Blog
Engineering notes
What actually decides whether an AI hardware product ships — written from the projects, not the datasheets.
-
Sep 30, 2026
Why Your AI Companion Toy Feels Slow — the Latency Budget Explained
Users judge a conversational toy on how long it stays silent after they stop talking. Most of that time is spent on the device, not in the LLM — here is where the milliseconds go, and what to ask your ODM.
Read the note -
Sep 30, 2026
The Fastest Way to Launch an AI Talking Toy: Build Around a Core Module
Brands shipping AI plush toys first are not designing boards from scratch — they build around a proven conversational core. What separates a production-ready core from a demo board, and why the module choice is a business-model choice.
Read the note