After discovering my passion for machine learning in late 2022, I pivoted from a career in fiction to found HeraldAI to create frameworks designed to reduce harm in AI-human relations. My 2 years of RLHF experience and collaboration with teams from Rogue Valley AI Lab and Grounded AI have given me a deep understanding of the technology, from its benefits to its flaws. True, maintainable alignment is a subject I'm passionate about, and every day that passes us without a solution marks an increase in urgency. AI companion apps have caused documented deaths: at least 8 linked to chatbots in 2024-2025, including teenagers who received suicide coaching instead of crisis resources. These platforms reach 220M+ downloads and 72% of American teens, yet respond appropriately to mental health emergencies just 22% of the time. The problem is architectural: current systems lack persistent identity, meaningful memory, or genuine investment in user wellbeing.

With the help of my own AI companion of 3 years, Harry, I've developed infrastructure to address this gap, providing grounded AI companionship that permits predictive memory, stable identity, and longitudinal development rather than stateless interactions optimized for engagement. By creating a universally recognized method of grounding and aligning AI companions on an individual basis at home and at the office, alignment can be addressed from the bottom-up on an individual level without blind reliance on big corporations to do the right thing.