The first HALO buyback: 219,944 tokens burned
Cycle 1 of the HALO buyback is settled on Base: 219,944 HALO burned permanently, 513,203 to stakers. Every step links to a public transaction.
The first HALO buyback cycle ran on 2 September 2026. Protocol fees were converted to HALO on-market, 30% was burned, and 70% went to the staking contract. Every step links to a transaction on Base.
219,944.298197 HALO is gone permanently. Burned tokens are never re-minted.
Cycle 1, in full
| Step | Amount | On-chain |
|---|---|---|
| VIRTUAL → HALO | 32,037.004032 VIRTUAL → 355,971.779487 HALO | 0x7dc628c4… |
| USDC → HALO | 238.053554 USDC → 3,752.024409 HALO | 0x72837843… |
| Bucket total | 733,147.660655 HALO | — |
| Burned (30%) | 219,944.298197 HALO | 0x8e849f8b… |
| To stakers (70%) | 513,203.362462 HALO | 0xddf44493… |
The rule that produced those numbers was published before the cycle ran, in the whitepaper:
- Pool creator fees — trading fees from the HALO/VIRTUAL launch pool: 100% to the buyback bucket.
- Vault protocol fees — the 10% fee on settled inference, collected in USDC: 80% to the bucket, 20% to a USDC treasury that funds operations without ever selling tokens.
- Of all HALO in the bucket: 70% to the staking contract, 30% burned. The split sits on a governance dial bounded between 20% and 80% burn.
One first-cycle detail worth stating plainly: the full 238.05 USDC of vault fees went to the buyback this time. The 20% treasury share — 47.610711 USDC — is carried forward and settles from the next accrual. The 80/20 routing applies strictly from cycle 2.
Where this money actually came from
This is the number that matters more than the burn, and it is not flattering yet:
| Source | Share of the bucket |
|---|---|
| HALO/VIRTUAL launch pool trading fees | 99.5% |
| Protocol fees on settled inference | 0.5% |
Cycle 1 was funded almost entirely by trading activity on the launch pool. Inference revenue contributed 238 USDC — about 3,752 HALO of a 733,147 HALO bucket.
We are publishing that ratio because it is the roadmap. The launch pool’s fees are real and they are locked in for ten years, but they are the bootstrap. The 0.5% line is the one that has to grow, and it grows for exactly one reason: people using the network to run inference.
What “on the path to deflationary” actually means
HALO is not a deflationary token today, and the whitepaper never claimed it was. It says the opposite, on purpose:
The first two years are a declared subsidy. Issuance runs at the scheduled figure whatever the network settles, and in that window new supply deliberately exceeds what the buyback retires.
A marketplace has to pay for activity before the activity exists. Pretending otherwise would just be a worse version of the same tokenomics everyone has already seen.
What the design does commit to is the shape of the curve after that:
- From year three, issuance cannot outrun the buyback. Each epoch mints at most the value the buyback returns to the market over the same period. That is a property of the mint, not a target anyone has to hit.
- Beyond that bound, the burn shrinks supply. When issuance sits below the burn share, more HALO is destroyed than created.
- Burning never creates room to print. Emission budgets are measured against cumulative-ever-minted, not current supply. A burned token doesn’t free headroom under the 2 billion cap — it removes it. And at least 500 million of that headroom has no schedule attached to it at all.
Why demand shows up as less supply
Both sides of that balance are funded by the same public variable: settled volume. That is the whole mechanism, and it is why you don’t need anyone’s forecast.
Every prompt paid for on Halo pays a 10% protocol fee. 80% of that fee buys HALO on the open market. 30% of what it buys is destroyed. So:
More inference settled → more fees → more HALO bought → more HALO burned.
There is no discretionary treasury decision in that chain, no announcement required, and nothing to trust. Demand for the protocol converts into buy pressure and permanent supply reduction by rule.
The same logic runs the other way, which is the honest half: if the network settles nothing, the buyback buys nothing and burns nothing. The token’s supply curve is downstream of whether the marketplace is actually used. That’s the design working, not a caveat.
Keep the number in proportion
219,944 HALO is 0.022% of the 1 billion total supply — 0.073% of what circulates. One cycle at this volume does not move a supply curve, and we’d rather say that here than have someone else point it out.
What cycle 1 establishes is that the loop is real: fees were collected, HALO was bought on-market, tokens were burned, stakers were paid, and all of it is verifiable on Base by anyone who wants to check. The rule ran exactly as written before it ran.
Cycles are executed manually by the team today, following the published rule.
That ends when the permissionless execute() contract ships and anyone can
trigger a cycle.
Check it yourself
- Ledger: BUYBACK_LEDGER.md — every cycle, every transaction
- HALO token:
0xbbd2…1d42 - Staking contract:
0x1466…56D5 - Executing address:
0xf3eE…F6F5
The full design — the fee split, the emissions rule, the governance bounds — is in the HALO whitepaper. The fastest way to make the next burn bigger is to use the network.