Epigen

Ask any AI "Should I buy this \$180 jacket?" and you get the same generic advice everyone gets, because the model knows nothing about you. Epigen is a personal memory layer for AI. It stores your own history as memories across life angles (shopping and family in this demo), retrieves only the ones relevant to the question, and answers with citations to the exact memories it used. In the demo, the same question gets two answers side by side. Plain Muse gives generic trade-offs. Epigen says don't buy: you walked away from a similar \$180 wool jacket, and your household rule says any purchase over \$150 is weighed against the apartment fund. Log the outcome as "Returned", ask again, and the answer changes. This is not basic RAG. Outcome feedback changes the next answer. One decision combines memories from different life angles. Every citation is verified against what was actually retrieved. And the memory is a reusable MCP layer that any AI tool can call, not a chat feature. The idea comes from epigenetics: every memory is stored, but only the relevant ones are expressed in each answer. The model stays general. The memory gets personal. Built from scratch Nov 3-4, 2026, during the hackathon window. Pre- existing: open-source libraries only. The commit history shows the full build, and the video shows only features built during the event. All people and data in the demo are synthetic.