Core Memory

WIP: This page will be updated regularly
Reimagining Memory: A New Interface for Capturing and Sharing Memories, 2024 ~ (On Going)

This project stared as an extension of Memorscope (Click the link!).

Imagine you’ve gone on a trip with your family and friends.
What if you could revisit those precious memories without sifting through hundreds of photos dumped into a shared album that you never actually look back at?

Memorscope is an AI-powered Collective Memory sharing device. It was presented at NeurIPS, one of the biggest AI conferences, and presented at MIT News.

But it didn’t solve the problem—it was more of an artwork. Still, it revealed a hint:

What if we could create one image that understands the context and stories behind your many photos?

What if we could talk to that picture?
What if we could zoom in—and dive deeper into the memory?
What if we could archive our lives as contextualized stories—not as a data graveyard?

I take many many photos.

Like, a lot.

As of today, I have more than 200,000 items in my photo album.

And I’m not alone.

I talked to friends, and many of them do the same.
We document everything — but then what?

What do we do with those photos?
Do we go back and look at them?

For me, the photo app just feels like a grid.
It’s not a memory — it’s just a storage system.
It’s a shelf, neatly arranged by time.

So what I wanted was to turn this grid into something else —
a story interface, not just a timeline.

Turing data into stories

Meanwhile,

The current technologies are asking questions to us :

Is memory what happened? Or is it what we choose to remember?

If you’re team ‘what we choose to remember’, then your friend is Google.
(Read: Google says Pixel cameras don’t take photos, they create memories)
They allow you to take a selfie and put whatever you want - like, a UFO!

If you’re team ‘what happened’, then your friend is Apple.
They offer traditional digital ‘camera features’, not a crazy ‘add a UFO feature’

What if there is an alternative?

What if, we could create a picture of one day of your life compressed not as data, but as a story?

What if, we could have one picture of our days, weeks, months, and years?

1 minute intro Video

Core Memory is a contextual memory AI interface that blends wearable sensing, generative AI, and ambient interaction to create a new way of engaging with personal memory.

Technical foundations

The product takes the form of:

• A wearable device with a camera, microphone, and IMU sensors for passive, context-rich capture (time, location, motion, audio-visual cues)

• A memory engine built on large language models and vision models, clustering and summarizing life fragments into narrative sequences

• A story generation layer that uses generative image models to produce blended visual memories—evoking emotional context, not just factual recall

• A gesture-driven projection system for ambient, physical interaction with memory

• A supporting mobile/web interface for browsing, editing, and sharing

Rather than serving as a passive archive, Core Memory acts as a contextual assistant for memory—retrieving, generating, and blending media based on time, location, people, and emotional tone.

This is not a one-step launch; it’s a step-by-step system. I will begin by releasing the mobile/web app to validate the interaction model and build an active user base. In parallel, I’ll develop and test the wearable hardware prototype with a smaller group, moving toward a seamless, integrated experience over time.

Ultimately, Core Memory lays the technical foundation for a longer-term vision: rethinking memory as a form of active reflection and human authorship, with AI supporting—not replacing—the storyteller.