Article

I Built My Daughter an AI Speech Friend (Founder Origin)

It was a Tuesday in February, maybe 8 p.m. My daughter was sitting on the living room floor in front of my laptop, and the prototype was running. A goofy cart

It was a Tuesday in February, maybe 8 p.m. My daughter was sitting on the living room floor in front of my laptop, and the prototype was running. A goofy cartoon character named Buddy said, "Ooh, what's that?" She held up a plastic giraffe. She said, "Raff." Buddy said, "A giraffe! Big giraffe." And she said, "Big raff," and laughed. My partner, Sarah, was standing in the kitchen doorway. She looked at me. I looked at her. Neither of us said anything for a second. That was more spontaneous speech than we'd heard from our daughter in six or seven days. Sarah said, quietly, "What is that thing?" I said, "I don't know yet." But I knew.

That's the middle of the story. Let me back up.

She Had Two Words, Then She Had None

I was a software engineer when my daughter was born. Comfortable career, decent money, the kind of stability that had taken real work to build after a pretty unstable childhood. I knew how to ship products. I knew how to manage teams. None of that prepared me for what happened next.

Her first year was textbook. Babbling on schedule. Two words by twelve months. We were a normal first-time-parent household, anxious about all the usual things.

Around fifteen months, the words disappeared. Eye contact got sparse. She had rigid preferences about how things should go and would melt down when they didn't. She lined up toys with an intensity that wasn't random.

I recognized the signs. My partner did too. We didn't say "autism" out loud for weeks. We watched. We hoped. The hoping felt productive and also felt like lying.

At the eighteen-month well visit, I brought it up with our pediatrician. She told me to wait. I left that appointment feeling worse than when I'd walked in. The one professional whose job it was to spot developmental concerns had just told me there was nothing to see.

I waited two more months. Then I called early intervention myself.

Someone Finally Listened

The evaluator came to our house. She watched my daughter play for an hour, asked us questions, used her structured assessment tools. At the end she said, "She is autistic. We will get her speech therapy. We will get her OT. We will start tomorrow."

I cried in my car in the parking lot of the early intervention office after dropping off the paperwork. Not because of the diagnosis. Because someone had finally heard us, and the help was immediate, and we didn't have to fight for it.

The shape of our life changed overnight. Two SLP sessions a week. One OT session. Home practice. Reading. Sensory bins. Floor time. We were now an ND family.

Two Hours Out of 168

The early intervention SLP was excellent. Play-based, naturalistic, child-led. She trained me to carry the work into our daily life. The techniques I've written about throughout this site (speech therapy at home for autistic kids) come directly from what she taught me.

Here's the thing, though: two SLP sessions a week is two hours of clinical work. The other 166 hours are on us. Even when I was doing home practice religiously, there were gaps. Days I was sick. Evenings I was wrecked from work. Times my daughter needed practice and I just wasn't available, physically or mentally.

So I went looking for apps. Something that could do even a fraction of what our SLP did. Something to fill the dead air between sessions.

I tried everything. Speech Blubs: drill-based, she was bored by session three and refused to open it after that. Articulation Station: designed for older kids working on specific phonemes, not a non-speaking two-year-old. Otsimo: closer, but still structured, still drill-shaped, still built for a different kind of kid. The talking flashcard apps: drill, test, drill, test.

Even the ones that marketed themselves as "AI" were really decision trees with recorded voices. The branding said one thing. The actual product was the same pre-scripted educational app that had existed for a decade.

None of them did what our SLP did. They didn't follow my daughter's lead. They didn't pause and wait. They didn't narrate her play. They didn't expand on her sounds. They quizzed her. She shut down. Every time.

The Technology Caught Up to the Need

I'm a software engineer. I'd been watching large language models develop. By 2023, and more so into 2024, the technology had gotten good enough that an LLM could plausibly hold a play-based conversation with a kid in real time.

This was genuinely new. The previous generation of voice assistants (Siri, Alexa) were command processors. They answered questions. They didn't converse. The new LLMs could actually engage, could respond to context, could follow a child's lead the way a human play partner would.

I started prototyping. Built a small version that could listen to a kid speak and respond in character, with the pacing and techniques of a play-based therapist. Tested it with my daughter.

That was the Tuesday night in the living room. "Big raff." Sarah in the doorway.

I spent the next year building it properly.

What I Decided This Had to Be

A few constraints I set before writing a single line of production code:

ND-affirming from the ground up. No deficit framing. No "fix" language. The character meets the child where they are.

Play-based, not drill-based. The interactions follow the principles of NDBI (naturalistic developmental behavioral intervention). No flashcard loops. No quiz mechanics.

Pacing tuned for ND processing time. Slower than typical kid apps. Longer pauses. Less rushing. Most children's apps are designed to maximize dopamine hits per minute. Ours is designed to give a kid room to think.

Voice data privacy. Not used to train external models. Parents control retention.

Honest positioning. Not a replacement for an SLP. Not AAC. A practice companion. That's it.

And honestly, the positioning constraint has been the most expensive one. A more aggressive company would claim this app helps kids talk faster, replaces therapy, addresses speech delay at its root. I refuse to make those claims because they'd be dishonest. The result is slower growth than the competitors who overstate. I'm okay with that. (Most days.)

Who Buddy Is

Buddy is the AI character in LittleWords. He has a personality. He plays with the child. He follows the child's lead. He uses techniques from play-based speech therapy: parallel talk, self-talk, expansion, expectant waiting.

He is not a clinician. He is not AAC. He is not a replacement for human therapy. He is ten minutes a day of play-based practice, designed with input from an SLP, voice-tuned through testing with ND kids, and built with COPPA compliance in mind.

He is what I wished existed when my daughter was little. So I made him.

The Stuff Nobody Warns You About

Building a kids' app is harder than building an adult app. It's like the difference between renovating a house and building one to earthquake code. The safety bar is higher. COPPA, parental controls, content moderation, voice data handling. All of it adds engineering hours that don't show up in the product's visible features.

The autism community has been burned. Repeatedly. By apps, by therapies, by influencers selling miracle protocols. Many of the parents I talk to have purchased things that didn't work. Trust takes time, consistency, and a willingness to hear skepticism without getting defensive.

The SLPs are watching. Some are excited. Some are skeptical. The skepticism is healthy. I want them to push back on anything we get wrong. I want them to use it with kids if it fits, not because my marketing was persuasive.

And the personal cost is real. Building a company while raising an ND kid is intense in a way that doesn't show up on a pitch deck. My partner has carried more than her share. I owe her the world.

What This Is Not

It is not a cure. There is nothing to cure. My daughter is not broken. She is autistic. She is herself. The app supports her language development. It does not change who she is.

It is not a substitute for an SLP. If you have access to therapy, use it. Use this alongside it.

It is not AAC. If your child needs AAC, get a real AAC system. We are clear about this and will stay clear about it.

It is not a screen time maximizer. The recommended session is ten minutes. No engagement tricks, no autoplay loops. We're not trying to keep your kid on the app longer than they should be.

Where Things Stand

LittleWords is in the waitlist phase as I write this. Founding Family pricing is $49 lifetime. Launch pricing will be $19 a month. We're accepting families slowly while we make sure the product actually works well for the kids using it.

If you want in, the waitlist is open. If you don't, that's fine. I'll keep writing this blog and building this product regardless. The mission is the work. The customers are how we sustain it.

If Your Kid Hasn't Been Evaluated

Get an evaluation. Early intervention in the US is free. The number is on your state's health department website. Don't wait for your pediatrician to bring it up. Call them yourself. It's the single most important thing I did, and I almost didn't do it.

FAQs

Are you a real dad with a real daughter? Yes. Her name goes on the team page when she's old enough to consent to it being there. The privacy line is hers, not mine.

Did you really build this yourself? I led the build. I have a team. The engineering is mine and the team's. The clinical design belongs to the SLP. The brand is mine. The decisions are mine.

Will the app keep getting better? Yes. We ship improvements regularly. The underlying AI will keep improving over the next several years. The whole product gets better as we learn what works for real kids in real homes.

Why is the price what it is? $49 lifetime for Founding Families is below cost for the first thousand users. We set it there to attract early families who'd help us shape the product through real feedback. Launch at $19/month is what we need to sustain the business. We're not trying to be the cheapest option. We're trying to be the honest one.

Is there an actual person to talk to if I have questions? Yes. The contact email goes to a real human (me, until we're bigger). I read everything.

---

Related reading: Autism dad hub · Speech therapy at home for autistic kids (pillar guide) · How my daughter found her words · Being an autism dad

Related Little Words guides

Important: Little Words is educational support for home practice. It is not a medical device, not an AAC replacement, and not a substitute for a licensed speech-language pathologist, pediatrician, or developmental evaluation.