The Tone Check started as a side note in a meeting. We were three weeks into building dip, and someone — I think it was Pauline — said, half-laughing, "you know what would actually help, is if there was something that read your message before you sent it."
We were not going to build it. The roadmap was tight. The calendar, the expenses, the messages: those were the three things, and the Tone Check was a fourth thing.
Then a parent we were testing with sent us a screenshot of a message exchange that ended badly. The first message was a calm question. The second was a defensive answer. The third was a sharper version of the question. By the seventh message, neither parent could remember what they had originally been asking about.
We kept that screenshot pinned to the wall for a week.
What it actually does
The Tone Check is a small text box. You paste in the message you are about to send, and it tells you how it will probably land.
That is the whole tool.
It does not rewrite the message for you, unless you ask. It does not stop you from sending. It just reads the message back to you, with the kind of attention you would have given it yourself if you had time. Most parents do not have time. Most parents are typing the message at 8:14 in the morning, with the school bag half-packed and the cereal half-eaten, and they hit send.
The Tone Check buys you four seconds.
What we kept arguing about
We argued about whether to suggest a softer version. Half the team thought yes, that's the point. Half the team thought no, that crosses a line — the parent should write the message, the tool should only reflect.
We landed on: the tool offers a softer version when you ask for it, and only when you ask. The default is reflection, not rewriting. It turns out four seconds of "this might land sharply" is enough for most parents to find their own gentler words. The rewrite is for the cases where you genuinely cannot.
We argued about the model. We tried four. The one we shipped is the one that produced the fewest false positives. Co-parents are sometimes terse with each other for legitimate reasons, and a tool that flags every short message as "harsh" becomes a tool nobody uses.
We argued about whether to count usage. We do not, beyond the rate limit. The Tone Check is not a feature we are trying to grow. It is a feature we are trying to make so quiet that you forget it exists, until you need it.
What we did not expect
We did not expect parents to use it on messages from the other parent.
We thought of it as a draft tool — paste your own draft, get a read. But what came back from the first batch of users was: "I pasted in a message my ex sent me, and the tool told me it was actually neutral, not the attack I read it as."
That use case never crossed our minds. We left the tool unchanged, because the same mechanism works for both. A message reads the way it reads, regardless of who wrote it. If the Tone Check reads a message as calm and you read it as a threat, the gap between those two readings is information.
A small thing, on purpose
The Tone Check is the smallest tool in dip. It is one screen. It does one thing. It saves you four seconds.
But four seconds is a lot, when you have been trying to get the morning out of the door, and the message you are about to send is going to set the tone of the day.
That is why we built it.