{"id":90706,"title":"Chiron - Grammarly for Math ✏️","tagline":"The Booth Brothers are building an iPad app that understands math as it's written","body":"Hey there 👋\n\nWe’re [James](https://x.com/J4mesBooth) and [Max](https://x.com/MaxB00th) - brothers and cofounders of [Chiron](https://projectchiron.org).\n\n# 📌 **TL;DR**\n\nChiron is an [iPad app](https://apps.apple.com/us/app/project-chiron/id6502346822) that understands math as it’s written - like Grammarly for algebra. We’re both Oxford math grads, have tutored for a decade, and we’re creating the future of learning which will be entirely 1-on-1.\n\n[Launch Video](https://youtu.be/Ec8Nwe0EaSU)\n\n# ❌ **Problem**\n\nWith AI that’s so great at math, why can’t it teach at near-human quality?\n\n* Math is handwritten ✏️\n* Math has to be practiced not read 🧮\n* AI is rubbish at spotting _your_ mistakes 🔍\n\n# ✅ **Solution**\n\n* Freeform writing on a whiteboard app ✍️\n* Accuracy of a calculator with symbolic logic 🔢\n* Real-time feedback on every stroke (\u0026lt; 300ms) ⚡\n\n# ⚙️ **How we do it?**\n\nWith tens of thousands of lines of Python and Wolfram Language code. Writing on an iPad, we convert the stroke data to LaTeX, parse it, ask a million symbolic questions about it with reference to the problem’s _ground truth_ (the lines with a blue index), and assess whether it is right, wrong or something in between.\n\n![uploaded image](/media/?type=post\u0026id=90706\u0026key=user_uploads/1522067/b06893f5-bab0-4021-b09e-c8d38b33e4ee)\n\nWe do **not** rely on the vibes of a language model to assess the details of their working - too slow and unreliable. Many students give up on LLMs for genuine collaboration when they mislead them on topics the student finds difficult.\n\n**We can handle a lot of math already.**\n\n![uploaded image](/media/?type=post\u0026id=90706\u0026key=user_uploads/1522067/df464a12-f051-4c0a-8c03-90e660079c40)\n\n# **💡 Core insight**\n\nWe tried all the AI tutor desktop formulations we could imagine, but beyond the obvious interface challenge, AI is exceptionally poor at spotting the subtle mistakes students often make. LLMs are great at doing correct math, but much worse at pulling out an error within the many lines of messy student reasoning.\n\nWe started performing a symbolic analysis of the student’s working as context for the AI to boost accuracy - **but then we realised we could give this straight to the student.** This was when we started to feel the magic.\n\n# **🙏 Our Ask**\n\nSo far, _every_ single teacher we’ve demoed our iPad app to in person has immediately started work greenlighting a pilot. Let us know if you can introduce us to any math teachers teaching between Algebra I and Calculus I.\n\nAnd you can check out Chiron on the [App Store](https://apps.apple.com/us/app/project-chiron/id6502346822) now - or share with any students who might benefit from it!\n\n![uploaded image](/media/?type=post\u0026id=90706\u0026key=user_uploads/1522067/7b18b9c0-1e00-49fa-a8bb-d6ab7bb7d080)\n\nCheers, James \u0026 Max","slug":"Nb0-chiron-grammarly-for-math","created_at":"2025-05-24T20:19:56.044Z","updated_at":"2026-05-25T02:48:10.064Z","total_vote_count":105,"url":"https://www.ycombinator.com/launches/Nb0-chiron-grammarly-for-math","share_image_url":"https://www.ycombinator.com/media/?type=post\u0026id=90706\u0026key=user_uploads/1522067/7b18b9c0-1e00-49fa-a8bb-d6ab7bb7d080","company":{"id":30459,"name":"Chiron","slug":"chiron","url":"https://projectchiron.org","logo":"https://bookface-images.s3.amazonaws.com/small_logos/413bccd180b2d3b183e43292dcbda31f7c7beaf3.png","batch":"Spring 2025","industry":"Education","tags":["AI-Enhanced Learning","Developer Tools","Education","Edtech"],"search_path":"https://bookface.ycombinator.com/company/30459"}}