Every other model starts with knowledge — here’s the lesson, memorise it, then we’ll test if you can use it. ThinkSmith starts before knowledge. Your child tackles a real-life challenge first, then learns what they need to solve it, with a 24/7 AI coach beside them. Then they teach it — by reviewing three peers’ solutions and coaching one of them through the gap. That’s when it sticks, and that’s what their Thinking Outcome shows you.
Most learners arrive unprepared and can’t absorb brand-new knowledge just by listening to it once. Re-teaching it in a video, slides, or a worksheet doesn’t fix that — it’s the same information in another format, still with no foothold to land on. ThinkSmith changes the order, not just the format.
Most learners cannot memorise and interpret a lecture full of brand-new knowledge. They get little from it, then struggle on the assignment and the exam. The knowledge had nowhere to land.
Same lesson, different format — useful material, still presented before the learner needs it. ThinkSmith links to Khan Academy tutorials too, but in the middle, not at the start: to solve a challenge your child has already attempted. It’s a link, not a partnership — Khan Academy is a free resource anyone can link to.
A team is only as strong as its weakest link. If most learners arrive unprepared, most cannot contribute. In ThinkSmith there is no weakest link and no free ride — the 3 peers are AI, always prepared, and your child must judge their reasoning themselves.
The order is the insight — and teaching others is where it locks in. Researchers have studied both for decades; ThinkSmith is the first platform built entirely around them, at the scale of one child working from home. The 17.31% exam improvement wasn't produced by better teachers or harder content. For the ten courses that could be compared directly, the instructors and courses were the same as in their previous, lecture-centred offerings — the students were new cohorts, and the only variable was this method.
Before any lesson, your child writes their own solution to a real-life challenge, in their own words — not multiple choice, not fill-in-the-blank. They are not expected to get it right. The attempt creates the foothold — and the questions — that the new knowledge will answer.
No lesson first — this is the step other models leave to chanceNow the new knowledge arrives — often as a link to a free Khan Academy tutorial — at the moment your child needs it to improve their solution, so it lands on a question they're already asking, not on a blank slate. The pilot study called this exactly what it is: foundational knowledge acquired “on a need-to-know basis.” Ms. Rivera, their 24/7 personal AI coach, is beside them throughout: guiding their thinking, never supplying the answer.
Your child takes the teacher's role, reading three AI peers' attempts at the same challenge, then selecting one and writing constructive feedback to that peer, coaching them through the gap. There is no weakest link and no free ride — the peers are always prepared, at genuinely varying quality, and judging their reasoning forces your child to rethink their own.
This is the step most platforms skip entirelyThey receive feedback on their own work and write a reflection on what changed in their thinking — the closest thing to a one-on-one teacher conference at this scale.
A synthesis across five competencies and a narrative of how their thinking developed that activity — and a clear read on whether they're ready to move on. They keep going if they want — the next challenge is theirs to choose.
ThinkSmith fits into how homeschooling actually works — not around it.
If a lesson-first approach leaves your child behind — if they can't absorb new knowledge by listening, freeze on a quiz, but light up when asked to reason through a real problem first — their kind of thinking has a place to develop and be seen here. ThinkSmith starts where they are strong: with the challenge.
One dashboard for every child. Email login or a simple 4-digit PIN for younger kids — no juggling separate accounts or logins per subject.
Older students register and run their own dashboard, no parent account required, while you still see the weekly rollup of how their thinking is developing.
One shared way to see thinking growth across a group of kids working on entirely different courses — and because every child works through their own challenge and teaches their own AI peers, no child's learning depends on how prepared the rest of the group arrived.
This is not a preference — it's how learning works. As the pilot study behind ThinkSmith puts it, following Duffy & Cunningham: learning is “an active process of constructing knowledge rather than acquiring knowledge.” Knowledge handed over first has nothing to be constructed with. The ThinkSmith order fixes that, four ways:
Before any lesson, your child commits to their own idea — even a wrong one. New knowledge then has something real to connect to, build on, or correct. Without that foothold, knowledge is simply heard and forgotten: there is nothing in the learner's mind for it to attach to.
Students retain less than 20% of a lecture a week later — Guskin, cited in the studyTrying first shows learners what they don't know — to themselves. So when the knowledge arrives, it answers a question they're already asking, and they use it immediately in their own solution. Used knowledge is understood and remembered; knowledge memorised in isolation, for a test that hasn't happened yet, mostly isn't.
Foundational knowledge, acquired “on a need-to-know basis” — the pilot studyA coach, a peer, or a teacher cannot guide thinking that hasn't happened yet — all they can do is repeat the lesson. An attempt, and three peers' attempts, give feedback something specific to work on: the actual misunderstanding, revealed. In the pilot, class time stopped being a lecture and became that feedback — because the engagement had already happened.
100% of surveyed instructors: engage first, class time for feedback, was more productiveReviewing three peers' solutions — and coaching one through a gap — forces your child to compare, analyse, and evaluate — higher-order thinking from the very start, not as a reward after memorising the fundamentals. The pilot study notes this reversal of the traditional fixed order also helps learners keep the foundational knowledge, because they acquired it while using it. And learners take ownership: they build their own learning strategies instead of waiting to be taught.
92.9% of surveyed students took more responsibility for their own learningPut together, that's the result the pilot measured: not better content, not better teachers — a better order, and final exam averages 17.31% higher across the ten courses that could be compared directly.
Plenty of tools claim to cover "any curriculum." Most quietly generate it — and get it wrong. ThinkSmith does the opposite: we find the official curriculum document for your child's grade, confirm it's the current version, and build real-life challenges from it — region by region. The curriculum decides what each challenge must cover; it is never turned into a lesson script to memorise. We'd rather map one region properly than claim fifty. So far, that means Grades 5–12, built for Texas and for Ontario, Canada:
About those links, plainly: Khan Academy is free, and its tutorials are there for anyone to link to. Where a tutorial can help with a challenge, ThinkSmith links to it — that is the whole relationship. Khan Academy does not endorse, sponsor, or support ThinkSmith, and ThinkSmith does not endorse, sponsor, or support Khan Academy.
Prefer your own materials? Upload your child's syllabus or course outline and ThinkSmith builds challenges matched to it directly — no waiting on region coverage.
Every message a child writes is screened for signs of distress before our AI coach is allowed to reply. If a report to a child-protection authority is ever appropriate, we disclose only a learner's name, grade, and region, and their parent's name and email — never a date of birth, address, or phone number. Learner records are held in Canada. We don't sell or share family information or show advertising.
Source: pilot study, Ubon Ratchathani University — 1,742 students, 22 courses, across seven fields: Biology, Chemistry, Computer Science, Information Technology, Mathematics, Physics and Engineering. In the pilot, students engaged with new knowledge before the lecture — tackling a challenge, reviewing peer ideas, and receiving structured feedback ahead of the deep-dive discussion, instead of sitting through passive instruction first. The ten comparison courses were taught by the same instructors who had previously offered them lecture-centred, so the variable being tested was the method — not the instructor, and not the course. One comparison course, Food Microbiology, showed no significant difference; it's reported in the paper, because the method is the point, not the marketing.
Read the full research paper →The method behind ThinkSmith has been tested for years — 1,742 students in the Ubon Ratchathani pilot study. This platform is new, and we're onboarding a small number of founding families directly with our team to get it right from day one.
Join our founding family trial →No mail app needed — the form opens right here.
ThinkSmith runs on two layers of protection: the foundation provided by the AI model itself, and the layer we build on top for our own classroom.
Tier 1 — Foundation. Our AI coach is built on Claude, made by Anthropic. Anthropic's own safety systems apply to everything the coach produces, preventing harmful, illegal, or exploitative content at the source. This protection is always on and does not depend on us.
Tier 2 — ThinkSmith. The foundation layer cannot know what a good lesson looks like, or what Ms. Rivera is supposed to be. That is our responsibility, and it is where we concentrate our work.
We are continuing to add to this layer. It is the part of the system we work on most.
To allow an authority to assess immediate risk and contact a parent, we share only:
We do not include a home address, date of birth, telephone number, location data or billing details in a safety report.
If a child discloses a concern involving someone in their own household, we report to the authority without contacting the family first. A system that routed every disclosure straight back to a parent account could place a child at greater risk.
Read our full Digital Safety Plan →
We're onboarding a small number of founding families directly. Send us a note and someone from our team will reply, usually within a working day.
When parents look for a learning approach for their children, they usually ask two questions: will this actually help my child learn better, and will it build the self-reliance they need for the future?
In 2008–2009, ThinkSmith founder Leslie Richards co-led a university pilot study involving 1,742 students across 22 courses in STEM, computer science and engineering at Ubon Ratchathani University.
The goal was simple: engage students in new knowledge before the lecture. Instead of sitting through passive hours of instruction, students used the ThinkSmith framework — active problem-solving, reviewing peer ideas, and receiving structured feedback ahead of the deep-dive discussion.
Academic improvement only happens when a learner’s mindset changes first. In a post-study survey of 684 students:
Of the 13 instructors (of 22) who completed the survey, 100% said having students engage with new information on their own and collaboratively before class, then using class time for feedback instead of a lecture, was more productive and a better use of everyone's time. The study opens with the problem they all named: “students don’t come to class prepared” — and notes that most students do poorly under lecture-centred instruction because they struggle with rote learning.
Traditional instruction asks students to sit quietly while an instructor talks, leaving the hardest part — applying the knowledge — for homework, when the learner is isolated and frustrated. The pilot inverted that with a five-stage loop:
Whether in a university lecture hall or at a homeschool kitchen table, the core finding holds: children learn far more from what they reason through themselves than from what they are told. ThinkSmith gives your child the structure, real-time coaching and feedback loop to become an independent, confident learner — supported by evidence from thousands of students.