Sign-up bonuses
A fintech app, marketplace or game pays a bonus for every new account.
Bonus hunters open dozens of accounts with fresh emails and cards. Atum links those accounts to the same principal, so the bonus goes out once.
Atum is proof of humanity without biometrics or documents. One person can run any number of accounts, wallets and agents. They all trace back to one principal, and anything scarce goes to that principal a set number of times.
A fintech app, marketplace or game pays a bonus for every new account.
Bonus hunters open dozens of accounts with fresh emails and cards. Atum links those accounts to the same principal, so the bonus goes out once.
An AI product gives new users credit to spend on requests, whether they type them or their agents do.
Without a check, one person can open a thousand accounts and walk away with a thousand credits. Atum traces them all to one principal, and the credit is granted once.
Everyone gets a daily allowance of free requests.
The allowance belongs to the principal, and every account and agent that person runs draws on the same pool. Signing up again doesn’t reset it.
A paid product, model or tool offers a free trial.
The trial is tied to the principal, not to the account or the agent. When it ends, a new account won’t start another one.
A platform covers its users’ first transactions so that getting started costs nothing.
The subsidy is capped per principal. An owner with a thousand wallets gets the same share as anyone else.
A service pays for every new user someone brings in.
A referral counts only when a new principal joins and does what the offer asks. Inviting your own second account earns nothing.
A store offers a discount on the first order, and an agent can find and apply it for its owner.
The discount applies once per principal, however many times the account, wallet or card changes.
Tickets, sneaker drops and limited releases sell out in seconds to people running dozens of accounts and bots.
A per-person limit holds across all of someone’s accounts and agents at once. It won’t stop a broker who pays other people to queue, since each of them is a separate life.
When a new model launches, there aren’t enough GPUs to go around.
Each principal gets one quota or one place in the queue. A thousand accounts or agents from the same owner still take one place.
A project pays people for testing, feedback or data collection.
Each principal is paid once per task, even if the work came in through several accounts or agents. The project hears from different people, not from one person many times over.
A platform rewards its most active users.
Rewards are counted per principal, so a thousand accounts or agents with one owner add up to one recipient.
A community votes on proposals, or someone asks an agent to vote for them.
Votes are counted per principal, so a thousand agents run by one person cast a single vote. Atum can’t stop different people from agreeing to vote the same way.
Marketplaces and app stores fill up with reviews one seller writes under many accounts, and more of them are now written by AI.
A platform can count reviews per principal and mark the ones that come from a person with a continuous history. Atum doesn’t prove the reviewer bought the product, and it can’t stop paid reviewers.
Online surveys and paid research panels lose a large share of answers to repeat respondents and to bots that pass attention checks.
A panel can take one response per principal and keep out agents posing as respondents. Careless or dishonest answers from real people are still the panel’s problem.
Quadratic funding gives extra weight to many small donors, so splitting one donor into many accounts pays off.
Matching is counted per principal, so splitting into accounts gains nothing. Atum can’t see collusion between different people.
On perpetual DEXs, prediction markets and on-chain terminals, farms trade volume back and forth between hundreds of wallets to collect points.
The phones behind one farm sit in the same room, move together and meet the same devices. Their histories overlap, and Atum sees that even when every wallet looks clean on its own.
One merchant runs dozens of P2P accounts, and a fake copy-trading leader pads a record of trades.
A platform only sees its own traffic. Atum adds a signal from the physical world about whether one principal or several people are behind the accounts.
When rewards are paid per device, farms keep thousands of emulators and phones.
An emulator can’t pass the chip’s hardware attestation. A thousand real phones in one room build one shared history, not a thousand separate ones.
Under x402 and similar protocols an agent calls an API dozens of times an hour and pays a tiny amount each time. Approving every payment by hand would cost more than the payments.
The merchant can see that a principal with a real history stands behind the agent and let it through, with no captcha and no manual approval. Agent payment protocols already have a field for this, and until now nothing could fill it.
An agent has to act for a person, within limits both sides understand.
The person signs a delegation with the key in their phone’s chip, covering operations, amounts, services and an end date. The service turns it into policy, and the person can revoke one agent without touching the rest.
AI now solves captchas, so they no longer stop bots, yet they still get in people’s way, and a site can’t tell a person’s agent from a scraper farm.
A site can set its limits per principal and let an agent with a real person behind it through without a captcha, up to that person’s budget. A person can still spend their own budget on scraping.
When an agent misbehaves, its owner can drop it and start a new one with a clean record.
The record builds up on the principal, so a new agent from the same person inherits it. Deciding what counts as misbehaviour is still up to the service.