Say who you are, and nothing else about you

On a Sunday night a prospect who had accepted my invitation got a first message from my account that read, in part, "I run Hellofew on my own account and I'm an affiliate, so I earn a share if you sign up." I built Hellofew. I have never been its affiliate. The next morning I read it twice, swore once, and went looking for how a sentence that was never true of me had gone out under my name.

The answer was not the model. Three weeks earlier I had tried a different setup line for the people helping me sell it, one that said affiliate, because for a day that was the plan. The plan changed, the line was retired, and every draft written under it that had not yet been approved was rewritten. This one had been approved. It sat in the queue, correct by every check the product ran, waiting for an acceptance that came three weeks later. The checks all ran at the moment of writing. None ran at the moment of sending.

The sentence that carries the trust

A cold message has four beats, and the second one is the sender saying who they are. It is one line, and it is the one line the reader uses to decide whether the other three are worth reading. "I built this" and "I'm an affiliate" produce different readers of the same next sentence. Get it wrong and the whole message is wrong, however good the hook was.

So the rule is now plain and it is enforced in code. The drafter is told exactly who the sender is, in the sender's own words, and is forbidden from saying anything else about them. Not building it if they use it. Not a friend of the founder if they are a customer. Not an affiliate unless the workspace says it is one, in that word. The people who help me sell it say what is true of them: they use it for their own outreach, a friend built it, they help people set it up.

Checked at the door, not at the desk

The fix for the queue was to stop trusting the draft's age. Every message is read again the moment it is about to go. If it carries a line the workspace no longer says about itself, it does not go. The action waits two hours, the draft is rewritten from today's positioning, and the next attempt sends the new one. An hourly sweep does the same for every queued draft in every workspace, so a retired line cannot wait three weeks for its turn again. After three rewrites that still produce it, the send is skipped and I get told, because then the line is in the workspace's own description and a person has to fix the words.

A real affiliate keeps the word. The check compares the draft with what the workspace says about itself, so a sender whose positioning says affiliate is allowed to say affiliate, and one that sells chatbots is allowed that word too. Stale means not what you say about yourself today. It does not mean a word on a list.

What I sent her instead

A one-line correction, in the reply, before anything else: I am not an affiliate, I built it, the earlier line was an old template of mine and it is gone. Then the page she had asked for. She had written one word, so the correction was one sentence. The numbers for how that account's messages have done since are on the numbers page, bad weeks included, which is the only kind of receipt a story like this one deserves.