Why Hindsight Replaced My Manual Reminders

Last night I Daydreamed about an AI that could quietly nudge a student team back on track. By morning, it had already happened: Hindsight had analyzed our overdue tasks and sent personalized, context‑aware reminders to every teammate who’d stalled. What We Built We started as a weekend experiment: a project‑management tool with a real AI backbone — not just another GPT‑4 wrapper with some canned prompts. I wanted a system that could assign tasks intelligently, summarize meetings on its own, track how the team performed over time, and actually remember what happened in the last sprint. Not a Jolt‑or‑Jira clone. Something that gets smarter the longer you use it. The result is a FastAPI backend with nine distinct routers — auth, projects, teams, tasks, decisions, ai, meetings, integrations, and reports — backed by a SQLite database managed through SQLAlchemy 2.0, and a React + Vite frontend running on :5173. The core AI calls go to either Groq (fast, cheaper inference) or OpenAI depending