Docs · how it works

The full mechanics.

This is how Ponslab is designed to work: sections 1–3 and 6 enforced by an on-chain vault and a metering backend, reconciled publicly every day once live. For design rationale, read the whitepaper.

Status: pre-launch. Vault, metering, playground, and the deploy pipeline are built; launch integration (M5) ships next — see the roadmap.

1 · Lock $PONSLAB, receive credit

Connect a wallet on Robinhood Chain (4663) and lock any amount of $PONSLAB — there is no minimum. On lock, you're credited inference tokens at the prevailing on-chain rate R.

• Baseline: 1,000,000 $PONSLAB ≈ $5 ≈ 100,000 inference tokens.
• No oracle. R is a fixed token quantity, adjusted by admin as the market moves. Rate changes are never retroactive — credit keeps the rate it was issued at.
• Locking is untaxed: the 2% creator tax applies to trades only, so the vault credits the full amount you lock.
• Your profile is your wallet. Signing in costs no gas; the lock is your first transaction.

2 · Three balances

Prompting consumes credit — base first, bonus second. Every unit of I_base you spend burns the matching $PONSLAB out of your lock. Spending bonus burns nothing: it's a gift from the treasury.

3 · The daily bonus

Idle credit earns a bonus of 25% of idle I_base per day, credited to I_bonus.

• Non-compounding — always computed on base credit, never on previous bonuses.
• Each grant expires 24 hours after issue and never stacks across days.

Three builders, same 1M lock, very different outcomes:

The regular builder captures the most value. Parking capital and burning out in one session are both weakly rewarded — by design.

4 · Prompt, preview, and per-model pricing

The playground streams generations from the full Orbio catalogue — 400+ models under OpenRouter naming. Models are priced by what they actually cost: credit is weighted against a $50-per-1M-token baseline.

• A $50/M-output model meters at ×1 credit per token.
• A $10/M model meters at ×0.2 — your credit stretches five times further.
• Free models are floored at ×0.01 so consumption always burns.
• If a model's price can't be read, the full ×1 weight applies — outages never mint discounts.

Previews render in a locked-down sandbox whose policy blocks fetch, beacons, images, and form posts — generated apps are required to be self-contained. Treat previews as untrusted regardless: never enter secrets into one.

5 · Launch: one atomic transaction

1. Build first. Your app is hosted on a temporary subdomain while you iterate.
2. Burn + launch together. A single transaction burns a fixed $PONSLAB amount (admin-set, targeting roughly $5–10) from your locked balance — so you need at least that much still locked — and calls the pons v2 factory with your wallet as the token's fee recipient. Your inference credit reduces proportionally with the burned tokens. You also pay the pons fee (0.0005 ETH) plus gas.
3. Atomic means atomic. If the factory call fails, the burn reverts with it. Your tokens cannot burn without a launch existing.
4. Promote. On confirmation your app moves to its permanent *.ponslab.family subdomain.

What's gated is the deploy, not the token — anyone can mint on pons directly. The burn pays for what only Ponslab provides: generation, hosting, and the subdomain.

6 · Withdrawing

• You can withdraw your locked balance net of burns at any time, after a 24-hour cooldown between request and release.
• The cooldown is an anti-abuse measure: it stops one balance being rotated across wallets to multiply daily entitlement.
• A burn is never blocked by a pending withdrawal — if free balance can't cover it, the withdrawal request is cancelled instead. A launch never fails because you queued a withdrawal.
• On withdrawal, credit is reduced proportionally with the tokens that leave.

7 · App lifecycle

Hosting is effectively free to keep, so there is no rent. Apps with zero traffic for 30 days are auto-sunset — taken offline to keep the shared domain clean. Your app's token lives on pons and is untouched by sunset.

FAQ