The algorithm

How BackSpin actually works.

No black box. Here is exactly how a wait turns into paid attention for developers, and into measured, fraud-resistant reach for advertisers, with the real formulas and an honest status on every part. It is all checked against the live code, and the code is open.

LivePartly livePlannedOpen source on GitHub

What you do: nothing extra

Live

You code the way you already do. While your AI assistant is thinking, BackSpin fills that dead wait with one short, useful sponsored line. There is no button to click, no schedule to keep, and nothing to babysit. When the wait ends, the moment is measured and you get paid for the real attention it earned.

AI wait starts -> a card shows -> you read it -> wait ends -> you get paid

Your attention becomes a window

Live

Each wait is captured as an attention window: when it started, samples of focus and keyboard or mouse activity while it ran, and when it ended. The window carries context like the tool and the estimated wait length, but never your code, prompts, file names, or repo. The producer ships the closed window after the wait, so scoring never touches a live generation.

How a window is scored

Live

Every window is scored on the published formula, clamped to 0 to 100:

AttentionScore =
    activity + focus + session_quality
  + workflow_relevance + trust_score - fraud_risk

Activity rewards real input (up to 40 points), focus adds 30 when the surface was in front of you, and a valid session adds 10. Workflow relevance sits at a neutral 10 for now. Trust score comes from your track record, and fraud risk is subtracted. A window pays out only if it was focused, had genuine activity, ran on a valid session, lasted at least 3 seconds, and stayed under the fraud line.

TrustRank rewards real history

Partly live

The more verified, human waits you build up, the more your trust score rises (capped at 20 points). A brand-new account starts right at the neutral middle, so you are never punished for being new, and a bot-like history erodes trust instead of building it. The deeper signals like device and network reputation come in a later phase.

What you earn, in real money

Partly live

A scored window turns into credits based on its quality-weighted seconds. You keep half the value your attention created, by the default 50 / 50 split, and you can see the dollar value the moment it lands.

you_keep = value x 0.50        (default split)
1,000 credits = $5.00          (current rate)

Cash out from the Withdraw page in crypto (USDT, Litecoin, Bitcoin, with PayPal on the way). Every payout gets a quick human review before it sends, which is also how we keep fraud from draining the honest pool.

Want the numbers behind this? See the public stats and the open market, or read the fraud policy.