Engram

What inspired us:

We got stuck on one question and could not let it go. Why does nobody talk about forgetting as a real problem? We talk about disease, climate, poverty. Nobody talks about the fact that we lose most of our day within hours, that we lose our entire childhood before age four, that 57 million people worldwide are watching their whole past disappear to dementia. Forgetting just feels like the weather. But the second we realized memory loss is a hardware limitation and not some unavoidable fact of being human, everything clicked. Hardware limitations are engineering problems. And engineering problems can be solved!

That is when we found the research that changed everything for us. A real hippocampal prosthesis, tested on real human beings, improved memory by 35 to 37 percent. A device wrote memory into people. That single fact blew our minds and became the seed for Engram.

What we learned:

We learned that brain-computer interfaces are moving incredibly fast. Companies are racing to build the hardware, the chips, the electrodes. But almost nobody is building the software layer that turns raw brain signal into something private and usable. That gap is enormous, and it is wide open right now. We also learned that memory itself works completely differently than we assumed. The brain does not record memory like a camera. It reconstructs memory every single time you recall something, and it quietly changes it in the process. That single insight shaped our entire architecture. We were not trying to build a recorder. We were trying to build something that captures the meaning of a moment, not just the footage of it.

How we built it:

We designed Engram as six layers, each one mirroring something the brain already does naturally. Capture takes in the raw moment. Encode turns it into a structured memory tagged with time, place, people, and emotion. Consolidate mirrors how the brain replays and strengthens memories during sleep. Vault keeps everything encrypted and local, with the user holding the only keys. Recall lets you ask questions in plain language and get real answers back. Grant lets outside apps request access to a single question, never a copy of your whole life.

We built a real working prototype that runs this entire pipeline today, using a webcam for facial expression, voice and text notes, and a local on-device memory store. We built an "Ask Engram" feature where you can literally ask your own memory questions in plain English. We also designed a full hardware concept, including a CAD model of the actual chip, so the project does not stop at software. It points all the way to the future where this lives on a real BCI.

Challenges we faced:

The biggest challenge was reasoning through something nobody has actually built yet. There was no existing playbook to copy. We had to build the entire architecture from first principles, asking why every existing solution fails, and then designing something that solves the actual root problem instead of just making a better recorder. Privacy was another huge challenge. The moment you are dealing with someone's entire life, you have to take ownership and security incredibly seriously. We had to design the system to be local-first and end-to-end encrypted from day one, not bolted on as an afterthought. And finally, connecting real peer-reviewed neuroscience to a believable, staged hardware roadmap took a lot of research and a lot of late nights. We wanted Engram to feel real, not like science fiction. We think we got there!