BrainDump: Speak Freely. Leave Organized.

People don't think in perfectly structured commands. We ramble, combine multiple thoughts, change our minds halfway through a sentence, and remember unrelated things at the same time. Most productivity tools force us to organize those thoughts before we can capture them. BrainDump takes the opposite approach: speak freely first, organize later. BrainDump is a voice-first personal inbox that uses AssemblyAI's Voice Agent API to turn natural, messy speech into structured tasks, reminders, ideas, and notes. A single brain dump can contain multiple intents, and BrainDump separates them into individual draft items that can be reviewed, corrected, and approved. The key interaction is conversational correction. Instead of manually editing a form, users can say things like "Actually, make that reminder 7 PM," and BrainDump updates the relevant item in context rather than creating a duplicate. Once approved, items are persisted in a personal inbox and can be searched or retrieved conversationally later. The architecture keeps the AI responsible for understanding and proposing actions while the backend remains the source of truth. AssemblyAI handles the real-time voice interaction, while a FastAPI backend validates tool actions and PostgreSQL stores the structured information. BrainDump is designed around one simple idea: capturing a thought should not require interrupting the flow of thinking. Speak freely. Leave organized.