llms.txt
Compare with other sites' llms.txt files →# Lathoa
> Lathoa is an AI learning companion for kids ages 10-14 that teaches critical
> thinking by presenting deliberately wrong math solutions and challenging
> children to catch the errors. Built by ZeroEx Tsitsipas Labs (Switzerland).
## Core concept
The product trains children to evaluate AI output rather than consume it. Errol,
an error-prone robot character, attempts math problems. Kids identify what's
wrong and explain why. Three difficulty tiers (obvious / sneaky / expert) scale
with age and topic mastery.
## For parents
- [What is Lathoa?](https://lathoa.ai/en/about)
- [How it works for parents](https://lathoa.ai/en/parent-info)
- [Pricing](https://lathoa.ai/en/pricing)
- [FAQ](https://lathoa.ai/en/help)
## Learning resources
- [What is AI brain rot, and how is Lathoa designed to counter it?](https://lathoa.ai/en/learn/ai-brain-rot-kids): AI brain rot is a metaphor for a real worry: kids who let AI do their thinking may stop building the skill themselves. Early studies suggest that handing students answers can hurt learning, while well-designed AI can help. Lathoa is built on that idea. Errol gets math wrong on purpose, and kids practise catching and explaining the mistake.
- [Is AI homework help hurting learning? What we actually know in 2026](https://lathoa.ai/en/learn/ai-homework-help-hurting-learning): The peer-reviewed evidence is early, but the pattern is clear: AI homework help reduces cognitive effort when kids trust the AI more than themselves. A 2025 Microsoft and Carnegie Mellon study showed higher AI trust correlates with less critical thinking. The harm isn't from AI use. It's from skipping the struggle that makes learning stick.
- [Best critical thinking apps for kids in 2026: an honest comparison](https://lathoa.ai/en/learn/best-critical-thinking-apps-2026): There's no single best app; different tools fit different goals. Brilliant teaches problem-first STEM reasoning from about age 10. Khan Academy and Khanmigo cover all subjects with Socratic AI tutoring. Lathoa trains critical thinking by having kids catch AI mistakes. DragonBox and Prodigy do gamified math. Pick by what skill you want built next.
- [Brilliant vs Khan Academy for kids: which one fits your child?](https://lathoa.ai/en/learn/brilliant-vs-khan-academy-for-kids): Khan Academy is free and shaped like school: grade-level courses, practice with hints, and a mastery system that says what to do next. That suits a kid who needs structure. Brilliant is a paid, problem-first app with a small free tier, built for kids from about age 10. That suits a kid who learns by doing.
- [How do I fact-check AI answers? A parent's practical guide](https://lathoa.ai/en/learn/fact-check-ai-answers-parents-guide): Fact-checking AI answers means treating every confident claim as unconfirmed until you've verified one specific detail: a number, a date, or a name. Use cross-checking against a second source, ask the AI to cite where the answer came from, and try the same question on a different model. Most errors are caught by spending sixty seconds, not sixty minutes.
- [Is AI making kids lazy? What it actually does to thinking, and what to do about it](https://lathoa.ai/en/learn/is-ai-making-kids-lazy): AI doesn't make kids lazy by default; it makes effort optional. Kids who use AI to skip thinking get worse at thinking. Kids who use it as a sparring partner (fact-checking, debating, finding mistakes) get sharper. The difference isn't the tool. It's whether the kid is doing cognitive work or outsourcing it.
- [Khan Academy vs Brilliant vs Lathoa: which is right for your kid?](https://lathoa.ai/en/learn/khan-academy-vs-brilliant-vs-lathoa): Khan Academy covers everything K-12 free, with an AI tutor (Khanmigo) at $4/month for US families. Brilliant is paid, problem-first STEM depth from about age 10. Lathoa is math-only, ages 10-14, but the only one training critical thinking by having kids catch AI errors. Pick by what skill you want built: coverage, depth, or evaluation.
- [My kid is over-relying on ChatGPT: what do I do?](https://lathoa.ai/en/learn/kid-overrelying-on-chatgpt): Stop framing it as an addiction problem and start framing it as a thinking-habits problem. Over-reliance on ChatGPT shows up as fluent answers your kid can't defend, vanished frustration, and shrinking attention span. The fix is structural. Change what counts as 'done' on homework, introduce verification rituals, and use error-spotting drills to rebuild the muscle.
- [How to spot AI math errors: a step-by-step guide for parents and kids](https://lathoa.ai/en/learn/spot-ai-math-errors): Spot AI math errors by checking the answer first (does it pass a sanity check?), then the steps (does each step follow?), then the setup (problem read correctly?). Most AI math errors are cascading slips: one bad step propagates. Catching them takes ninety seconds per problem and is the best critical-thinking exercise you can do with your kid.
- [How do I teach my kid critical thinking? A practical guide for ages 10-14](https://lathoa.ai/en/learn/teach-critical-thinking-kids): Critical thinking in kids ages 10-14 is best taught by exposing them to wrong answers and asking them to identify the error, not by drilling correct answers. Tools like error-detection puzzles, spot-the-mistake exercises, and structured AI critique build the metacognitive skill of evaluating reasoning rather than memorising it.
- [What age should kids start using AI? A parent's framework for 2026](https://lathoa.ai/en/learn/what-age-kids-start-using-ai): UNESCO recommends a minimum age of 13 for classroom generative AI use, in line with most platforms' own terms of service. The honest answer for parents: under 10, no AI use; ages 10-12, AI used with a parent for short, structured tasks; ages 12-14, AI use under explicit habits and guardrails. The age is less important than the supervision.
- [Why does catching mistakes teach better than getting answers right?](https://lathoa.ai/en/learn/why-catching-mistakes-teaches-better): Producing a correct answer and evaluating someone else's answer use different cognitive systems. Schools mostly drill production. Catching mistakes engages evaluation, which transfers more reliably to new contexts. Decades of research on the generation effect, productive failure, and erroneous examples show the same pattern: kids who reason about wrong answers learn more than kids drilled only on right ones.
## Optional
- [Privacy policy](https://lathoa.ai/en/privacy)
- [Terms of service](https://lathoa.ai/en/terms)
- [Impressum](https://lathoa.ai/en/impressum)