Echo cancellation for real-time output is a nasty problem. I hit similar walls with a browser-based take on this (livesuggest.ai) — you get easier distribution but lose a lot of control compared to native. The Founders real-time suggestions piece is what I'd want to try most. What's the end-to-end latency like in practice?
Been running something similar called LiveSuggest (livesuggest.ai) for a while. The browser-only vs native desktop tradeoff is interesting. No install means anyone can try it in 30 seconds and it works on any OS, but you give up the local-first angle. Curious if you considered a web extension route at any point, or if local model support was always the requirement that pushed you toward a native app.
The MCP angle is clever, especially for devs already living in Cursor or Claude. I've been working on the other end of the workflow with livesuggest.ai, real-time suggestions during the call rather than memory after. Local-first for privacy makes sense, we went browser-based with no audio storage for similar reasons. Complementary tools depending on where you feel the gap.
Makes sense for coding, hard to distinguish from just looking up answers. I use LiveSuggest.ai for regular work meetings and it's a different experience. more like having a second brain for staying on track mid-conversation, not solving problems for you. Different use case entirely.
Cool project, the on-device STT angle is solid. I've been working on something adjacent, LiveSuggest, which focuses more on real-time suggestions during the call rather than transcription, same no-bot approach via system audio capture. Different use case but the "no bot joining" thing resonates, people really don't want that. livesuggest.ai if curious.