Hacker News new | ask | show | jobs
Show HN: learnchess.ai – Local-first chess coaching app with AI (learnchess.ai)
2 points by nikolasburk 100 days ago
I've been playing chess for a few years with my ELO stuck around 1600 on Lichess. I wanted an app that combined puzzle training, interactive lessons, game analysis, and AI coaching in one place. But all the apps I used over the years either felt overwhelming, had UX that wasn't up to my expectations for a modern app or were lacking some fundamental features that I really wanted. So, I built LearnChess.ai.

The technical angle I'm most excited about: it's built local-first using LiveStore (event-sourced, SQLite on-device), which means zero loading delays for user data and full offline support. Sync to the cloud happens in the background for cross-device usage. The stack is TanStack Start (web), Expo (mobile), and it runs entirely on Cloudflare.

Features in the current beta:

- Puzzle trainer with community submissions

- Guided lessons for openings and endgames

- Game import and analysis from Lichess & Chess.com

- Opening explorer based on your own game history

- Stockfish running on a dedicated Hetzner server (faster than in-browser WASM)

- AI coaching integrated directly in the analysis board

- Coordinate training, beginner track, leaderboards, ...

Available on web, iOS (TestFlight), and Android (nearly ready).

Would love your feedback — both on the product and the architecture choices!

1 comments

Nice job~~

As a someone who recently built a chess app with Stockfish (WASM with a fallback to the server, see my profile if interested), just be aware that if your application becomes even moderately successful, a couple dozen users hitting your Stockfish server concurrently is very likely to be slower than running individual Stockfish WASM on the client side unless you significantly beef up your backend.