Lathoa App Introduces Deliberately Wrong AI to Teach Kids Critical Thinking

What Lathoa Does

Lathoa — the “critical‑thinking app for kids 10‑14” — presents math problems that are solved step‑by‑step by an AI‑driven robot Er .

Deliberate error: One step in Errol’s solution is intentionally wrong. The child must identify the faulty step, explain why it is wrong, and then earn a rank‑up.
Skill focus: The app rewards detecting a mistake rather than arriving at the correct answer, training a habit that will be valuable whenever a child interacts with AI.
Age range: Designed for 10‑ to 14‑year‑olds; difficulty adapts from single‑step checks for younger users to multi‑step word problems for older ones.

How the Learning Loop Works

Errol solves aloud – The AI walks through a problem, showing each intermediate step.
One step is wrong – The mistake is hidden among the steps (tiers: Obvious, Sneaky, Expert).
Child spots the error – The student highlights the incorrect step; Errol’s confidence icon visibly drops.
Explain in own words – The child types a brief explanation of the error, reinforcing understanding.
Reveal the correct solution – The app displays the proper solution side‑by‑side and awards a rank (Rookie → Inspector → Detective → Sherlock).

The loop repeats without timers or leaderboards, emphasizing thoughtful analysis over speed.

Pedagogical Rationale

Error‑detection learning – By confronting a plausible but wrong answer, children practice pattern‑matching for faulty reasoning, a skill not covered by traditional tutoring platforms that only reward the right answer.
Active explanation – Requiring a written justification forces students to articulate the underlying concept, which research shows improves retention.
Progressive difficulty – Three error tiers scale with age and skill, ensuring the challenge remains appropriate as the child improves.

Product Offering & Pricing

Plan	Price	Kids	Features
Free	$0	1	Obvious tier only, 3 cases/day, no replay, no parent dashboard
Learner	$/year (exact figure not disclosed)	1	All three tiers, unlimited cases, replay past catches
Parent (most popular)	$/year	1	Learner features + parent dashboard, weekly progress digest, misconception flags
Family	$/year	Up to 4	Parent features for each child, side‑by‑side sibling progress

A 7‑day trial of the Parent plan is available without a credit card; the first 50 families receive a half‑price yearly discount (35 spots remaining at the time of posting).

Safety & Privacy Measures

Verified content – Each AI‑generated problem is passed through a separate verification step; if the AI attempts an outright false statement (e.g., “2 + 2 = 5”), it is rejected before reaching the child.
No open‑ended chat – Interaction is limited to structured cases, eliminating the risk of unsupervised dialogue with a language model.
COPPA compliance – The service does not profile children, sells data, or employ dark‑pattern monetisation.
Device‑agnostic – Runs in any browser, on iOS and Android; multiple profiles can share a single device via a 4‑digit PIN.

Community Feedback from Hacker News

Positive reception – Commenters note the similarity to existing “error‑injection” exercises used in Khan Academy and classroom settings, confirming the pedagogical relevance.
Desire for more samples – Some users (e.g., lemming) request additional example problems to assess difficulty, especially for pre‑algebra levels.
UX observations – Minor UI quirks were mentioned, such as the robot mascot’s shadow moving with its bounce, but overall design was praised as well‑put‑together.
Skepticism about AI‑generated copy – One comment (demibabs) suggested the site’s copy appears AI‑generated, which could undermine trust despite the app’s goals.
Broader implications – Several commenters highlighted the societal value of teaching kids to question AI outputs, noting that deliberate‑error training could inoculate future users against misinformation and propaganda.

Comparison to Traditional Tutors

Traditional tutoring platforms (Khan Academy, private tutors) focus on guiding students to the correct answer. Lathoa flips this paradigm: the correct answer is provided only after the learner has identified and explained the AI’s mistake. This complementary approach targets a distinct competency—recognising when an authoritative‑seeming source is wrong.

Future Roadmap

The current curriculum covers arithmetic, fractions, ratios, pre‑algebra, and early geometry aligned to Common Core grades 4‑8. The team plans to add reading comprehension, basic logic, and later science modules, expanding the critical‑thinking framework beyond mathematics.

Bottom Line

Lathoa introduces a novel, AI‑driven learning experience that deliberately injects errors to teach children aged 10‑14 how to spot faulty reasoning. By rewarding the question rather than the answer, the app aims to cultivate a lifelong habit of critical evaluation—an essential skill in an era where AI‑generated content is ubiquitous.