Your microphone, only if you ask

They answer the phone.
You keep the last word.

Voice agents that handle a call on their own and stop before anything that matters. Small enough to run on a CPU, so the audio never has to leave the building.

The whole idea

An agent that cannot be stopped is not autonomy. It is exposure.

Most voice products ask you to choose: let it act, or listen to every call. The interesting line is narrower than that. Let it act on the ninety percent that is answering questions, and make it stop on the ten percent that spends money or changes a record.

Inbound · Northside Dental

An agent takes a call on its own. Watch what happens when it reaches the part that spends money.

Approval lists, not prompts
You name the actions worth stopping for — money out, a cancellation, anything touching a record. Everything else the agent just handles. The list is the contract, and it is readable by the person who signed it.
It waits well
A held call does not go silent. The agent keeps the caller company, tells them the truth about the wait, and hands over a transcript so whoever picks it up is not starting cold.
Every yes has a name on it
Approvals and refusals are both recorded against the action. When somebody asks who authorised it, that is a lookup and not an investigation.
Small on purpose

It runs on the computer already behind the desk.

A voice model does not need a datacentre to answer a phone. Ours are built to fit on ordinary CPUs, which changes three things at once — and the third is the one that closes deals.

No GPU bill

The unit economics of a phone line stop being an inference invoice. That is what makes it sellable to a dental practice rather than only to an enterprise.

Answers immediately

Latency you can hear is latency that loses the caller. Running next to the phone system removes the round trip that causes it.

The audio stays put

Nothing leaves the building unless you send it. For anyone handling health, legal or financial calls, that is not a feature — it is the reason they can say yes at all.

It does not go stale

Agents that read the work as it is published.

Small models get better by being taught, not by being replaced. A reading agent works through new papers and open source on its own, keeps the few findings that would change how the voice models are trained or served, and leaves the rest. The same discipline as the phone: act freely on the gathering, stop before the change.

  1. 01
    Read

    Everything new in speech, quantisation and small-model training.

  2. 02
    Discard

    The overwhelming majority. A finding that changes nothing is noise.

  3. 03
    Propose

    A specific change, with the reason and the expected cost.

  4. 04
    Wait

    A person approves it before it touches a model in production.

Put one on a phone line you care about.

We are taking a small number of first lines — the kind where a wrong answer actually costs something, because that is where the guardrail has to earn its place.

One reply from a person, not a sequence.