I'm working on Learntime (https://learntime.ai), an open-source/saas learning app that combines AI mentorship with evidence-based study techniques. It's designed for students and lifelong learners who want to retain information more effectively and perform better in exams. It focuses on un-prompted recall of entire topics (not just flashcards), supported by an AI mentor that provides hints and tracks progress. It uses spaced repetition to optimise review schedules and includes a quick quiz feature targeting weak areas. This is my first solo open-source project, and it's been quite a journey! I initially started with LangChain, thinking it would simplify the AI integration. However, I found myself spending more time wrestling with abstractions than solving core problems. Eventually, I dropped it in favour of a more direct approach. The biggest technical challenge has been implementing free recall scoring with LLMs. I've been through a cycle of different techniques - prompt engineering, few-shot learning, fine-tuning - and back again! All of this while welcoming my third child and approaching 50! It's been a balancing act, but the project keeps me energised. Feedback, similar experiences or insights welcome!