Nobody in the family uses the family AI platform I build - really bummed about it

A developer's ambitious project to create a completely private, local alternative to Alexa has hit a painful reality check after discovering his family never uses the sophisticated system he built. Inspired by privacy concerns after Amazon announced using Alexa recordings for training, he spent over a year developing a platform with voice control, smart home automation, documentation search via Bookstack, and internet capabilities. The system runs on substantial hardware including multiple high-end NVIDIA GPUs (3090s, 4090s, 4080, 5060Ti) worth thousands of dollars, an EPYC processor with 24/48 cores, and 256GB of RAM. Despite achieving technical success with features like memory functions and RAG (retrieval-augmented generation), user adoption was zero—his wife logged in once since Christmas, his son once in 17 days, and his daughter never.

The experience highlights a common disconnect in tech development between engineering excellence and user needs. The developer's survey revealed his family simply didn't care about data localization enough to change their habits, despite his privacy motivations. This realization has led him to dramatically scale back his ambitions, considering replacing the $20K+ hardware setup with a single 48GB Mac Mini running GPT-OSS:20B. The story serves as a cautionary tale about building solutions without continuous user validation, even for personal projects. For the broader AI community, it underscores that technical capability alone doesn't guarantee adoption, and that privacy concerns—while significant to developers—may not motivate average users to change established behaviors with commercial assistants like Alexa.