01 — Maths, explained
Let's Math
Any primary-school maths question, turned into a correct, narrated, animated video.
An LLM is fluent. It is not a verifier.
Let's Math turns any primary-school maths question into a correct, narrated, animated video, typed, photographed as homework, or uploaded as a PDF. An LLM extracts the structure of the question, but deterministic code exists specifically to verify the arithmetic and override the model when it is wrong, which is the product's own admission that the model can't be trusted with correctness alone.
The model reads. Code decides.
An LLM extracts the structure of the question; deterministic code verifies the arithmetic and overrides the model when it disagrees, or declines the cases it can't verify safely, with no silent fallbacks. The teaching visual is chosen by code, never by the model: one select.py per topic, across 16 topics. Narration is assembled into timed beats, Deepgram Aura speaks them, and Remotion renders a portrait video. The same question always produces the same video, because the cache key is a hash of the question.
It declines rather than guessing.
If there is no dedicated scene for a pattern, the product declines rather than showing a placeholder. The LLM is allowed to understand the question; code owns correctness, the choice of pattern, and the choice of visual. It has been checked against 9 curricula, across 16 topics and 190 sub-concepts, for classes 1 to 5 with KG and class 6 as safety bands.
- Topics
- 16
- Sub-concepts
- 190
- Curricula checked against
- 9
Built with
- Python
- FastAPI
- AWS Lambda
- Next.js 15.5
- React 18
- Tailwind 3.4
- Remotion 4
- Deepgram Aura