Buffy is built around a few core ideas. Understanding them will help you get the most out of the product.
Activities: habits, tasks, and routines
Everything you want to track or act on is an activity in one model:
- Habit — A repeated behavior (e.g. “drink water,” “evening walk”) with a schedule (e.g. daily, weekdays) and optional time window.
- Task — A one-off with an outcome and often a due date (e.g. “ship report by Friday”).
- Routine — A sequence of steps on a schedule (e.g. “morning startup: water, planning, stretch” between 7:30–8:00).
They share the same structure: schedule, context, and history. That’s why Buffy can show one daily briefing, one reminder flow, and one history for all of them.
Channels
Channels are where you talk to Buffy and where reminders show up:
- ChatGPT — Planning, defining routines, and longer conversations.
- Telegram — Quick nudges and “done / snooze / skip” on the go.
- Slack — Team rituals, standups, and shared routines.
You’re not locked to one. The same behavior core powers all of them; you choose where to plan and where to be reminded.
Memory
Buffy uses three layers so it can adapt over time:
- Short-term — Recent conversation and context (e.g. “move that to tomorrow,” “mark the second one done”).
- Episodic — What actually happened: completions, skips, snoozes, response times. This is the ground truth for “how often did I do this?” and “when do I usually complete?”
- Semantic — Patterns derived from history (e.g. “morning nudges work better than evening”) so suggestions and reminder timing can improve.
Reminders
Reminders are tied to activities and time windows, not just a fixed clock time:
- One nudge near the start of the window, optional follow-up near the end, then quiet.
- Clear exits: done, snooze, skip.
- Behavior and channel choice can adapt using memory (e.g. nudge in Telegram if you usually respond there).