How it works
One line. Two engines. One gate that decides what ships.
Twenty agents across five layers turn a request into shipped units. This page goes one level deeper than the homepage into why the line is built the way it is.
The six stations
Every unit walks the same route.
Whether it's a lead or a page, a unit enters at Intake and only leaves at Ship. Nothing skips the gate.
01IntakeYour ICP or topic map becomes a run definition. Suppression rules and exclusions load before anything else runs.
02SourceCandidates are pulled from public data and provider APIs. Suppressed entries never enter the pipeline.
03QualifyEach candidate is scored with a written rationale attached. No rationale, no pass.
04VerifyContacts are deliverability-checked. Written claims are checked against the evidence that produced them.
05 — the gateSpike or passA quality rubric decides. Unsupported claims, thin rationale, undeliverable contacts and regulated-industry promises are spiked and recycled here.
06ShipSurvivors are staged to your dashboard, emailed, or published to your CMS. You're billed only for what ships.
Station 05, in detail
What the gate is actually checking.
The rubric groups into a few categories. We're not publishing the literal checklist — that's an operational detail, not a marketing claim — but here's what it's built to catch.
EvidenceIs this traceable?Every factual claim in a written page or a lead rationale needs a source it came from. No source, no pass.
FitDoes this match the ICP?A lead that scores well on paper but doesn't fit the actual definition you gave is held rather than shipped.
DeliverabilityWill this actually reach someone?Contacts are checked for real deliverability before they're billed — a bounce doesn't become your problem after the fact.
DuplicationHave we shipped this already?Repeats against your own delivered history or suppression list are caught before they ship a second time.
Regulated languageIs this an unsafe promise?Claims that read as guarantees in regulated categories — financial, medical, legal outcomes — are blocked at the gate, not left to a disclaimer.
Thin outputIs there actually enough here?A rationale or a page that's technically present but too shallow to be useful gets held for a rewrite rather than passed on a technicality.
Who's watching the watcher
A few operating principles.
Cost is capped before it happensEvery run checks against a cost ceiling before it starts, and again before each expensive step. A run that would breach your monthly cap is refused up front — not explained afterward.
You approve, the machine executesNothing publishes or sends without a human decision point in the loop — either your standing settings or an explicit approval in your dashboard.
The agent that checks isn't the agent that made itA sourcer never scores its own candidate. A writer never approves its own page. The gate is a separate checkpoint from production, every time.
Nothing ships on a guessIf a unit can't produce a rationale or a traceable source, it doesn't get "close enough" treatment — it's held, rewritten, or spiked.