bayti (in-progress)

during my recent AI studies, i was thinking about having a complete local AI assistant that would help me to manage my household. know our habits, needs, and focus on assisting us : my family and myself.

• imagine an agent that know the children school schedule, and that i just ask to prepare a study week for them, to write exercices and lessons.
• to write me a complete grocery list based on our previous ones and sent it to me by whatsapp
• to connect it to lights and homepod and ask it if i correctly shut down the lights in the living. if not, to do it.
• and so many more fields to explore.
• to add to our shared calendar all the next UFC events based
• check the next birthdates and suggest gifts based on our current budget

An agent needs context to be efficient. so to have the best house assistant, i needed a workspace for the whole family, so built Bayti¹.

that’s what i did. I built the platform with Convex (to also learn it).

Eventually, I imagine Bayti becoming an intelligent interface for the home itself: connected to devices such as HomePod and other services, able to understand household routines, control parts of the house, surface relevant information and coordinate the family.

“Turn off the living-room lights.”

“Someone from the family is at the door.”

“What do we need to pay this week?”

“Add this invoice and remind me before it’s due.”

Bayti was initially just a solution to my own problems. I haven’t had enough time to pursue the full vision, so I released the project as open source for now.

But it represents an important transition in the way I build software: after spending a long time experimenting with AI, I became comfortable enough with the technology to stop treating it as a feature and start designing products where intelligence is part of the architecture.

I absolutely want to come back to this one.

https://github.com/cowboyshibuya/bayti