I built a voice coding agent to skip the typing
I built Rubber Duck, a voice coding agent. A Todoist Ramble clone for Things 3 that streams audio to OpenAI Realtime with gpt-4o-mini and runs tool calls. Commandment, an open source WisprFlow alternative. Each started from one observation: typing is a bottleneck when the listener speaks natural language.
Based on voice AI projects built between late 2025 and early 2026.
Two voice tools, one annoyance
I built Rubber Duck, a voice coding agent. It streams audio to OpenAI Realtime with gpt-4o-mini, runs tool calls, and drops what I ramble into Things 3. Then commandment, an open source WisprFlow alternative, in one night. Both came from the same annoyance: when the thing listening is an LLM, typing is the slow part.
You feel it after using them for a while. Talking all day is uncomfortable in its own way, but I keep reaching for the mic instead of the keyboard. That is the tell.
Voice removes the translation layer
You do not work out how to type what you mean. You say it, and the agent does it. The friction in typing was never words per minute. It is turning a thought into structured input, and voice skips that step. When the listener is an LLM, the mouth beats the fingers for tasks, code, and controlling tools.
Until voice tools mature, a high performing keyboard is still the fastest physical interface between thought and shipped code.
The interface is collapsing
Every tool I build now points the same way. Speech will replace typing for AI work. The only open question is how fast the tooling catches up to the demand, which is already here.