I'm having a lot of fun using and writing my new language bot.
First time in ages that I find it fun to study languages, and I really think the bot is good at remembering my preferences.
It's the first time I've ever implemented - or really though about - how to manage context. As well as the obvious stuff - like compaction - my bot also has some memory-like systems that are for language-learning specifically. For example, it keeps track of mistakes and successes in grammar topics; all labelled and categorized by the llm.
This allows it to keep a compact summary of the mistakes I'm making in vocab and grammar and so on.
And also a 'learner profile' where the llm keeps a summary of my preferences and basic life facts. For example it has learned that "Bitcoin" is one of my interests
I'm not a language teacher, so I'm making sure that I design the system such that I don't have to make a decision that might be a bad decision; everything is llm-guided
It's just for me and one friend, not planning to share
View quoted note →
View quoted note →

