She almost sent the email. Then she did this instead.

Today's Tip: Never send a hard reply after 10pm again

The best thing Claude does isn't writing your emails for you. It’s stopping you from sending the one you already wrote.

Let me explain…

Suzy Nakashima almost sent an email last Tuesday night that she would have regretted for a long time.

She’s 63. Works part-time at the reference desk at the Des Moines Public Library, three days a week, mostly helping patrons find books and print things off the internet. Warm person. Steady. Not somebody who fires off angry emails.

But her son-in-law had sent a note that afternoon about Thanksgiving. It was polite on the surface. And it wasn’t polite underneath. Something about how the drive to Des Moines is a lot with the kids this year, and would Suzy and her husband consider coming to them instead… and by the way, could Suzy not bring the sweet potato dish, because the kids don’t like the marshmallows.

Suzy read it twice. Then she started typing a reply.

By 10:30 that night she had three paragraphs. All true. All fair. And in tone, a small nuclear device.

Here’s what she did instead of hitting send. She opened Claude on her laptop, pasted her whole reply into the chat, and asked one question. Not “make this better.” Not “rewrite this for me.” Something more careful.

She asked Claude to read the message back to her the way her son-in-law would read it… and to tell her honestly what the message was doing under the words.

Claude came back with something Suzy did not expect. It said the message was, in one sentence, “You’ve hurt my feelings and I want you to know it.” Not a discussion of Thanksgiving. Not a counter-offer on the sweet potatoes. A wounded response wearing the clothes of a logistical reply.

It also flagged three lines specifically. One said, “I understand you’re very busy.” Claude noted that phrase, said in that context, reads as sarcasm no matter how it was meant. Another line said, “The kids have eaten my sweet potatoes for twelve years without complaint.” Claude said this reads as a challenge, not a comment.

Suzy read that back and thought… yes. That’s exactly what I was doing. I just couldn’t see it while I was typing.

She spent the next fifteen minutes cutting the message down to five sentences. She kept the honest part — that she was hurt by the request, that Thanksgiving at their house had meaning for her, that she was open to talking about it. She cut everything sharp.

She sent it Wednesday morning after coffee. Her daughter called that same afternoon.

I know what you’re thinking. “I don’t need a chatbot to tell me my email sounds angry. I can tell.”

Sometimes you can. And sometimes you’re standing too close to see it. The whole reason a message gets composed at 10:30 at night after a hard day is that you are not, at that moment, the best judge of how it reads. You wrote it in one emotional state… and it will be received in another one entirely, by someone whose day you know nothing about.

Claude doesn’t know your family history. That’s exactly why it’s useful for this. It reads the words. Not the twenty years of context you’re carrying into them.

The old way to catch a bad email was to sleep on it. That still works. But most of us don’t sleep on it. We send it, then we lie awake.

This is the same idea, done in three minutes, with a second set of eyes.

★ Quick Wins

Skill Builder: The Read-It-Back Prompt

There’s a habit worth building with any AI tool, and it’s simple. Before you ask it to rewrite something, ask it to read it back to you first.

Most people go straight from draft to “fix this.” That skips the most useful step in the whole process. When you ask an AI to reflect your writing back — what it’s saying, how it’s landing, where it turns sharp — you almost always find that you already know how to fix it. You just needed to see it from the outside.

This works on more than emails. It works on a text you’re about to send your sister. It works on a review you’re leaving for a business. It works on a note you’re putting in your grandchild’s birthday card. Anywhere the words matter more than the information.

The prompt is short and can be saved once and reused for years. It doesn’t ask the AI to be creative. It asks the AI to be a careful reader.

Save this prompt somewhere you can find it. A note on your phone. A sticky note on the fridge. Anywhere you can reach when it’s late and the reply feels urgent.

The value of AI for writing isn’t that it types faster than you. It’s that it can read what you wrote without any of your feelings about it. That’s the second set of eyes we’ve always wished we had.