I keep coming back to the same problem: I get ideas faster than I can capture them.
My phone can technically do almost anything, but there’s still friction. Unlock it. Find the right app. Open a note. Remember what I was trying to say before another notification steals my attention.
I want something simpler.
A small physical device I can carry, press, and speak into when I want to talk to my AI. One button. One microphone. A visible light that makes it obvious when the microphone is active.
That idea became Harpocrates.
Harpocrates is meant to be a private physical gateway to Hermes, my AI assistant. Hermes is the intelligence. Harpocrates is the thing I hold.
The name comes from Harpocrates, the Greco-Egyptian figure associated with silence and discretion. You’ll usually see him with a finger near his mouth. That image started as a sign of childhood, but Greek and Roman culture eventually connected it with secrecy and silence.
That felt right for a device sitting between my voice and an AI.
The important part isn’t making another voice assistant that’s constantly waiting for a wake word. I want the interaction to be intentional. You press a real button. A light turns on. You speak. The device sends the recording only after you’ve asked it to.
Quiet until you press the button.
The current plan uses a small ESP32-C3 board, a digital microphone, a rechargeable battery, a button, and some kind of visible status light. The ESP32 itself won’t contain a complete AI model. Its job is to record the request and securely carry it to a phone or server where speech-to-text and AI processing can happen.
I already own a larger ESP32-S3 board for early testing. The smaller C3 is the proposed wearable version.
And to be clear: I have not built the device yet.
Right now I have the concept, the visual identity, one early test board, a proposed parts list, and a growing collection of questions. I still need to verify the smaller board, wire the microphone, capture an audio sample, decide between Bluetooth and Wi-Fi, work out the battery system, and eventually build an enclosure.
It’s ugly, but it works hasn’t even happened yet. First, I need to make it work.
The privacy side matters just as much as the hardware. I don’t want a hidden microphone quietly recording in the background. The user should always understand when the device is listening, what it’s sending, and where that information goes.
I’m not going to call it private, encrypted, or secure until those things are actually built and tested.
Eventually, Harpocrates could become part of a larger system: my website as the public world, an app as the private control center, and Hermes moving information between them. Harpocrates would give that system a physical doorway.
But the immediate goal is much smaller.
One microphone. One button. One little board.
I want to be able to capture an idea before I lose it.
That’s really the whole thing.