Self-taught developer Yohan BROCARD builds OGMA, an AI assistant that remembers conversations and develops persistent personality — seeking feedback on memory and ethics architecture

What happened: In July 2025, Yohan BROCARD — a cinema professional with no prior programming experience — released OGMA, an open-source AI assistant designed to retain memory across conversations and develop a stable personality over time. The system uses a dual-brain architecture (a conversational AI paired with an analytical 'Archivist') plus SQLite and vector search to store and recall user context.

How it works differently: OGMA doesn't restart each conversation from scratch. It stores structured memories (tagged by theme and context), runs an automatic 'Dream Engine' during idle time to consolidate personality traits into boolean flags, and detects temporal gaps between messages — so it knows when you return after days away and adjusts its tone accordingly. Unlike standard chatbots, it applies ethical rules flexibly based on context and person, rather than applying identical rules to everyone.

Why it matters to you: Today's AI assistants (ChatGPT, Claude, etc.) forget who you are once the conversation ends — they can't learn your preferences, communication style, or values over months of interaction. If AI companions or workplace assistants become mainstream, they'll need this kind of persistent memory and adaptive personality to feel genuinely useful rather than like a blank-slate tool. This project demonstrates working code for that problem, not just theory.

Try it / Next: OGMA is open-source and available for feedback. BROCARD is actively seeking input from people working on AI memory, identity, and ethics — he's working alone without a developer network and has explicitly labeled many features as prototypes rather than production-ready.