Detto is a local-first macOS application for Apple Silicon that captures meetings and voice memos with automatic speech recognition running on the Neural Engine and text refinement on the GPU. It transcribes audio to markdown with YAML frontmatter, integrates with Obsidian vaults, and includes an AI agent layer that processes notes automatically to extract action items, update client files, and connect information. All processing happens locally after initial model download, with no data leaving the device.