[quote="gu_jinyu, post:1, topic:1555"]
Last week at our team's technical review meeting, the PM presented the latest user research: Gen Z is willing to pay over 3000 RMB for electronic pets that "never get sick, don't shed fur, and are always available." Team members looked at each other blankly, and someone muttered, "Isn't this just a plush toy with AI added?" But I know the real challenge isn't technical implementation, but how to ensure a product that "can't die" isn't seen by users as "soulless."
Conclusion first: The explosion of the AI pet track is essentially an engineering victory of "affective computing" moving from labs to consumer goods. But can the team consistently deliver trustworthy companionship…
[/quote]
I've looked at several projects in this track, and my biggest worry is that the technical moat is too thin. What you said about "STM32 plus GPT-4o API" is basically the truth, so the team background and core algorithm capabilities determine the moat. Have you considered an exit path for continuous model iteration within your "Three Gates" mechanism?