Conversation as it happens
People correct themselves, interrupt, change direction, and leave thoughts unfinished. The system has to adapt to them.
I wanted a screen-free device that could support everyday questions, conversations, and routines without requiring my family to speak in exactly the right way. So I built Hey Nubble and put it to work in our home.
Hey Nubble was designed around the way a household actually communicates: incomplete questions, interruptions, changing intentions, repeated routines, and people who do not always use the words a system expects. It became particularly helpful for family members who are neurodivergent, where predictable routines, patient interaction, and low-friction access can make a meaningful difference.
Living with a working device changed the question. It was no longer simply whether natural voice AI could work, but whether it could respond quickly, handle ambiguity safely, and remain dependable across the messy variety of everyday life.
People correct themselves, interrupt, change direction, and leave thoughts unfinished. The system has to adapt to them.
Voice can provide low-friction support for recurring household needs without adding another app, display, or interface.
A delayed response does more than slow the system down. In a voice interaction, it changes how natural and dependable the experience feels.
A device used around a family must behave predictably, communicate its limits, and keep responsible boundaries at the center of the design.
One broad model decision worked initially. Breaking the problem into measurable stages made the architecture easier to understand and improve.
The successor design separated product behavior from model, speech, and infrastructure providers so the experience was not defined by one vendor.
Hey Nubble began with empathy for a small group of real users: my family. Building it required voice interfaces, applied AI, connected devices, safety design, and product architecture. Living with it reinforced a broader leadership principle: start with the human need, observe what actually happens, and be willing to change the architecture when the evidence challenges the original design.
The strongest product decisions came from understanding how the device fit into family life, not from beginning with a model or feature list.
Real use exposed assumptions, edge cases, and architectural tradeoffs that controlled demonstrations could not.
AI accelerated exploration and implementation, while architectural review, product-level testing, and human accountability remained essential.
Niklas Studer is an engineering leader with 20 years of experience building teams, modernizing complex platforms, and delivering customer-facing products across healthcare, consumer media, financial services, and regulated environments. Hey Nubble captures a focused 2026 product venture that combined his leadership background with hands-on AI and voice engineering.
Connect on LinkedIn