A lot of AI features don't need the cloud. I build them to run on-device, so the data never leaves the user's machine. Audient records a meeting and summarizes it locally, Whisper transcription plus a local LLM, zero data egress. imlistening does real-time offline dictation on macOS. SLMAndYou runs quantized small language models right on a phone. If you want an AI feature that's private and actually ships, that's my thing. More at kumarsomesh.com.