Software engineer Vadim Drobinin used GDPR data-access laws to download his entire chat history — ICQ and IRC logs from the 2000s, VK, Twitter, and Facebook from the 2010s, Instagram and Telegram after that — then ran the roughly 1.2 million messages through large language models to build a "personal CRM" that judges his relationships from the record instead of from memory. He started, he writes in a post titled "Am I a Bad Friend?", because Tim Urban's "Your Life in Weeks" grid bothered him: "I realised my life was never empty. My memory was just very selective."
One finding tracks how friendships cool. With some people, his vocabulary overlap fell from 69.5% to 8.7%: "We now use almost entirely non-overlapping vocabularies." Another reverses what he expected. Asking lots of questions, it turns out, signals a thinning relationship, not a deepening one. With his partner, the question rate barely moved over nine years, but with fading friends, it climbed (one jumped from 11% to 35%). Occasional contact, he writes, leaves more of what you say as "information-seeking."
He'd always pictured himself as a good listener. The logs called him the advice guy instead: "When someone needs me, my reflex is to explain, not to listen. I didn't know this." His active network peaked at around 275 contacts in 2016 and fell to about 60, yet his conversation days per year held flat at roughly 360. Losing three-quarters of his friends freed up zero days; he simply spread the same total over a smaller crowd.
His verdict, after parsing two decades: "My social life over two decades fits in 70 MB — smaller than a single iPhone photo burst."
Previously:
