2026-08-09

What We Cannot Verify

A site whose central argument is that this genre's advice decays owes an account of which of its own pages are decaying. A related workplace perspective is covered in Monitask’s how employees can tell if they are being monitored.

Six things we do not know

Which of these pages will be wrong in five years. Some will. The ones describing how a platform currently weights things are the candidates — the same construction that killed our 2010 material.

How platforms actually weight anything. Recommendation systems are not published. Everything anybody writes about them, including here, is inference from observed behaviour and vendor statements.

Whether the vendor figures are right. Email conversion multiples, campaign benchmarks, engagement ratios — sourced from companies selling the thing measured, quoted as theirs and not verified. For current technology context, see Ars Technica.

Whether the archive dates are exact. Forty came from the articles, thirty from the first comment in a feed, which is a lower bound. Seventeen have none.

What works for your genre. Almost everything published in this field generalises from a few genres with large audiences.

And whether any of it works for you. One catalogue, one audience. The calculation that answers it is yours and nobody else can run it.

What is offered instead

A dated archive that can be checked against what happened.

A distinction — tactic against mechanism — that holds up when tested against fifteen years of our own material.

Calculations you run, on your numbers, requiring nothing from us.

And sources labelled by who paid for them.

Why this page rather than a disclaimer

Because a disclaimer protects the publisher and a list informs the reader.

A reader who knows which parts are inference can weight them. One given uniform confidence cannot, and will be confidently wrong in a direction they cannot predict.

And because the alternative would be absurd here. A site arguing that this genre overstates its certainty cannot itself present inference as fact.

The correction we want

Evidence from a release. What you did, what you measured, what happened, over what period.

Particularly where it contradicts a page here. The only real evidence this subject has is somebody's actual numbers, and almost nobody publishes theirs.

The commitment

Nothing on this site gets silently updated.

Quiet rewriting with a fresh date is what erases the record, and it is the practice this site exists to argue against. Changes are logged with what changed and when.

The short version