We didn't invent any of this.
We just built the app around it.
The comic format. Literal language at the bottom levels, idioms at the top. Adaptive difficulty per skill. None of these were our ideas. Each one is backed by a stack of peer-reviewed work going back to Carol Gray writing Social Stories on a typewriter in 1991. What we did was put them all in one place that an actual autistic 8-year-old will open twice.
since Carol Gray, 1991
we lean on directly
tracked, one by one
that move on their own
Six findings that shaped what's in the app
For each one, here's the research and the part of the app it ended up in.
Social Stories™
Carol Gray started writing short first-person narratives for autistic kids in 1991. Thirty years and several meta-analyses later, the verdict is steady: brief, predictable stories about real situations help kids understand them and feel less anxious entering them. Every panel in our app is a piece of one of these.
Gray, 1991 · Kokina & Kern, 2010Interest-led content
Generic stick figures don't get read twice. Stories built around things autistic kids actually care about do. The research is steady on this: interest-led material beats interest-neutral material, by a lot. So our library spans trains, animals, space, oceans, building, sports and the small everyday worlds kids live in.
Winter-Messiers, 2007 · Dunst et al., 2012Visual processing strengths
Lots of autistic kids think in pictures before words. Temple Grandin literally wrote a book called that. Comic panels carry the social information in the image: the person's face, what they're holding, what's behind them. Less working memory burned on parsing language; more left for figuring out the actual social puzzle.
Grandin, 1995 · Quill, 1997Adaptive difficulty (IRT)
Item Response Theory is the maths that holds together standardised tests. We use it differently. It watches how your kid answers on each of the 12 skills, separately, and adjusts only that skill's next story. So your kid can be solid on emotion recognition and still circling level 1 on false belief, and the app won't pretend otherwise.
Embretson & Reise, 2000Literal language, until it isn't
"Her face fell." For a literal-minded kid, that's a horror movie. At the bottom levels, we strip all of it. No idioms, no sarcasm, no "broke the ice." Sentences are short and concrete. Only around level 5 do figurative phrases reappear, and only because non-literal language is the skill being taught at that point.
Tager-Flusberg et al., 2005Generalisation through variation
The classic 1977 paper everyone in applied behaviour analysis cites makes one point: if you only ever teach a skill in one setting with one set of characters, the kid only ever uses it in that setting. Same skill, many different stories, many different worlds. That's what makes it stick. It's also why our library covers each skill with dozens of distinct stories, not three.
Stokes & Baer, 1977The newest part of the app, and the research under it
For younger and still-developing kids, the app now teaches the skills that come before stories: pointing, responding to a name, knowing faces. None of that is improvised either. Here's the work it rests on, and where it shows up.
Joint attention: the keystone skill
Pointing and shared looking are among the earliest skills to falter in autism, and among the most predictive of later language and social growth. Decades of work show joint attention can be directly taught. That's exactly what the Teaching Pointing activity does, using your own family photos.
Mundy et al., 1990 · Kasari et al., 2006Parents as the intervention
The landmark PACT trial coached parents to interact differently with their child, and the gains were still measurable nearly six years later, the first randomised evidence of lasting symptom change from early intervention. Every Foundations activity is parent-coached by design.
Green et al. (PACT), 2010 · Pickles et al., 2016Learning the natural way
Naturalistic Developmental Behavioral Interventions, the field's modern consensus, embed learning in the child's own routines, follow the child's lead, and reward effort gently instead of drilling. We pair that with the child's real interests, which measurably lifts engagement.
Schreibman et al., 2015 · Boyd et al., 2007Real faces, real names
Children on the spectrum learn and retain better from real photographs than from generic cartoons, and responding to one's own name is a foundational skill that can be deliberately practised. So Foundations uses the family's own photos and the parent's own voice, recorded once.
Carter & Hartley, 2021 · Nadig et al., 2007From "open the app" to "done", in about three minutes
The structure is the same every time. Predictability matters; surprises are for birthdays.
Set up the kid once
First name, age, the three things they love, anything sensory you want us to know. That profile flavours every story from then on.
Read the comic
Three or four panels. Same kid your kid recognises, same world they're obsessed with. A real social moment, at their current level.
Answer the one question
Multiple choice. Tap an answer. The right one shows up underlined; a short explanation tells them why, not just that they got it.
The app moves the bar
That answer updates one of the 12 skill scores. The next story is calibrated to where they actually are now, on that specific skill.
You see what they see
The Parent section shows accuracy per skill, current level, day streak, and every miss with the right answer. PDF export if a therapist asks.
Three minutes, done
We kept sessions short on purpose. A daily three-minute session beats a Saturday hour for actually moving the needle.