Build My Brain

Landing!

Most AI assistants tell you what happened. We built one that lets you explore what could happen next. The inspiration was our own week: a hackathon tonight, a CS 61A midterm in two weeks, a linear algebra problem set, four application campaigns and a research dream, with no tool that shows what one choice tonight does to the rest of the month. So we fed the agent our real life: calendar exports, course PDFs, internship and club trackers, a resume, research notes. Cognee turns it into a personal knowledge graph. An AWS Strands agent recalls that context, pulls fresh public opportunities through Bright Data, and branches every decision into a scored scenario map. Change a decision and the future moves: three more hours of midterm prep takes midterm readiness from 54 to 72 while coursework stability slips from 81 to 77. Pick the path you want and it becomes three concrete actions and calendar blocks, behind a confirmation you control.

What we learned is that a brain is only as trustworthy as its evidence trail. Every inference on the profile carries a confidence score, a "why?" drawer back to the source, and a reject button, and we refused to let the model invent probabilities: the LLM only proposes factor values, the score itself is deterministic code with named weights, and every number is labeled a scenario estimate rather than a prediction. The hardest parts were keeping five hours of build honest about that line, making the personal data feel personal without ever inferring anything sensitive, and shipping an end-to-end demo that works with or without API keys. Your future is not predicted. It is shaped.