Meera Pillai had been covering the Indian tech industry for nine years, first at Economic Times and then at her own newsletter, which had 180,000 subscribers and a reputation for the kind of technology criticism that was inconvenient for founders but useful for everyone else.She had been tracking Digital Pind since the Series A. Not because she thought it was doing anything specifically wrong — the consent architecture was better than most Indian consumer tech she had reviewed, the data handling disclosed, the product doing roughly what it said it would do. She was tracking it because she had a practised instinct for the kind of technology that moved faster than the questions asked about it, and Digital Pind was moving very fast.Her first article had been in January 2024, before launch: "Who Owns the Dead? Digital Afterlife Companies and India's Absent Regulatory Framework." Karan Mehta had declined to comment. The Ministry of Electronics had not responded. The piece got 80,000 reads.Her second piece, in March, was more technical: "How SMRITI Works, and What It Can't Know." She had spoken to three AI researchers who explained the architecture, confirmed the persona vector approach, and agreed that a fine-tuned LLM could not access information absent from its training data. This was technically settled. The piece was reassuring.She had already started the third piece in her head by April.The emails had started arriving in April. Not many at first — three, four a week. By May, twenty a week. Users who had noticed something in their SMRITI conversations that they could not explain. The specific shape of the inexplicable was consistent across correspondents: