Infrastructure for attention that can be trusted

BackSpin producers live where AI work actually happens: editor extensions, command-line tools, browser surfaces, and agent integrations. They observe product-specific wait states and present discovery without reading prompts, source code, model responses, or the contents of a person's work.

The integration method changes by host because each environment exposes different safe, reversible extension points. The contract is the same: report a bounded attention window and let the service decide what it means.

A producer can never mint credits by itself. It opens a session, signs its account-bound events, and submits a completed window. The API checks the signature, server time, session state, eligibility rules, replay resistance, reward caps, campaign availability, and anti-farm signals before it creates any accounting record.

That separation is deliberate. The client is responsible for a useful experience. The service is responsible for the market, the money, and the audit trail.

Reward shares, campaign limits, frequency caps, velocity rules, eligibility requirements, and payout decisions are owned by the server. Admin settings change that server-held configuration. A client can display the current product experience, but it cannot choose its own reward multiplier or invent a successful result.

This also makes operational decisions traceable. When a policy changes, the decision is made once in the service and applied consistently across CLI, editor, browser, and agent surfaces.

The open algorithm explains the scoring, ranking, market, and revenue ideas that people need to evaluate BackSpin. It is available at github.com/hackazer/backspin-algorithm.

The production service keeps abuse thresholds, challenge material, risk correlation details, and enforcement criteria private. Publishing those details would turn a transparency document into a playbook for defeating the controls that protect legitimate users and advertisers.