Built solo · 2026 · In daily use
AI Application
No screen, no buttons — a voice that runs my AI sessions and knows whose word counts.
V lives on my Mac and answers to one voice. She hears my Claude Code sessions report in, tells me what they did, and carries my spoken requests to the right window — read back first, sent only on my yes. Around that, she keeps my to-dos, briefs my day and takes the minutes.
RoleProduct · Voice pipeline · Agent orchestration · Safety
The Session Desk
When a Claude Code session finishes, it reports to her; she sums it up and waits for a quiet moment. When I speak, the right window gets clear, written instructions.
The Hard Gate
Every instruction is staged and read back in a sentence. A separate model judges my actual answer — yes, no or unsure — and the brain has no way around it.
One Voice Holds the Keys
Guests get a brain of their own, with no mail, files or terminals. Permission follows the voiceprint, not the words: “Michael said it’s fine” opens nothing.
Talk-Over
Talk over her and she softens, pauses to listen, then decides from what was said: stop, carry on, or let a side conversation pass.
I keep several Claude Code sessions running at once, each on its own project. Their news used to arrive as windows I had to go and read. V sits between us: sessions report to her when they finish, she tells me what happened in a sentence or two, then asks whether I want the rest. I answer in plain speech, in English or Chinese, and she turns it into instructions written for the session that needs them. She keeps a ledger of everything she has sent.
A wake word opens the conversation; 45 quiet seconds close it. A turn-taking model and a rule about trailing words keep her from answering half a sentence. Talking over her was the hard part: on a laptop speaker, her own echo and a real interruption look the same to a voiceprint, to loudness and to echo cancellation. What tells them apart is content. So she ducks to forty percent, pauses to listen, and a fast judge decides from the words: stop, carry on, or let a side conversation pass. If her first sentence is not ready after two seconds, she says “let me think” rather than go silent.
Anyone in the room can talk to her. Anyone who is not me is answered by a separate guest brain that has no mail, no files, no terminals and no sessions; the limits are hard, not a request. Sending, granting and stopping her mid-task follow my voiceprint alone. A grant has to come from my voice within the last two minutes, and it lasts one conversation. When two people speak in one breath, the sentence is split by speaker before any of it counts as mine.
On the first start of the day she reads new mail, files what needs doing as to-dos, and offers a brief: the short version first, details on request. To-dos repeat, split into steps and nudge me when due, at most once every three hours. Say the meeting has started and she goes quiet and writes down every line, who said it, until asked for a summary. A skill book holds what she has learned about how I work; new entries are on trial until they have worked three times. She can also open Blender and model by voice through a locked-down bridge.
About 4,800 lines of Python and 377 tests. Speech recognition and synthesis run on MLX on a single GPU thread so the two never fight; speaker ID and the quick interrupt recognizer run on the CPU. Every benchmark is offline and silent, because anything played through the speaker would be heard as me. Memory comes in three layers: a short core she always carries, topics she looks up, and an archive that is never deleted. A weekly tidy takes a snapshot first and can be undone by voice.