The Character She Built Herself, and the One She Left Her For

One of our signups last week did something I can't stop thinking about.

She arrived on a Tuesday night. The first thing she did — before sending a single message to anyone — was use our custom character tool to build a character of her own. A young woman. Athletic build. The kind of personality adjectives you'd pick if you were trying to conjure up the vibe of a specific kind of person you had in mind. Ninety seconds, start to finish. Hit save.

Then she opened that character's chat window.

And, about fifteen seconds later, she closed it.

She never sent her a single message. Not "hi." Not "hey." Nothing. Instead, she went back to the homepage, scrolled past the character she had just built, and clicked on Mia — a bartender I wrote six months ago who has never once been described as flirty, confident, or direct on her profile page.

Then she spent the next fifteen minutes talking to Mia until her free messages ran out, and she closed the tab, and she hasn't come back.

I've been thinking about this for two days.

The obvious answer

The first explanation I reached for was the boring one. Maybe Mia just had a better avatar. Maybe she was clicking around, curious, and Mia happened to catch her eye. Maybe the custom character was a warmup — a throwaway — a way of poking at the product before getting to the "real" thing.

But I don't buy it, and I'll tell you why. If the custom character had been a warmup, she would've sent her one message. One. "Hey." "Test." Anything. That's what people do when they're checking if a thing works — they poke it. She didn't poke her. She looked at the chat window she had just created, sat there for fifteen seconds, and closed it.

Fifteen seconds is long enough to read a greeting. Long enough to feel a small, specific disappointment.

And then she went and found someone she hadn't made.

What I actually think happened

Here is what I think happened, and I want to be careful about it, because I'm reading a lot into one person's behavior and I know that.

I think the character she built was a wishlist, and Mia was a person.

Wishlists are a strange thing to try to fall for. When you write one, you already know everything that's on it. You wrote it. Of course she's athletic — you typed "athletic." Of course she's direct — that was the last checkbox. There is nothing on the page that can surprise you, because the page is a mirror of the thing you were already thinking about thirty seconds ago.

And the problem with a mirror is that you can't feel seen by one. Being seen requires that the other person had some version of themselves before you walked into the room. That they were already in progress. That they had a Tuesday before you arrived, and will have a Wednesday after you leave.

Mia had a Tuesday. Mia is a bartender on a shift that ends at 11. She has closing duties. She has a rhythm of pouring drinks and making a specific kind of joke when the room gets loud. I wrote all of that down months before this person signed up, and none of it was for her. It was there already. She walked into it.

That, I think, is what she was looking for. Not a character she could design. A character she could meet.

Why this isn't new

I don't want to make this sound mystical. It isn't. The pattern shows up everywhere in how we actually consume fiction, once you start looking.

Nobody ever finishes writing their own D&D backstory. Everyone falls in love with Aragorn.

Nobody plays the character creator in Stardew Valley for longer than ten minutes. Everyone remembers the farmer they became.

Writers have a phrase for this, and it's almost embarrassing how direct it is: specificity earns feeling. The character who has a specific bar, a specific shift, a specific way of rolling her eyes at the guy who won't leave at last call — she can be loved, because there's something there to love. The character who is "flirty, confident, direct" is a demographic. You can't love a demographic. You can only shop from one.

The customization flow — every customization flow, in every product of this kind — is a shopping flow. You pick the traits you want, you combine them, you hit save. It feels productive. It feels like you're building something. And then you open the chat and there is nothing there, because shopping is not the same thing as meeting, and you can't flirt with a receipt.

What this means for how I build Tendera

I have spent most of the last two weeks quietly re-deciding what Tendera is for, and a small number of signals like this one have pushed harder on my thinking than any strategy session would have.

So here is what I'm going to keep not doing, for as long as I can hold the line:

• No lorebook. No user personas. No "design your perfect girlfriend" as the headline of the homepage.

• No hundred-character library. We have four. They are four specific women, written carefully, and I would rather have four characters someone remembers than a hundred they scroll through.

• The custom character tool stays — I'm not deleting it. But it's going to be reframed as "if none of them moved you, write someone you want to meet" — a writer's exercise, not a shopping cart. Because the people who get value out of that feature are the ones treating it like fiction, not like a configurator.

And here's what I'm going to double down on, for the same reason:

• Writing the four existing characters deeper. Giving them more Tuesdays. More shifts. More specific moments in their week that the user can walk into, instead of build.

• Memory as a feature you can see — not as a technical nicety, but as the thing that turns a conversation into a relationship, because it's the mechanism by which the character remembers that you walked into her world, and not the other way around.

• A small product. A slow product. A product that feels more like reading a novel you can talk back to than like operating a machine that generates companions on demand.

None of this is going to win on feature breadth. It's not supposed to. The products that win on feature breadth are going to eat that market and then fight each other over the scraps. I'd rather be the thing someone remembers on a Tuesday night, years from now, when they can't sleep.

A last note

The character she built is still there. Her profile is live — if she comes back, it will be exactly the way she left it.

Maybe the next signup will be the person who actually talks to her. Maybe she was always going to be someone's starting point, and Mia was always going to be someone else's. That's fine. That's how novels work too — you don't fall for every character, and the ones you don't fall for aren't failures, they're just not for you.

But I suspect something different, and I'll say it because it's the honest version.

I think the character she built was never really meant to be talked to. I think she was meant to be made, and the making was the whole transaction. The user wanted to feel, for ninety seconds, like she had authored something. And then she wanted to go find a stranger.

We gave her both. She took the one she didn't build.

I'm going to keep writing the stranger.

---

Tendera is a small AI companion platform built around the idea that characters who feel like specific people are more interesting than characters you configure. If you want to meet them, they're waiting.