Skip to content
ATON
CommunityInvites open with the campaign.

Earned quickly.
Kept slowly.

Discord and Telegram, both moderated from the first day rather than after the first problem. Referral rewards are live from the start, and every one of them sits in a holding window before it becomes irreversible.

A programme that pays the instant somebody clicks is a programme being farmed by lunchtime. The wait is the part that makes the rewards worth anything, so it is the first thing on the page rather than a line in the terms.

Discord

Development chat, testnet coordination, and where most questions get answered fastest.

Invite opens with the campaign

Telegram

One channel, everyone in it, no tiers and no private rooms. Announcements and lighter conversation, on the same rewards and scoring backend as Discord.

Invite opens with the campaign

Referrals and rewards

Bring people in and you earn credit. Smaller tiers pay out as a discount on your subscription. Larger tiers pay out as a reduction in network gas fees, which is an on-chain entitlement rather than a promise, banked before mainnet and honoured when the chain is live.

Referrals are single level. You earn on the people you bring, not on the people they bring. Deeper structures start to look like a different kind of scheme with a different set of rules attached, and that is not a line worth walking up to for a marginal gain in growth.

Smaller referral tiers

Subscription discount

Applied to your Atheron AI plan when it goes live.

Larger referral tiers

Gas fee reduction

Banked before mainnet and honoured at launch.

Why rewards have a waiting period

Rewards go live from day one, which is the right call for growth and the wrong call for fraud resistance if nothing sits between earning and keeping. So there is a pending window, between fourteen and thirty days, before a reward becomes irreversible.

Engagement is scored for quality rather than volume, so posting the word "gm" four hundred times is not a strategy. There is light-touch human verification and a signal for machine-written text, and that signal is treated as one weak input among several rather than as a verdict. Text detectors are unreliable enough that treating one as a gate would be worse than not having it.