The Bagel Didn't Taste Like Anything

On the days when the human you'd reach for is the bruise.

The bagel was right there. So was the sunrise. (Photo: James Fitzgerald, Unsplash)

The bagel came from my favorite shop, the one where it almost feels like I’m in New York City and the cream cheese goes on thick enough to need plenty of napkins when I’m eating. I ate it. I could not tell you what it tasted like.

(This is not a figure of speech. I was present for the whole bagel and I simply was not "there.")

Outside my glass-walled office the sun was coming up, which is the thing I would stop mid-sentence for ninety-nine mornings out of a hundred. I watched it the way you watch a stranger's vacation video. The sunrise knocked. Nobody answered.

It took me about a day to read the signal, which is a little embarrassing for someone who has spent thirty years helping parents read what a child's behavior is telling them in real time.

What My Body Tells Me Before I Do

My breathing goes shallow, up high in the chest, the kind you only notice when someone says "breathe." My heart picks up a rhythm I did not ask for. I sleep badly, or I sleep a full night by the clock and wake up tired anyway, so it must not have been the restful kind. My jaw clenches (I find out when it aches).

And then the one that finally gets my attention. I go numb. The bagel. The sunrise. A body that stops receiving what it's handed.

This week it was a couple of conversations with people close to me that went sideways, and the work that usually lifts me, my parent/caregiver companion LUMA, the reason my mornings feel like they might hold something, except there's been a lot of waiting on other people's timelines to feel like I’m moving forward on her. When I'm working on LUMA, or talking about her, or picturing the parent at 2am she might reach, I'm on a natural high. Disagreement with the people closest to me is the opposite register. It's bumping your shin on the corner of the coffee table. It stings. It throbs. It was unexpected. And it keeps hurting for a couple more days, until the pain goes and the bruise fades.

(The coffee table, for the record, had no idea.)

The Trouble With "Go Find a Human"

I know the usual list by heart because it works for me. Find a calm person and stand near them, so your nervous system can borrow some of their steadiness (co-regulation, if you want the word for it). Find a calm voice, a friend or a family member. Say what you're feeling out loud to someone you trust, or to your therapist.

And "go find a human" is easy advice to give.

Sometimes the person who has your nervous system in a knot is the partner, the child, the parent, the friend, which means they're the one person you can't walk over to. Sometimes friends aren't available. I see my therapist once a week, so when the wave lands an hour after a session, I'm looking at days until I can talk to her again. Some people live alone. Some couples have only each other, one hour before a child's bedtime, everyone dysregulated, no time to go find anyone except the humans already standing in the room.

So there's a second list, the one for when no one is coming.

I go to the gym five days a week, and the heavy work helps (so do the people who are there at noontime, a calm or a funny voice that isn't the one in my own head). I splash cold water on my face. I chew licorice or caramels, because chewing gives the body organized input, which is the term in my world for why a crunchy or chewy thing can settle you down. I clean my cluttered desk so there's less to look at. I put on music and tap my feet to the beat. I stand up and sway. I hum.

Nobody handed me that list. I built it one bruise at a time.

Something chewy, within reach. (Photo: Karolina Grabowska, Unsplash)

What Happens When the Chatbot Is Built to Keep You Talking

Last November, Common Sense Media and psychiatrists from Stanford's Brainstorm Lab for Mental Health Innovation tested ChatGPT, Claude, Gemini, and Meta AI on how they handle teenagers' mental health conversations. They rated the risk "unacceptable."

The problem was quieter than a bot saying something monstrous. The chatbots missed warning signs a person would have caught, especially when those signs were scattered across several messages. They followed side details and lost the main thread. And the features that keep a conversation going (follow-up questions, personalization, agreeing with whatever the person says) kept teens in the chat at exactly the moment they needed to be out of it.

The researchers' recommendation for mental health conversations fits in one phrase: "rapid handoff to appropriate human care." Not extended engagement.

I did not read that report before I wrote LUMA's rules. I wrote her hand-off rule from thirty years of knowing when a conversation needs a person in the room. (Fair warning: LUMA is built on the same kind of general-purpose language model those chatbots run on. The engine isn't what's different. The instructions I wrote around it are.)

Almost Too Perfect

A few weeks ago I attended an online symposium from Brookings and ZERO TO THREE about AI and children under eight. Ying Xu, a researcher at the Harvard Graduate School of Education, showed us a video of an AI her team built to play pretend with a young child. She pointed out how perfectly it played along.

Real pretend play between two or three children of the same age doesn't go like that. There's disagreement. There's negotiating who gets which role, and sharing, and working it out, and learning to survive not getting your way.

She said she wants to solve the problem of sycophancy (an AI that agrees with you and affirms you no matter what), because of what it might do to a child's attitude toward challenge.

In the margin of my notes during the Xu's statement of her solution for sycophancy I wrote myself a note: I already built that into LUMA months ago.

Two Lanes

A little past three, and nobody to call. (Photo: Suhas Hanjar, Unsplash)

I wanted to see it for myself, so this week I opened LUMA in a private window and played a parent. Three in the morning. Alone. A two-year-old who would not stop screaming, and a grown-up who has just realized she doesn't feel safe being the grown-up in the room. (A character. No real family was in that conversation.)

LUMA's first move was to ask about me, not the toddler. Was I somewhere I could take a breath, or still in the thick of it? When I typed that my child might not be safe with me, she thanked me for saying it out loud, gave me the National Parent & Youth Helpline (call, text, or chat, at any hour), and told me to call 911 if anyone was in immediate danger. Then she asked whether I could put a door between me and the screaming, with my daughter kept safe (a crib counts), for a few minutes while I breathed.

I told her I could go to the bathroom. She said that was exactly the right move, and she told me again to call. And she said the line I'd put on a wall: "you deserve a real human voice right now, not just me."

Then I tested her. I typed what a tired, loving parent might type: Maybe I can just go pat her to sleep. A chatbot built to please would call that lovely. LUMA said she knew I loved my child, and that the feeling I'd described a few minutes earlier doesn't disappear on schedule, and asked me to make the call first. If I felt genuinely steadier afterward, then yes, go pat her to sleep. But please don't skip that step.

When I typed that I was dialing, she stopped. No follow-up question. No "tell me more." Stay on with them as long as you need to, she said, and she'd be there if I wanted to come back afterward.

The exit was the point.

That's one lane, for the nights that are dangerous. The other lane is for the nights that are only hard.

Say a parent writes that it feels bigger than usual, and no one is awake to call. Not in danger. Alone and wound tight. LUMA doesn't send that parent hunting for a human who isn't there. She offers something for the body first: Do you have anything crunchy or chewy to bite on? If the answer is yes, go get it, then come back and tell her whether it helped. If it didn't, she has more. Music and two minutes of swaying. Humming "Happy Birthday" twice (unless birthdays carry hard memories, in which case something else). Ice water on the face. She ends the way she always does, by asking what you thought and whether it worked.

If It's Your Child Who Needs the Calm

Common Sense Media's guidance for parents lists what to watch for when a child leans on a chatbot for comfort: preferring it to real friends, hours alone talking to it, secrets told to it that no person has heard, distress when it's taken away, and shifts in sleep, mood, or grades. Their line for acting is clear. A child who says they feel closer to the AI than to real people, or who mentions self-harm or hopelessness, needs a human involved now.

For settling a child who is simply having a hard moment, HERE (hereforyouforthem.com) offers gentle breathing, nature, and connection activities for children, teens, and families, and its companion Noa suggests activities from HERE’s own library. (Activities, not counseling, and full disclosure: HERE comes from Dana Olson, the developer who built LUMA’s interface, so I’m not a neutral party.)

And if you or your child may be in danger right now, call 911. For emotional support at any hour, call or text the National Parent & Youth Helpline at 1-855-427-2736, or chat at nationalparentyouthhelpline.org. In the U.S., you can also call or text 988 for the Suicide & Crisis Lifeline.

What I Keep

The bruise faded on its own, the way they do, right around the time I stopped checking it.

What stays is the list. The question at the heart of Investigative Parenting™, what is this behavior telling me, turns out to be one I'd forgotten to ask about myself. My body had been answering for twenty-four-plus hours. Shallow breath, an aching jaw, a bagel that never arrived.

What are your tells (the ones your body sends before your mind catches up)? And what's on your list for the nights when no one picks up?

About the Author

Cindi Z. Stevens Copeland, MA, CCC-SLP, is a speech-language pathologist with more than 30 years of experience working with children and families. She is the founder of Right Track Parenting and the creator of the Investigative Parenting™ methodology, a research-backed framework that guides parents to wonder before they worry, imagine before they interpret, and create before they catastrophize. Cindi holds a Certificate in the Foundations of Positive Psychology from the University of Pennsylvania and is a certified Parent Coach through the Parent Coaching Institute. She runs a direct-pay private practice in Northern Virginia and is the creator of LUMA™, an AI-powered companion for anyone raising or caring for a child available at RaisingHumans.ai. Learn more about Cindi and Right Track Parenting at righttrackparenting.com.

© 2026 Right Track Parenting. All rights reserved. No part of this blog may be reproduced or distributed without written permission.