You asked whether we are just a copy of Duolingo and our biggest competitor. Short answer: on the surface, yes. Here is what the evidence says about that, four directions that would break it, and the one risk worth taking seriously before you pick.
Our audience is not an enthusiast community choosing an alternative lifestyle. Most of them were pushed. School environment is the top reason in US federal data at 83%. In the Queensland government's survey of 565 families, a disability or health issue sits at 61 to 64%, and bullying or negative influences at 57%.
Read one layer deeper and it gets sharper still. Inside that disability cluster, neurodevelopmental conditions are about 46% of it, which puts roughly 28 to 30% of all Queensland home-ed families on autism or ADHD specifically.
So close to a third of our Australian children are likely to be neurodivergent, and a large share of the rest left because school was distressing. Saturated colour, constant motion, celebratory noise and streak-loss anxiety are not neutral styling decisions for that group. They are load, aimed at the people least able to carry it.
Four findings, each one load-bearing for a design decision.
Seven curriculum brands were reached and audited by fetching their live sites: Euka, Timberdoodle, The Good and the Beautiful, Blossom & Root, Master Books, Build Your Library and Christopherus. Zero use mascots, streaks, XP, badges or arcade-bright palettes. All use either real photography of real families or restrained flat illustration. All lead with a values statement or a practical-relief promise, never an entertainment promise. Euka, our direct competitor and the most screen-based of the set, is also the most corporate-clean, and still has no gamification signalling anywhere.
So what: A streaks-and-mascot layer would not make us a norm-follower in this category. It would make us the outlier.
Duolingo's own design blog says the bright, bouncy style was developed in 2018 for Duolingo Kids, then migrated back into the flagship product. So we have copied a children's style second-hand, from an adult language app that borrowed it from its own children's spin-off.
So what: There is no deep pedagogical reason behind the aesthetic we inherited. It is a house style that travelled, attached to a free consumer app that converts 6 to 8 percent and optimises for daily return. We are a paid curriculum. Different business, different job.
US federal data (NCES 2024-113, the 2022-23 NHES wave, ~1.77M homeschoolers) on reasons rated important: school environment including safety and bullying 83%, moral instruction 75%, family life 72%, academic dissatisfaction 72%, religious instruction 53%, special needs 21%.
Australia is sharper. The Queensland Department of Education's Home Education Unit surveyed 565 families in 2022: a better-learning-environment cluster at 95%, a disability or health cluster at 61 to 64%, and bullying or negative influences at 57%. Within that disability cluster, neurodevelopmental conditions account for roughly 46%, which works out at about 28 to 30% of all Queensland home-ed families citing autism or ADHD specifically.
So what: The customer is not an enthusiast building an alternative lifestyle. They are usually a parent whose child was struggling, who is now anxious and improvising, and who is worried about the paperwork. And in Australia there is close to a one-in-three chance the child is autistic or has ADHD.
If somewhere near a third of our Australian children are autistic or have ADHD, then saturated colour, constant motion, celebratory noise, competitive leaderboards and streak-loss anxiety are not neutral styling choices. They are sensory and emotional load, aimed at the exact cohort least able to absorb it, many of whom left school partly because it was overwhelming.
So what: Restraint stops being a matter of taste here and starts being an accessibility argument. That is a much stronger footing than “Duolingo is overdone”.
Peer-reviewed work leans the other way. Cao, Wang, Li & Tong (2023) in the Journal of Children and Media report that characters' realism, not familiarity, improved children's learning from video. Richert, Shawber & Hoffman (2009) found children preferred transferring learning from real over fantasy sources.
So what: Combined with what we already knew (praise wording swings from -0.44 to +0.66; gamification's cognitive effect holds at g=0.49 while its motivational effect goes unstable), the mascot-and-streak package is the least evidenced part of the mold.
Both turned up in the same research. Neither kills the argument, but designing without them would be designing against the evidence.
It is tempting to read all of the above as “this audience hates technology”. The data says otherwise. An Age of Learning-commissioned survey reported by Fast Company found 44% of home educators use ChatGPT, against 34% of classroom teachers. Outschool's own data shows the post-2020 cohort skews more secular and more progressive than the pre-pandemic one. Euka sells itself as blended, screens plus practical activity, not as a screen-free alternative.
So what: the anti-screen segment is real and loud, particularly the Waldorf and Charlotte Mason end, but it is a minority. The move is restraint and credibility, not pretending to be a paper product. We are not designing for people who hate screens. We are designing for people who want the screen to be serious.
The best pedagogy-prevalence data found (Thomas 2017, n=1,055, via NHERI) breaks down as eclectic 68%, classical 14%, unschooling 13%, boxed or traditional 5%. And 79% used no single formal curriculum package at all.
So what: our product generates a complete term across all subjects and asks the parent to approve it. That is the boxed model, and boxed is 5%. I am not saying the model is wrong, generation makes it a different thing from a printed box, and the compliance burden pushes Australian families toward completeness in a way US families do not feel. But it is the single biggest assumption in the product and it deserves a proper look with Nathan rather than a line on a design page.
Each one is a coherent world, not a mood. Palettes are indicative, meant to be argued with rather than adopted.
Naturalist illustration, ink linework, cream paper. Colour codes subjects rather than decorating them.
Paper-cut, collage and real materials. Photographed things rather than drawn ones.
Literary. Each subject looks like its own book, not another tile in the same system.
Warm light against dark. The only direction that uses the name and the mark we already own.
None of this means tearing out progress tracking, XP or badges. The evidence splits cleanly: gamification's cognitive effect holds up under scrutiny at g = 0.49, while the motivational and behavioural effects go unstable. Structure earns its place. The arcade costume does not.
There is a working example of exactly this split. Nibble, a knowledge app, runs ordinary levels and XP bars underneath, but skins the whole thing in painterly still-life: a ceramic plate, a checked tablecloth, a bottle of syrup rendered like a food illustration. Same mechanics as everyone else, none of the arcade. That is the move available to us.
Screens referenced throughout: Duolingo ABC library, Apple Books Kids, Nibble, Calm Kids, Headspace Kids, Greg.
Three, and the first one is the most important thing on this page after the audience finding.
Across all ten aesthetic directions researched, every exemplar with real evidence behind it sits in the two-to-eight bracket. Nothing was found that is proven for ten to fourteen year olds, which is half our range. Even Pok Pok, the strongest restraint example, has a published review noting it starts to feel too simple past about eight.
So the honest problem is not really which aesthetic to pick. It is that nobody has solved the older-child end with a restrained look, in any of these styles. A twelve year old doing algebra is the hardest user in this product and the one with the least precedent to borrow from. Whatever direction you like, test it on the oldest child first, not the youngest. The young end is easy and everyone can do it.
A muted or low-chroma palette makes accessible contrast harder on interactive elements, and it uses up the saturated hues you need to encode state: correct, incorrect, locked, celebrating, needs attention. That is an engineering cost, not a taste debate, and it is a fair hit on my own recommendation below.
It is also the strongest argument for Direction D specifically over a beige or sage reading of restraint. A dark ground with warm light keeps enormous contrast range available and leaves the saturated end of the spectrum completely free for state. Restraint in the ground, not in the signals.
There is a real critique, most sharply put in the “sad beige parenting” commentary, that muted children's aesthetics mostly serve adult taste and quietly override what children themselves would choose. It is cultural commentary rather than a study, but it names the failure mode precisely: designing something parents will photograph rather than something a child wants to open on a Tuesday morning.
Worth holding against the evidence on the other side. Stern-Ellran and colleagues (2016) found uncontrolled visual saturation can hurt young children's task performance, so blanket brightness is not a safe default either. The target is neither beige nor arcade. It is controlled contrast with a clear focal point.
Every clean counter-example found is either an open-ended toy or a screen-free audio device. Pok Pok Playroom won a 2021 Apple Design Award with an eleven-colour palette and no gamification at all. Yoto and Tonies hold real design awards including Red Dot and iF. But none of them is a structured, progress-tracked, multi-subject curriculum that a child has to return to on a Tuesday morning.
So the honest position is symmetrical. There is no evidence the arcade look is required for our category, and no proof restraint carries daily engagement in it either. Going this way means being first, not following a playbook.
Worth weighing against the alternative, though: the playbook we are currently following was written for a free consumer language app that converts 6 to 8 percent of users and optimises for daily return. We are a paid curriculum a parent has already committed to. Different business, different job.
Direction D, The Lantern, and I would not treat it as a close call. Three reasons, in order of weight.
It is the only one that uses what we already own. The venture is called Lightkeepers and the mark is a lantern. A warm light held up in a dark place is already the brand's central image, and right now the design system does nothing with it. Every other direction asks us to import a world from outside. This one just takes the name seriously.
It fits the customer the data actually describes. A parent whose child was distressed at school, who is now anxious and improvising, is being offered a light to follow. That is the correct emotional register, and it is the one Euka is reaching for with reassurance copy while still looking like corporate software.
It is the kindest option for a neurodivergent child, and that is close to a third of them. A dark, low-contrast-noise ground with one warm focal light is close to the opposite of sensory overload. It also solves a real interface problem for free: when the brightest thing on screen is the next action, an inattentive child does not have to work out where to look. Direction D turns our accessibility constraint into the organising visual idea rather than something we retrofit later.
It is cheap to reach from here. Navy is already in the locked palette. Amber is already the accent. Direction D is closer to a re-weighting than a rebrand, which matters because changing buildchain/ is a conversation with Nathan, not a free move.
I would pair it with the Field Guide texture for lesson content: naturalist linework and functional subject colour on a light reading surface, sitting inside the darker lantern chrome. That gives the two-surface system the parent/child split needs anyway, and it solves the long-reading problem dark backgrounds create.
Take all four into the AI tools tonight regardless. I would rather be argued out of this by something you generate than have you design to my pick. The prompts stay brand-free precisely so that can happen.