merchant early and resolves disputes. Quai already gives us two monies and programmable settlement. Hartii has built two of the missing parts and tested them on the testnet: Counterflow for clearing and settlement, and the Unlock Desk for funding. The thesis is that the rest can be built as signed, conditional promises that no one has to trust, only check.
Authorize
A payer signs a promise with limits and conditions. A risk co-signer adds a second signature or declines.
Clear
Counterflow pairs opposite flows and nets them, so most value never moves twice.
Settle
One transaction pays every side of a batch, or none of it happens.
Fund
The Unlock Desk turns money that arrives later into money now, at a priced discount.
Dispute
Funds wait in a short hold window; a signed refund or an arbiter decides.
In this thesis a bank is anyone who co-signs for payers or collects for merchants: the issuer and acquirer roles of a card network.
02
How a card network works
Four parties and a network in the middle. Authorization takes seconds; clearing and settlement happen later, in batches.
Authorize
The issuer checks the card, the balance or credit line and its fraud rules, then says yes or no in seconds. Nothing has moved yet.
Clear and settle
Later, in batches, the network adds up what every bank owes every other bank and moves only the net amounts, typically the next business day.
Who is paid
The merchant pays a fee on each sale, shared between the issuer, the network and the acquirer. Totals vary by country and card; the shape is what matters here.
03
Card network, part by part, on Quai
| Card network job | Hartii piece | State |
|---|---|---|
| Authorization | Conditional signed intent + risk co-signature | design |
| Bank conformance | One registered build, heartbeat attestations, authority that voids itself | design |
| Clearing and netting | Counterflow: pairs opposite flows in one batch | proven on Orchard |
| Settlement | One transaction pays all sides or reverts | proven on Orchard |
| Early merchant funding | Unlock Desk: sell money that arrives later, at a discount | proven on Orchard |
| Card holds and captures | Escrowed hold with a capture deadline | design |
| Disputes and refunds | Hold window, signed refunds, arbiter | design |
| Wallets in people’s pockets | Pelagus and Blip, the wallets Quai users already have | wallets live today |
| Network fee | A small protocol fee under published rules | design |
04
Authorization is a signed promise with limits
The payer signs once. The promise names exactly who, how much, until when and under which conditions. A risk co-signer, the issuer’s role, adds a second short-lived signature or declines.
PAYMENT INTENT
- who pays
- you
- to whom
- one shop
- at most
- 40 QUAI
- until
- a short window
- rules
- the ones you set
signed by payer
CO-SIGNATURE
- risk check
- passed
- valid for
- seconds
signed by risk co-signer
AUTHORIZED
The settlement contract can pull up to 40 QUAI for this merchant, once, in this window.
Nothing in the intent gives anyone else power over the payer’s wallet. It is a permission with edges, checked by code at the moment money moves.
05
Signatures that void themselves
The core idea. A signed promise is only honoured if every condition holds when it is used. Break one and the same signature is worthless. Switch the conditions to see it.
Signature
VALID
Every condition holds. Settlement may pull the funds.
Every rule is checked at the moment money moves, and any single failure stops it. One action from the payer can void a single promise or every promise at once.
06
Every bank runs the same code, checked against the chain
Here a bank is anyone who co-signs for payers or collects for merchants. There is one registered build. Every bank runs it around the clock, and anyone can check it against the chain at any moment.
Network
CLEARING
All 5 banks run the registered build and match the chain. Payments clear.
- 1
One build
The bank software is open and built reproducibly. Its hash is registered on chain, and changing it takes a public delay.
- 2
A heartbeat
Every few blocks each bank posts an attestation: the build it runs and the state it sees.
- 3
A check anyone can run
Watchers, other banks and every signer re-read the chain through their own node before they act, so nobody takes a coordinator’s word. A mismatch or a silence is public within blocks.
- 4
Authority that voids itself
A bank whose attestation is wrong or late loses its authority on its own: the settlement contract stops taking its signatures. No vote, no operator.
Decentralised by default: no bank is special, and none keeps its power by being trusted. The only exceptions are the recovery steps written down in advance, each narrow, public and time-limited. design The registry, the attestations and the voiding rule are not built yet.
07
Holds and captures, like a hotel deposit
Hold
The intent moves funds into escrow under the payer’s name. The merchant knows the money exists; the payer knows it is not yet spent.
Capture
Before the deadline the merchant captures up to the held amount. A smaller capture refunds the rest in the same step.
Release
No capture in time and the hold expires: anyone can trigger the refund, so nobody’s money waits on a server.
08
Clearing: Counterflow pairs opposite orders
People selling QUAI and people selling wrapped Qi are paired against each other. Matched value swaps directly. Only the leftover needs a provider, who fills it from their own stock up to a hard limit.
One batch
Opposite orders are paired in one batch. If any order cannot be honoured, nothing in the batch moves.
A provider for the rest
Whatever is left over is filled from a provider’s own stock, within firm limits.
Nobody can block it
Every side is paid in the same step, and no single wallet can hold up anyone else’s payment.
09
Counterflow, tested
On the Orchard testnet proven
| Trial | Result |
|---|---|
| A QUAI seller and a seller of wrapped Qi paired, both paid in one transaction and paid out by withdrawal | pass |
| A batch with one impossible order is refused, nothing moves | pass |
| One-sided order filled from provider stock | pass |
| Expired order refunded to its maker | pass |
What comes next
Four trials with test coins on a public testnet; a test token stands in for wrapped Qi. Netting pays when payments flow both ways, so the next step is real two-way payments between a first merchant and a first payer.
10
Funding: money later, sold for money now
The Unlock Desk is a market for Quai mining rewards that are locked for 3, 6 or 12 months. A miner sells a locked reward at a discount for QUAI today; the buyer receives the full amount when it unlocks. The same machine can fund a merchant waiting on settlement.
Both sides choose the trade: cash now for the miner, the full amount later for the buyer.
11
Unlock Desk, tested
On the Orchard testnet proven, 7 of 7
Seven of seven trials passed with test coins: selling a locked reward, buying it, selling into a standing bid, selling automatically inside a miner’s own floor, a resale, refusing early or unauthorised claims, and paying every owner the full amount at unlock.
It runs on the Orchard testnet with Pelagus. Test coins only.
What comes next
A mining node of our own on Orchard, to test the protocol’s real payout of locked rewards before anything touches mainnet.
12
Disputes without a call centre
On a card network the issuer can pull money back. On a chain, money that moved stays moved, so the protection has to sit before the money leaves the hold.
Hold window
A merchant’s capture lands in a short hold before it is final. Most payments simply pass through it.
Signed refund
A merchant can sign a refund intent at any time, with the same self-voiding rules as a payment.
Arbiter
A disputed hold freezes. A named arbiter, later a small panel, signs the outcome. The rules and every decision are public.
Design, not built. The hard part is not code: it is who the arbiter is, what evidence counts and how long a merchant waits. Those are product decisions to make before a line is written.
13
When something breaks
Decentralised does not mean nobody can fix anything. It means every fix is written down in advance: who may act, how far and for how long. Pick a failure to see what still works.
Lit: the steps this failure can use, from the least power to the most
14
Built so no one can capture it
Every method this network can use to stay decentralised, hard to censor and private, with its honest status today.
Decentralised
- design One registered build, made reproducibly so anyone can match its hash.
- design Heartbeat attestations from every bank, checked against the chain.
- design Authority that voids itself on a mismatch or a silence.
- design Open entry: anyone with the registered build and a bond can run a bank.
- design Two independent versions of the bank software, so one bug cannot take all.
- design Co-signing by a threshold of independent signers, never by one key.
- design Signers that re-read the chain through their own node before they sign.
- design Fail closed: a changed detail stops the flow, and a retry never pays twice.
- design Upgrades only after a public timelock.
- proven Settlement in one transaction: every side is paid, or nothing moves.
- proven Payouts by withdrawal, so no wallet can block anyone else’s batch.
Hard to censor
- live RPC fallback, live today in Hartii Labs: when one node cannot be reached, another is used.
- Quai Anyone can run their own Quai node and read the chain directly.
- proven Expired orders refund to their makers with no operator involved.
- design Holds that release themselves after a deadline; anyone can trigger it.
- design Offline signing: a signed payment can be handed to any relayer.
- design A forced exit with your own key after a fixed delay.
- design A permissionless matcher fallback after a timeout.
- design Static pages with a published hash, mirrored on other domains.
Private
- Quai Qi payment codes: payer and payee are not linked on the chain.
- design Native Qi in clearing: a shared two-of-three key that looks like any Qi address.
- design A one-time payee address for every payment.
- design Only hashes on chain; invoices and documents stay off it.
- design Sealed memos that only the merchant’s bank can read.
- design Orders in fixed sizes, so an amount says little on its own.
- design Net positions on chain instead of single payments.
- design View keys for audits and disputes, shown only to those who need them.
- early idea Intents that stay sealed until the batch closes.
- live This page stores nothing, sets no cookies and runs no trackers.
live runs in production today · proven passed on the Orchard testnet · Quai part of the Quai protocol · design planned, not built
15
Who can see what
Private by default, provable on demand. To pair and net orders a matcher needs four fields, and nothing more.
Needs, per order
- Side
- Amount
- Limit
- Deadline
Never needs
- Who pays
- Who is paid
- Why
- Their history
The co-signer always sees what it has to check, and the merchant’s bank sees its own payments; privacy means that nobody else does. Qi’s payment codes protect native Qi, but wrapped Qi is as public as QUAI, so on the contract side privacy comes from the tools above, or from keeping Qi native, the next chapter.
16
Keeping Qi native
Wrapped Qi is as public as QUAI. Qi can stay native instead, so the privacy it was built with survives clearing, with no bridge and no wrapper to trust.
No one acts alone
Buyer, seller and a neutral arbiter each hold one share of a key made for this trade alone. Any two can sign; nobody can sign by themselves.
Private on chain
Qi signatures are Schnorr signatures, so a shared key looks like any other Qi address. Each trade gets fresh addresses that are never reused.
Ordered, then checked
Qi is locked first. Watchers that read the chain through their own nodes confirm it is final before any QUAI moves.
proven on Orchard Trades have run this way on the testnet from start to finish: native Qi locked and confirmed, QUAI locked in a contract, both sides released by two of the three. So have the hard cases: a buyer who never pays, an agreed refund, a dispute decided each way with the arbiter, and a restart after a crash. All of it ran on one machine through one shared node; separate machines and nodes come next.
Two chains never settle in one step. Ordering, deadlines, bonds and finality checks keep the gap small and visible, but they cannot undo a payment that is already final. That is why the arbiter exists, and why its powers are fixed before a trade starts. design
17
The architecture, layer by layer
Select a layer.
18
Under the hood: what Quai gives us
Two monies on one chain, programmable settlement, and a mining reward that can be locked for a bonus. Facts from the go-quai source and the qu.ai docs.
QUAI
The account-model coin. Smart contracts, gas, every Hartii product today. Settlement logic lives here.
Qi
The UTXO coin with private payment codes, designed by Quai for everyday payments. It does not run inside contracts, so a contract cannot pay or receive native Qi in the same step.
WQi
Wrapped Qi, an ordinary token on the contract side. This is what a contract can move atomically. Its backing is the wrapper’s, not a law of nature.
Conversions and locks
QUAI and Qi convert natively, but conversions unlock after about two weeks, and a mining reward can be locked 3, 6 or 12 months for a bonus that is shrinking block by block.
| Measured or read | Value | Source |
|---|---|---|
| Live zones | Cyprus-1 only | qu.ai networks page |
| Lock bonus today, 3 / 6 / 12 months | +2.88% / +8.23% / +20.58% | go-quai source, Nov 2026 |
| QUAI mined per day | about 909,000 | Quai mainnet, Nov 2026 |
| Reward that takes no lock bonus | 99.78% | Quai mainnet, Nov 2026 |
| Locked rewards owned by a contract | 0 | Quai mainnet, Nov 2026 |
| Block time | about 5.2 s | 50,000-block epoch ≈ 3 days |
19
What backs Qi, exactly
Qi is minted in direct proportion to proof of work, on a fixed schedule. That is a precise claim, and a narrower one than “energy-backed”.
| Claim | Does the code back it? |
|---|---|
| Each Qi costs the same number of hashes, at a given time and share mix | Yes. By construction: 8×10¹² × m KawPoW-equivalent hashes per Qi. |
| Each Qi carries the same energy | No. Energy is hashes times the joules per hash of the hardware used, which the protocol never sees. |
| Each Qi costs the same in dollars | No. That also needs electricity prices and the market price of QUAI. |
| Qi is pegged to energy, or redeemable for it | No. There is no peg, backing or redemption in the protocol. |
The exact rules, read from the go-quai source, release v0.57.1
Qi per block
Qi = D′ ÷ (8×10⁹ × m) qits
1 Qi is 1,000 qits. D′ is the network’s hash-based difficulty, anchored to SHA-256 work. m rises from 1 to 4 on a fixed schedule, about 1.66 today, so later Qi takes more hashes as hardware improves.
QUAI per block
QUAI = κ × log₂(D′ ÷ 3×10¹¹)
κ is about 13.26 QUAI and is frozen by the protocol for now. QUAI grows with the logarithm of the work; Qi grows in a straight line with it.
The rate between them
QUAI per Qi = QUAI reward ÷ Qi reward
Both read from the same block. As the network adds hashpower the protocol gives fewer QUAI per Qi, because Qi issuance rises in a straight line while QUAI rises with a logarithm.
Difficulty
target = 2²⁵⁶ ÷ D
Difficulty is the expected number of hash attempts per block. Zone, region and prime blocks come from the same hash, and the reward does not depend on the level.
Mining turns electricity into heat, and that cannot be undone. The protocol mints Qi against proof of that work. It measures neither joules nor free energy: proof of work is evidence that energy was spent, not a meter of how much.
What it means for payments: Qi is a unit of work on a tightening schedule, not a stablecoin. Nobody can mint more without more work, and its price still comes from markets. That makes it an honest settlement unit, and it is why this network quotes prices and never promises them.
20
Scale: honest numbers
| Item | Today | What it means |
|---|---|---|
| Live zones | 1 (Cyprus-1) | All settlement runs in one zone; more zones mean partitioned books, not one global queue. |
| Block time | about 5.2 s | Authorization is off-chain and instant; settlement follows in blocks. |
| Two-way volume | unproven | The deciding number. Without steady demand in both directions netting saves little. |
A card network’s speed comes from authorizing off-line and settling later in bulk. This design does the same: signatures are checked instantly, money moves in batches, and only the net crosses the chain.
21
How we will act on it
In order, and only forward when the step before is proven. No dates: each stage starts when the last one has passed.
- 1 · now
Prove it on the testnet
Self-voiding promises, the bank registry, holds and captures, and native Qi clearing, each on Orchard with test coins. The Unlock Desk (7 of 7), Counterflow (4 of 4) and native Qi trades, with their refund, dispute and restart drills, have passed.
- 2
Break it on purpose
Drills for refunds, faults, disputes, restarts, chain reorganisations and silent banks. Then an independent review of every contract and signer.
- 3
A capped beta
After that review and legal advice: real value with hard limits per payment and per day, a small set of banks, everything public.
- 4
First merchant, first payer
Real payments in both directions, measured in the open. Only then does it grow.
22
A living document
This page will be updated every time a piece is built, tested or proven wrong. Each edition is listed below with what changed, so nothing is quietly rewritten.
- 0.1 · 15 Nov 2026 First edition. Unlock Desk 7/7 and Counterflow 4/4 on Orchard; signed-intent and invalidation design.
- 0.1.1 · 15 Nov 2026 Tightened wording.
- 0.2 · 15 Nov 2026 A living network on the cover that checks every bank for the same code. New chapters: what backs Qi, every bank runs the same code, when something breaks, built so no one can capture it, who can see what, and what the matcher needs to see.
- 0.3 · 15 Nov 2026 Keeping Qi native: clearing without a bridge or a wrapper. How we will act on it, stage by stage. Signers that re-read the chain for themselves. The story comes first and the protocol facts sit under the hood; the cover has keyboard controls.
- 0.4 · 16 Nov 2026 A first native Qi trade ran end to end on Orchard: Qi locked first, QUAI locked second, both released by two of three. It ran on one machine with one shared node; separate ones are next.
- 0.5 · 16 Nov 2026 The hard cases of a native Qi trade ran on Orchard: a buyer who never pays, an agreed refund, a dispute decided for each side, and two trades finished after a crash. Still one machine and one shared node.
Next to prove
- A signed intent with self-voiding rules, on Orchard, with Pelagus and Blip signatures.
- A bank registry with heartbeat attestations and authority that voids itself, on Orchard.
- A threshold of co-signers, with a payer-only path when they fall silent.
- Holds and captures with an expiry anyone can trigger.
- Native Qi trades with each role on its own machine and node.
- Our own mining node on Orchard, to test the real lockup payout.
- Real two-way demand: a first merchant and a first payer.
Hartii · hartii.org/eco · Nothing here is an offer, a promise of return or financial advice.