Before a child finally speaks, what pitfalls do parents have to get through first?

A child not speaking is hard to notice at first, because no child can speak at birth. A year or two just goes by. By the time you and your wife notice your child should be talking but isn't, the pediatrician usually says "it's fine" first.

Then you keep noticing something's off, the doctor starts frowning, and pulls out a diagnostic questionnaire. After filling it out, you convince yourself you were just too pessimistic, that you filled it out wrong, that the evidence isn't really there. Until one day you see kids the same age as yours having full back-and-forth conversations right next to him, and you finally have to admit: something really is wrong.

Next comes an entire pit of "what do we do now." A parade of quacks will show up trying to sell you all kinds of things that don't actually work.

After knocking on every door worth knocking on, you'll eventually land at a speech therapist. This step actually helps — therapists know a lot of methods you don't. But what they can do is still limited, because a lot of it depends on the child's innate capacity. If the child is deaf, mute, or has a physical impairment, a therapist has plenty of ways to help them communicate. But if the block is in cognition itself, all you can do is drill it over and over and take whatever progress you get.

Why off-the-shelf AAC devices didn't work for his son

Speech therapy often gives kids who struggle to talk a device called AAC. It's basically a tablet full of words and symbols — you combine symbols into a sentence and the tablet speaks the words aloud. It's essentially a digital version of pointing at pictures to talk, barely more sophisticated than a file directory.

The author's son used AAC for about a year and never showed any interest — he'd fiddle with it for a few minutes, then drop it and go play with something else. There was one at school and one at therapy; the author never bothered getting one for home, because the kid just ignored it entirely. Every therapist and teacher agreed: he had no interest in the tool, and there was no forcing it.

The problem was the symbol system itself. The red octagon meant "stop," and his son had no idea — he didn't even understand what "stop" meant in that context. The arrows, the tiny stick figures, same story, incomprehensible. The very thing AAC asked of him was exactly the thing he struggled with most.

Swapping abstract icons for "his own life"

The author's day job is AI engineering (he describes himself as a Forward Deployed AI engineer). He sat down and, using vibe coding (intuition-driven coding: describing what you want to an AI in conversation and letting it generate the code directly, while you just review and tweak the details), spent two hours building a simple website with basic navigation, then generated images for a few hundred vocabulary words in ChatGPT.

He loaded the website onto a touchscreen laptop at home and handed it to his son. His son was instantly hooked. To him, it was his own life laid out like a picture book, like his favorite cartoon — he recognized himself in the images without any effort. He pressed grandpa's photo over and over, then said the longest sentence he'd ever said in his life: "I really love you a lot" — the exact phrase grandpa always said to him.

They photographed every food his son ate, taught him that "the photo of an outstretched hand" meant "I want," then taught him how to flip to the food list. For four years, the author had never been able to tell what his kid wanted to eat — now it suddenly became effortless: "I want orange," "I want peanut butter crackers," "I want French toast sticks." Which toy was upsetting him? Tap the photo, done. His son loves it when the author presses his head hard against his — so the author made a dedicated button that, when pressed, says in his voice, "I need a head squeeze." The entire product was built around this one child. These small-looking but massive wins left the author genuinely annoyed at himself for not doing this sooner.

The two things no off-the-shelf device had: data, and "teaching"

Once his son got hooked, the author started getting genuinely angry at the entire industry. He found that most AAC devices on the market lacked even the two most basic things.

So the author built both in: metrics, plus a teaching mode that actively taught his son. After two weeks, his son's ability to look at a word and pick the matching image out of a set improved by more than double. What he built was the loop below, running continuously.

He also gave the app a schedule: every so often it plays a song, and when his son hears it, he walks over on his own to play a matching game or watch a slideshow. The app asks him every 45 minutes whether he needs the bathroom. On the drive to appointments, the author taps a few things on his phone and his son's tablet switches into "preview mode," specifically teaching him what he's about to encounter at the clinic. When the author wants to ask his son something and the kid doesn't get it right away, he just types it into the phone app, and the words show up on his son's screen rendered in the symbol set he's familiar with, like hieroglyphics — his son gets it instantly, because the meaning shows up exactly where he's already paying attention.

After he started talking, came an "unhappy" kid

Three weeks in, his son became a genuinely unhappy kid — actually angry. Because he now had a "voice," but wasn't used to having one yet. Understanding came with responsibility attached.

Before, when he was hungry or upset, it was one vague emotion and his parents had to guess. Now they'd point at the tablet and say, over and over, "tap it, say it" — making him express it himself. For the first time, he had to take responsibility for "saying it." His speaking frequency went up too, roughly five times what it used to be.

That bite of waffle didn't come easy. The author had tried every trick to get his son to eat more variety, including "setting rules" — none of it ever worked, about as effective as trying to give a cat a pill. This time, they made a few small pieces of content in the app: a song about how great it is to try new foods, paired with photos of his son trying new foods himself, narrated in Dad's voice. And then, his son ate the waffle.

From "built for my son" to "usable by other families' kids"

Now the author is thinking about what else to add to help other parents. Turning a tool custom-built for his own kid into something any family can pick up in ten minutes means solving a whole pile of technical problems.

For instance, he first needs to build a generic taxonomy of early childhood vocabulary — a "scaffold." That way, when parents casually snap photos at home, the system can automatically sort them into fixed categories, which is what makes all the downstream automation possible.

He wants the parent experience to be as simple as possible: tell the system your kid's favorite color, pick a cartoon style your kid likes, maybe read a 30-second script to record your voice, upload a few photos — and the communication board builds itself. He absorbs all the complexity himself, so that even a mom juggling four kids and running ragged still has the bandwidth to set the whole thing up in ten minutes.

There's an even harder problem: how do you quantify whether a child has actually mastered something. It varies wildly by kid. The author's son is weak on joint attentionThe ability to share focus on the same thing with another person — like when you point at something and the child follows your gaze. It's a foundational skill for language and social development, and many autistic kids are weaker at it.: if the author manages to hold his attention during a test, he'll get almost everything right; if not, he'll just mash buttons at random until the test ends. The exact same set of questions can produce wildly different results.

His day job happens to involve something in a similar direction (currently in the final sprint before launch, which is also why he's been so busy lately). Once that's built out, he plans to let parents just talk to the app in plain language, and have the app adjust the interface in real time.

How much it costs, and who already wants it

The dedicated AAC devices mentioned earlier can be expensive. Buying the hardware directly from a vendor can run over seven thousand dollars; if you already own an iPad, a subscription is reasonably priced.

By the author's math, he can run this app for $9.99 a month, which also covers the cost of all the AI API calls needed to make it work as well for other kids as it does for his own son; if you want voice cloning, it's $19.99 a month, since keeping that feature always-on costs more.

His son's speech therapy clinic wants to use it on other kids, and so does the school. The author hasn't made a single pitch — this is what he means by "accidentally started a small business." There's still plenty of paperwork to sort out; for instance, he can't call it a "medical device," and he can't call it "AAC" either — in the marketplace, he can only describe plainly what it does, and sell it as that.

His goal is that once a family no longer needs to keep generating new images, he'll make it free, or at least keep a bare-bones free tier — because after a few thousand images have been generated for a kid, there's not much left to generate. This is the lowest price he can responsibly offer, since once it's live, he has to keep the service running for every single kid using it. He says in a few more weeks, other families will be able to try it out.

The author's day job is AI engineering. Instead of swapping in another set of icons, he spent two hours building a simple app that replaced every image with his son's own real life — not just any bagel, but "his" specific one; not just any toy, but "his" specific one — all rendered in his son's favorite cartoon's art style, with his own cloned voice narrating every button.

Once the content changed, his son was instantly hooked — he pressed grandpa's photo over and over, saying the longest sentence of his life: "I really love you." But just making it comprehensible wasn't enough: what actually doubled his recognition ability in two weeks was another thing most devices on the market are missing.

That "other thing" is actually two things: data (logging which button the child pressed and whether he got it right) and a teaching mode (drilling specifically what he hasn't mastered). Most AAC devices on the market have neither, since they were built for people who could already speak. The author added both, chaining them into a loop that runs every day.

Start with the cost. Getting a child a working communication tool the old way (dedicated AAC hardware) routinely starts at seven thousand dollars; this app runs $9.99 a month (or $19.99 with voice cloning) — a difference of hundreds of times over.