I pick up new words everywhere: browsing, reading books, watching something. I wanted every one of them to go straight into a learning pipeline and be right there on my phone, not lost in a translator tab.
It’s also my own problem solved with my own code and architecture. I’m the user, so I literally watch my own data flow through the system, feel where it hurts, and fix it.
There’s a browser extension for Chrome, Edge and Firefox, and a mobile app that drills words like Anki, but with typing and saying them out loud. Translations come from DeepL, examples and pictures are generated, audio is self-hosted TTS.
The backend is a .NET modular monolith with Kafka between modules, MongoDB, Redis and MinIO, deployed to my own kubeadm cluster on Hetzner via ArgoCD.
It’s a private project for now: the app is in a closed beta, the code is not public.