FolioU1BUnit 1
As at
Sun 22 Nov 2026
Entries
13
Not yet
12
Page seal 7CEB, 074°03′. Drawn from the SHA 256 of this page’s source text7CEB
Page seal7CEB, 074°03′

Mail Enforcement

Quality gate (target)
Fri 18 Dec 2026
First sale (target)
Mon 21 Dec 2026
Turnaround
Runs to reject or day 120, whichever comes first

Stage B of Unit 1, sold on its own. We take a domain that already watches its mail (DMARC at p=none) and move it, one approved step at a time, to quarantine and then to reject. At launch this job runs in runbook mode: AI agents read the reports, write each policy step and test the result; you, or your agency, apply each step; we do not log in to your systems. At launch, a second AI model from another company and a named reviewer under contract will check every job.1 At reject, receivers that check DMARC refuse forged mail in the domain's name. Each step waits until the reports show the client's own senders passing. Launch price: 120 USD direct; agencies that resell it pay us 76 USD (wholesale) and set their own price to you. Full price, once we can make the changes ourselves: 150 USD direct, or 95 USD wholesale.

Quality test (Bench) gate: Fri 18 Dec 2026 (target). Not yet: Bench gate passed and sealed2 Until it passes, Mail Enforcement is not on sale to anyone outside the firm.

Lines of this unit

LineLaunch priceLaunch price, through a partnerFull price, directFull price, wholesaleGate, target
Mail Enforcement (Stage B alone)120 USD76 USD150 USD95 USDFri 18 Dec 2026
Domain and Mail Gap Closure, Stages A and B312 USD200 USD390 USD250 USDSat 5 Dec 2026
Mail Alignment (Stage A alone)232 USD152 USD290 USD190 USDSat 5 Dec 2026

Who it is for

A domain that already has DMARC at p=none and at least 14 days of aggregate reports we can read. For an order in the first weeks after the gate, that means a domain whose existing reports reach us through its DMARC vendor's export. A domain that finishes Stage A with us reaches 14 report days about two weeks after Stage A ends.

The ramp

  • Quarantine (failing mail goes to spam) after 14 days of clean reports.
  • Reject (failing mail is refused) after a further 42 days, the last 30 of them with mail from senders you have not approved under 1 percent.
  • Each step is an approval. Nothing moves until the client or partner approves it in writing.
  • A sender that still fails holds the ramp, with a written hold reason, instead of bouncing the client's invoices.
  • The end point is reject, or day 120 from the order, whichever comes first.3

The 42 day step is our own design. Red Sift says quarantine to reject typically takes 6 to 8 weeks; DuoCircle suggests under 1 percent unknown traffic for 30 days before reject, and gives 30 to 90 days from quarantine to reject (both read 22 Nov 2026, S28).

Why it takes about 6 weeks

For a domain that already has 14 days of clean reports when you order, quarantine can be the first step, and reject comes 42 days later, about 6 weeks. A full Unit 1 order takes about 8 weeks, because Stage A and the first 14 report days come first. Some sellers state faster figures: a 45 day median (Valimail, S22) or about four monitored weeks (O365HQ, S24). We are slower on purpose. We wait for 30 days with mail from senders you have not approved under 1 percent before reject, so a forgotten billing or helpdesk sender shows up in the reports before it bounces.

How it is accepted

Stage B is accepted at the policy the report data allows, with any hold reason written down. The record is issued at that point. Rechecks then run at 30, 90 and 365 days after reject.

Warranty

  • Stage B: if reject is not reached by day 120 because our fix was wrong, we refund the Stage B share. In Tier 1, in runbook mode, that is 120 USD direct or 76 USD wholesale; at full price it is 150 USD direct or 95 USD wholesale. Other tiers get the same share of their lower price. It is not more than you paid.
  • Rechecks: we recheck at 30, 90 and 365 days after reject. If a 30 or 90 day check fails because our fix was wrong, we redo it free and credit you 30 percent of the price paid. If a 365 day check fails for that reason, we redo it free.
  • Who decides the cause: we propose one from a short list (you changed the setting back, your provider changed something, our fix was wrong, or other), with the evidence. You or your partner confirm it. If you disagree, the cause you choose goes on the record.4 We publish how often each cause happens, including other.
  • In runbook mode, if a test fails because a change was applied differently from the runbook, we point to the line, and you apply it again at no extra cost. The clocks pause until it is applied.5

The 30 percent credit comes from a reserve we fund.6 This is the Stage B and recheck part of the one Unit 1 warranty rule, word for word.

Runbook mode

The client applies each policy change from our runbook. We hold no write access. At launch every order runs this way; we make changes ourselves only once insurance that covers AI agent work is in place.7 Runbook mode is the launch price, 20 percent under the full price.

Price

The firm's own prices, in US dollars.

DirectWholesale
Mail Enforcement, launch price120 USD76 USD
Mail Enforcement, full price (once we can make the changes ourselves)150 USD95 USD

Tier shares apply by the client's region. Payment is taken in full at order.8

What the buyer gives

  • Proof the domain sits at p=none, with 14 days of aggregate reports we can read: a vendor export by API, or a report route on the client's own provider.
  • Proof of control, or runbook mode.
  • A written approval at each step.
  • For each sender that fails, the settings or DKIM record it shows you, or the reason it cannot be given.

What the buyer gets back

  • A step by step runbook for each policy change.
  • A written hold reason for every sender that holds the ramp.
  • A signed Stage B Proof of Fix record at the policy reached.9
  • Rechecks at 30, 90 and 365 days after reject, each bound to the warranty.

Who checks it

AI agents read the reports, write each step and test the result. At launch the client or their agency applies each step, and we do not log in to the client's systems. A second AI model from a different company checks each decision.10 At launch, a named reviewer under contract will review and sign every job.11 Is client data used to train AI? No, unless the client ticks a consent box that allows it.

The Bench

50 graded cases built from synthetic aggregate reports in the public report format. No domain reaches reject before this gate, so the reject decision is first graded on synthetic cases, then regraded on the first real reject with mail traffic. Result: Not yet: cases, pass rate, date12

Sources

Not yet in place12

Nothing is on sale yet; each numbered mark on this page points to a condition below. Every condition, with its state, is on the status board.

  1. 1second model vendor connected; two contracted reviewers, target Mon 30 Nov 2026
  2. 2Bench gate passed and sealed
  3. 3start of the 120 day count for Mail Enforcement bought alone confirmed in the terms
  4. 4dispute rule set in the terms, including what happens to the refund while a cause is disputed
  5. 5rule confirmed in the terms
  6. 6recheck reserve funded
  7. 7cover that includes AI agent work bound
  8. 8company formed and checkout live
  9. 9client record signing live in our record log
  10. 10second model vendor connected
  11. 11two contracted reviewers, target Mon 30 Nov 2026
  12. 12cases, pass rate, date

Each item is also marked where it stands on this page.