Software engineer Vadim Drobinin analyzed two decades of his personal chat logs using large language models to create a digital archive that evaluates his historical social relationships and communication patterns.

Key Points

• Drobinin processed 1.2 million messages from platforms including ICQ, IRC, VK, Twitter, Facebook, Instagram, and Telegram.
• Data analysis revealed that increased question-asking often correlates with thinning friendships rather than deepening connections.
• Vocabulary overlap with certain contacts dropped from 69.5% to 8.7% over time, indicating a significant shift in shared language.
• The project identified a personal tendency to offer unsolicited advice rather than active listening during conversations.
• Despite his active network shrinking from 275 to 60 contacts since 2016, his total annual conversation volume remained consistent.

Why it Matters

This experiment demonstrates how individuals can leverage personal data access laws and generative AI to gain objective insights into their own behavioral patterns and social history. It highlights the potential for personal CRM tools to challenge subjective memories and provide a data-driven perspective on long-term interpersonal dynamics.