W1: Three Project Ideas
Idea 1: an agent for knowledge. I write and collect a lot of information but rarely organize it. Knowledge only becomes useful when it is connected to other knowledge, and making those links by hand is slow. This agent would read my whole Obsidian vault, suggest links between notes and explain why each one makes sense, so I can spend my time producing new ideas instead of indexing old ones. The spatial side of the idea comes from Giulio Camillo's sixteenth-century Theatre of Memory, a room meant to hold the sum of all knowledge.
Idea 2: an agent for reliving memory. Old photos, recordings and letters carry triggers that take us back to specific moments. Like the "Eulogy" episode of Black Mirror, this agent would connect those fragments into a virtual scene that you can walk through, correct and rebuild, so remembering becomes something you actively do rather than something that fades.
Idea 3: an agent for recording dreams. Dreams fade within minutes of waking and are hard to write down in that state. This agent would take a dream by voice as soon as you wake up, turn it into a visual record, and track recurring people, places and moods over months, showing patterns you can't see in a single dream.
W2: Human-human interaction
This week I met two classmates to talk about our final project ideas and the agents we were building.
Matti Gruener and I talked through my idea of an Obsidian-based visual "memory theater", a second brain you explore with an agent, and the Jev-powered on-screen concierge I built this week. We also discussed visual UX, where the truly agentic part of an idea ends and the interface begins, and how AI is changing design work and who owns it.
With Lilith Yu, I shared my interest in building a personal knowledge concierge agent. Lilith shared her idea for a new kind of OS with no apps or traditional UI, just a modular workspace that agents rearrange around what you are trying to do. I shared my own experience with modular workspace apps like AeroSpace and how one could be adapted into an agentic workspace.
W3: Final project group
I pitched my dream project to Mateo Larrea, and we decided to team up on it for the final project around the idea of an agent that helps you record your dreams when you wake up, plus a soundscape that helps you fall asleep.
Our expertise come in handy when setting out scope of the works. I am more interested in building the agent and its long-term memory, and the visual and spatial side. Mateo prefers to works on audio world modeling, which generates sound environments from a space or a prompt and follows how they change over time, and he brings experience with memory and retrieval.
Two ideas came out of our call. The agent could generate a scene from your description of a dream, to help you remember more of it. And before sleep, the sound and light around you could change with your mood and your day. However this is our first conversation and we both want to make something bigger than a mobile chat bot or app, so we will be meeting in person to discuss the more concrete scope of the final project soon.
