This Week I Almost Quit

I’m going to tell you about my week. Not the curated version. The real one.

I almost shut Enchantably down. Not metaphorically. Not in the dramatic, 2am, pour-a-glass-of-wine way that founders talk about at conferences when the story already has a happy ending. I mean I sat at my kitchen table on a Wednesday night and seriously considered walking away from the thing I’ve poured the last year of my life into.

Here’s what happened.

Monday

A customer messaged me. She’d been trying — really trying — to get her child’s book right. She’d gone through 140 credits. A hundred and forty. Each credit costs me real money — compute, image generation, the AI infrastructure I wrote about in a previous post. She’d been regenerating illustrations over and over, refining and adjusting, because the images weren’t coming out the way she’d hoped. A lot of redos. A lot of patience spent on a product that was asking too much patience of her.

She said she was disappointed. She said she wouldn’t proceed with refining. She was done.

I stared at that message for a long time. Not because of the cost, although that stung. Because this was a parent who came to Enchantably hoping to make something special for her child, and the experience I built wasn’t good enough. She gave it 140 chances. That’s not a casual user giving up after one try. That’s someone who wanted it to work and it didn’t.

That one sat in my chest all day.

Tuesday

Another customer reached out. The last page of her book had disappeared. Just — gone. The story had been built, the illustrations refined, and then the final page — vanished.

This one had an extra twist: it wasn’t a children’s book. This customer was making a personalized storybook as a gift for her bridesmaids. A novel, completely unexpected use of Enchantably that I hadn’t imagined when I built it — adults recreating their memories together, their friendships, their shared moments, illustrated in a book they’d keep forever.

It was one of the sweetest things I’d seen someone do with the platform. And I’d broken it.

When I dug into the bug, I found the cause: it was me. The disappearing page was a side effect of the image regeneration fix I’d been working on. In trying to make illustrations better, I’d knocked the last page off the shelf. The thing I was building to solve Monday’s problem had created Tuesday’s problem. Founder life in a nutshell.

But beyond the bug, this customer made me think about something bigger. She was trying to recreate real memories — specific places, specific faces, specific moments between real adults. That is exponentially harder for AI to nail than a trip to space or a visit to Tooth Town. Fantasy is forgiving. Memory is not. If the park bench doesn’t look right, if the dress is the wrong color, if the friend’s hair isn’t quite how she remembers — the magic breaks. The gap between what the AI generates and what lives in someone’s heart is wider for real memories than for imagined adventures, and I hadn’t fully reckoned with that.

It’s a problem I want to solve. Because the idea of Enchantably as a place where adults can preserve their stories — not just children’s bedtime adventures but the moments that made us who we are — is too beautiful to abandon just because it’s hard.

Wednesday

Wednesday was the worst.

I’d already had a brutal week at my day job. Another firedrill. Another intense set of deliverables dropped on my desk with an impossible timeline. I’m a Practice Architect — I’m the person they call when something is on fire, and something is always on fire. By Wednesday evening I was running on caffeine and stubbornness, with nothing left for anyone or anything.

And then my 5-year-old looked at me and said: “You don’t play with us.”

Not angry. Not whiny. Just... factual. The way kids say things that are true, without padding or preamble. A simple observation from a person who has been watching his mother disappear behind a laptop screen for the better part of a year.

He’s right. He’s absolutely right.

I have been behind a screen for most of the last year. Building Enchantably. Working my day job. Answering emails at dinner. Refining story arcs after bedtime. Debugging code on weekends. I wrote a whole blog post about the conveyor belt — about wanting to be more present, about getting lost in what my kids are doing instead of rushing them through the day — and the truth is, I’ve been doing the opposite. I haven’t been rushing them. I’ve been absent.

It’s summer. The days are long. I should be using them — playing in the yard, splashing around in the pool, driving to the beach with the windows down and nowhere to be. Instead I’m at home, working until late on my day job and then trying to squeeze Enchantably into the hours that are left, which are the hours that should belong to my kids and my sleep and my sanity.

That sentence from my son — “you don’t play with us” — landed harder than the missing page and the 140 credits combined. Because the whole reason I built Enchantably was for my kids. And somewhere along the way, building the thing for them became the thing keeping me from them.

I sat with that on Wednesday night and I thought: maybe it’s time to stop.

Friday

And then, just to round out the week — the AI broke.

Not in a dramatic, interesting way. In the slow, maddening, staring-at-error-codes way. The image model I use for illustrations exists in two tiers: 1K and 2K resolution, which are now generally available with full support and stable capacity, and 4K — the resolution I need for the beautiful, high-quality illustrations that make Enchantably books look like they belong on a real bookshelf.

4K is not GA yet. It’s been “any day now” for weeks. I keep checking. I keep waiting. Any day.

In the meantime, I’m stuck on the preview version. And here’s the cruel irony: once the model went GA at the lower resolutions, they toned down capacity on the preview tier. The servers I depend on for every single illustration got quieter. Not turned off — just... deprioritized. The queue got longer. The timeouts got more frequent.

So on Friday, I sat at my laptop trying to generate illustrations for a book, and what I got was this: timeout. Timeout. 429 — too many requests. Timeout. 429. Retry. Timeout.

Over and over. For hours.

There is a special kind of helplessness in being completely dependent on infrastructure you don’t control. I can optimize my code. I can refine my prompts. I can streamline the pipeline. But I cannot make Google flip the switch on 4K GA. I cannot will the servers into having more capacity. I cannot do anything except wait, and retry, and wait, and watch the error codes stack up like a monument to my own powerlessness.

I waited for days to pass. That’s the only strategy available. Try again tomorrow. Maybe tomorrow the queue will be shorter. Maybe tomorrow will be the day it goes GA. Maybe tomorrow.

Building on the frontier of AI means living with this. It means your product’s quality is tethered to decisions being made in rooms you’ll never enter, on timelines you can’t influence. It means some weeks you feel like you’re building the future, and other weeks you feel like you’re refreshing an error page and hoping.

This was a hoping week.

What my husband said

I told Ryan. All of it. The disappearing page, the disappointed customer, the cost of the credits, the thing our son said, the feeling of being pulled in four directions and failing at all of them. I told him I didn’t know if I could keep doing this.

He listened. He’s good at that. And then he said something that I haven’t been able to get out of my head:

“If we save one kid. Just one kid from a tragedy. It’ll all be worth it.”

He was talking about Safety Magic. The stories I cry every time I make. The ones that teach kids what to do in a fire, what to do if they’re lost, how to protect their body. The ones that make my children uncomfortable in a way that makes them think, that make them ask questions like “what if we’re on the second story,” that lodge safety rules in their heads through songs they hum without knowing why.

One kid. Not a million users. Not a viral launch. Not a metric on a dashboard. One kid who remembers the song when the smoke alarm goes off. One kid who knows their parent’s phone number because it was in the story. One kid who tells a trusted adult because a book gave them the language to say something feels wrong.

Is that worth the disappearing pages? The 140 credits? The Wednesday nights? The months of building instead of playing?

I think it might be.

Saturday

I got up on Saturday and I worked for five hours on Enchantably.

Not on marketing. Not on blog posts. Not on the parts of the business that feel productive and visible. On the unsexy, unglamorous, nobody-will-ever-know infrastructure that makes the product better.

I refined the way avatars get generated — the system that creates a child’s likeness in the illustrations. It was inconsistent. Sometimes the child on page one looked different from the child on page eight, and that tiny fracture breaks the spell. I rebuilt the logic until it held.

I overhauled the image regeneration pipeline. The customer who used 140 credits shouldn’t have needed 140 credits. The illustrations should have been closer to right on the first pass, and the regeneration process needed to be smarter, more responsive, more likely to get it right on the third try instead of the thirtieth. I made it better. Not perfect. Better.

And I added a new feature: manual illustration prompt editing. So that when a parent can edit the prompt the model receives before it gets sent. Less guessing. Hopefully better images.

Then I spent an hour on the body autonomy storyline.

I’m not going to say much about that one right now. It’s for my older son. It’s about learning that his body belongs to him and that he has the right to set boundaries — and that the people who love him will respect those boundaries. It’s also, if I’m honest, about me learning to respect his boundaries. To stop overriding his “no” when it’s inconvenient. To model the thing the story teaches.

That hour was the hardest hour of building I’ve done on Enchantably. Harder than any technical problem. Harder than any disappearing page.

And it was the hour that reminded me why I’m here.

The truth about this week

Here’s what I want to say to anyone who’s building something and having a week like mine.

It doesn’t get easier. The pages still disappear. The customers still leave disappointed. Your kid still tells you the truth you don’t want to hear. The day job still demands everything you have, and the thing you’re building on the side still demands everything that’s left, and “everything that’s left” is almost nothing, and you give it anyway.

I am not going to pretend this is sustainable. I don’t know if it is. I am tired in a way that sleep doesn’t fix. I am behind on my day job and behind on Enchantably and behind on being the parent I want to be. I am writing this blog post at a time when I should probably be doing any of those three things instead.

But Ryan was right. If one kid is safer because this thing exists — one kid who hums a fire safety song, one kid who knows their address because a story taught them, one kid who understands that their body is theirs — then the disappearing pages are problems to fix, not reasons to quit.

The disappointed customer is a signal to build better tools, not a verdict on whether this should exist.

And my son — my beautiful, honest, five-year-old son who told me the truth on a Wednesday — he’s the reason I’m going to figure out how to do this differently. Not stop building. But build with more margin. Build and also play. Build and also be the mom who gets on the floor and looks at the rock with the face and doesn’t check her phone.

I don’t have that figured out yet. I’m telling you that honestly. But I’m still here. The pages are getting fixed. The tools are getting better. The body autonomy story is going to be something I’m proud of.

And this week, for the first time in a while, I left my laptop in the house and sat in the yard and watched my kids run through the sprinkler and I didn’t time it.

That felt like a start.