System Map — click any node
Qalnet is a private, Apache-2.0 fork of the IOTA node (Sui-lineage, Move objects, Starfish-class DAG-BFT consensus) run across qalarc's own fleet. Click nodes for detail.
The life of a QALS — animated
Follow one Australian dollar: it enters the segregated Reserve, mints 1 B-QALS, pays for compute, and burns. Press play, or step manually.
Why this can't break
Every B-QALS in existence was minted only after the Oracle Committee confirmed a matching AU$1 landed in the segregated client-money account. The chain is the liability ledger — anyone can compute the invariant:
reserve_aud_cents ≥ b_qals_supply × 100 (every block)
The two token classes
| B-QALS | G-QALS | |
|---|---|---|
| Born | AU$1 deposited | growth allocations |
| Backing | AU$1, in Reserve | none (FloorVault floor) |
| Redeem | AU$0.995 anytime | market / floor |
| When used | burns + revenue | burns (deflationary) |
Compute Job Lifecycle — click the states
On Qalnet, compute is a tracked object: escrowed, attested, settleable, slashable. Click any state to see the rules, who can act, and what the chain enforces.
Compute receipt NFT
Every SETTLED job mints a ComputeReceipt: model hash, input anchors, GPU-seconds, cost, provider DID. It doubles as provenance for AI outputs (C2PA-style, on-chain) and as collateral in the Qalx receivables vault.
Vouchers — compute as a gift
ComputeVoucher objects are transferable GPU-hour credits. Move's linear types make double-redemption impossible — the compiler enforces it, not a promise.
Credit Engine — simulate an account
Every app user, device and AI agent gets a CreditAccount with chain-enforced caps. Simulate one: drag the sliders and watch a month of metered usage settle.
Holds (pre-authorisations) auto-expire — no orphaned holds, ever. Caps are enforced by the Move VM: an agent cannot overspend even if compromised.
AU$1 Backing Simulator — 36 months
The honest engine room: B-QALS are born 1:1 from deposits and burn on use (left chart — the invariant always holds). G-QALS carry no claim, but a revenue-funded FloorVault ratchets their floor (right chart — coverage is honest and slow; parity needs scale). Drag the assumptions.
No simulator output is a promise. G-QALS coverage typically stays under a cent for years — the wiki dashboard publishes the real number monthly.
Safety & Verification Systems
The stack that makes Qals boringly safe: structural invariants, multisig humans, timelocks, and public proof. Toggle any control to see what fails without it.
Proof-of-Reserves (live example)
| Item | Value | Source |
|---|---|---|
| B-QALS outstanding (on-chain) | — | Qalnet RPC (public) |
| Liabilities (AU$) | — | supply × AU$1 |
| Reserve (segregated AUD) | — | Oracle Committee attestation |
| Coverage | — | reserve ÷ liabilities |
| Last attestation | — | signed statement, on-chain |
Values update from Module 5's month-36 simulation — try it.
Who can do what (privilege map)
| Role | Power | Check |
|---|---|---|
| Oracle Committee | attest reserves, trigger investigate | 3-of-5 multisig |
| Treasury Council | fees/floor/LTV params | 2-of-3 + 48h timelock |
| Emergency pause | halt mint/redeem/pools | multisig, 72h auto-expire |
| Validators (Fleet) | order transactions | Starfish BFT, 3f+1 |
| Move VM | everything else | linear types — code can't cheat |