The Diagnostic
One of your own threads, read against the loop. Paste it, read what comes back, and decide for yourself whether it looks like how the conversation went.
Three threads are free. After that it asks for an email address, so it can tell you what comes next; there is no meter and nothing to buy on the page.
What you paste
Any thread where work was being arranged: an email chain, a run of chat messages, a text exchange with a contractor. Short is fine; a dozen messages is plenty. It helps if the thread has at least one ask in it, since the Diagnostic reads requests and the answers to them and has little to say about a thread that is only talk. Names can stay in or come out as you prefer; the Diagnostic strips names it recognizes before the thread goes anywhere, and keeps nothing afterward either way.
What comes back
A reading of the thread in three parts, in the order they happened. Here is what it makes of a short, invented exchange between an estimator and a colleague:
Priya, Monday 9:12 — Tom, I need the revised cabinet estimate for the Feeney kitchen, itemized with the pantry broken out, by Thursday noon. The Feeneys are in at 2.
Tom, Monday 9:40 — Sure, I'll try to get to it tomorrow.
Priya, Thursday 12:15 — Tom? The estimate?
Tom, Thursday 12:20 — Almost. Give me an hour.
The moves. Each message is marked with what it does in the loop, and the three parts of a promise are checked against the request.
The loops. The moves are tied together into the promises they belong to, and each one is placed on the loop.
What was left unsaid. A short plain reading, written to the person who pasted the thread, never about a third party's character.
Priya's request had everything a stranger would need to check it. Tom's answer took none of it on: "tomorrow" was offered, "I'll try" kept it from being a promise, and nobody said so at the time, so both of you carried Thursday noon without an agreement under it. The thread ends with the first real date Tom has named, an hour after the deadline. If this is a thread of yours, the question worth sitting with is whether, on Monday, either of you knew the yes was a hedge.
That last part is Clem's voice. Clem is the coach you meet properly on the next rung, in your inbox; here Clem only reads, asks and offers. Clem never scores, never says "you should," and never says "I will," since Clem is an agent and agents do not make promises.
What it does not do
It does not keep your thread. The text goes from the page to a function that reads it once, with a language model behind it, and returns the reading; it is not written to disk, logged, or used to train anything. The site counts how many threads have been read, and that number is the only thing recorded. The full statement is short and says exactly this.
It does not rate you or anyone in the thread. There are no scores, grades or rankings anywhere in the product, and hedging is shown as a way of operating with uses and costs, not as a mistake.
It does not read your inbox. That is the next rung, and only if you ask for it.