A modular wearable · Physical-MCP (PCP)
aibao = AI Bisected Autonomous Orchestration

Program it by speaking.
A body you can wear.

Camera, microphone, speaker, haptic motor, drive motor, servo, display — each one is a small, independent computer. Say a single sentence, and the hub turns it into coordinated action across these modules.

— interested Learn more

No prototype. Nothing for sale, no payment taken. One click simply tells me it’s worth building.

✦Hub
📷 Camera 🎙️ Microphone 🔊 Speaker 📳 Haptics ⚙️ Motor 🦾 Servo 🖥️ Display

Story

It started with a simple question.

Why should a smart device’s abilities be fixed at the factory?

Today’s wearables come as one sealed unit. If the camera fails, you replace the whole thing. If you want a silent vibration alert, no model offers one. If you want it to remind you differently while you cook, you wait for the manufacturer to ship an update.

What I want to take apart

Abilities locked inside a single enclosure. Features chosen by the manufacturer rather than the person wearing the device. Changing one rule means learning to code first.

My answer

Take it apart. Each module is its own ESP32 that can describe what it does. Plug it in and the hub discovers it; unplug it and that ability is gone. The modules you wear today are the abilities you have today.

Then make programming as simple as saying one sentence: “When the kitchen timer goes off, don’t make a sound — just buzz twice.” Say it once, and it keeps doing exactly that.

Capabilities

Four ideas that make up a body.

These are design goals, not existing features. See “Progress” below for what has actually been done.

🧩

Separate modules

Every module is a small, independent computer rather than a part on one large board. If one breaks, you replace that one. If you want a new ability, you add a module.

  • One ESP32 per module, each running on its own
  • Each module declares what it can do, how quickly it responds, and what it can’t run alongside
  • Plug it in and the hub finds it; unplug it and the ability disappears
  • You decide each day which modules to wear
🗣️

Speaking is programming

Instead of opening an app to change settings, you simply state the rule you want. Once said, it stays in effect — there’s no need to repeat yourself.

  • A rule spoken once is compiled into a small program that lives on the module
  • Decisions are made on your body, not by asking the cloud every time
  • Keeps working without a signal, and uses less power
  • Anything said by mistake can be undone
🎙️

Four-microphone hub

The hub isn’t the brain; it’s the switchboard. The model that makes the decisions runs in the cloud or on a computer at your home — a small chip can’t run it.

  • Discovers modules and describes their abilities to the model
  • Schedules actions — for example, so the microphone doesn’t record the speaker
  • Four microphones can tell which direction you’re speaking from
  • Falls back gracefully when offline instead of failing entirely
👕

Worn on the body

Each module sits where it belongs: the camera facing forward, haptics against your skin, the display where you can glance down at it.

  • The camera sees what’s in front of you
  • The haptic motor alerts you silently
  • The display is there only when you need to look
  • Form factor and power supply are still undecided — see the FAQ

Use cases

One sentence, one new feature.

These are uses I have in mind. None of them works yet.

🍳

When your hands are wet

“When the kitchen timer goes off, don’t make a sound — just buzz twice.” Say it once, and it works that way every time you cook. No touching your phone, and no alarm ringing through the house.

MicrophoneHaptics
🦯

When it’s hard to see

When the camera sees a person or object approaching, the haptic motor alerts you; when it sees text, the speaker reads it aloud. An extra pair of eyes for people with low vision.

CameraHapticsSpeaker
👀

When you’re focused on work

“If someone walks toward me, show it on the display.” With headphones on and your back to the door, you no longer need to keep looking over your shoulder.

CameraDisplay
❓

The open question

I haven’t worked out what motors and servos should do on a wearable. Assisting movement? A camera that turns on its own? If you have an answer, tell me at the bottom of the page.

MotorServoYour idea

Progress

From idea to wearable, in five steps.

Completed and current steps are marked. The rest haven’t started.

01
Done

Project start

Settled the core idea: every module is an ESP32 that can describe itself.

02
In progress

Finding out who wants it

That’s the page you’re reading now.

03
Not started

The first module

The haptic module learns to describe itself: what it can do, how quickly it responds, and what it can’t run alongside.

04
Not started

An unfamiliar module

An AI makes good use of a module it has never seen before, relying only on that self-description.

05
Not started

Speaking is programming

A spoken rule is compiled into a program that lives on the module, with no connection needed afterward.

What you’ve said

Good or bad — in your words.

Only feedback from real people, published after I’ve read it. Nothing here is made up, and critical feedback stays up too, unless it’s spam or abuse.

—
interested
—
messages
7
modules
0
prototypes

No messages have been published yet — the first one could be yours.

FAQ

Things you might want to know.

If your question isn’t answered here, leave it at the bottom of the page — I can’t reply individually, but I read every one.

No. There is no prototype and no price, and I’m not taking payments or pre-orders. Clicking “I want this” only tells me whether it’s worth building — that number decides whether I put more time and money into parts.
The goal is no: saying a sentence is the programming. That part hasn’t been built yet — it’s step 5 under “Progress”.
Not by design. Understanding the rule you speak requires a model in the cloud or on a computer at home. Once it’s understood, the rule is compiled into a small program that lives on the module, and every decision after that happens on your body. It keeps working where there’s no signal.
It can’t run one. I’ve tested this on the same chip: the largest language model that fits has fewer than 30 million parameters, which isn’t enough to decide which modules to use or how they should work together. So the hub acts only as a switchboard — it discovers modules, schedules actions, and falls back gracefully when the connection drops. Decisions are left to a larger model.
Honestly, I haven’t worked that out yet. It also determines the safety requirements: once something on your body can exert force, “stop immediately when commands stop arriving” is no longer a nice-to-have — it’s the minimum. If you have ideas, tell me below.
Not decided yet. With separate batteries, you’d charge several things every day; with a shared one, there’s a cable running across your body. It’s a question of form, not just circuitry.
The counter keeps only an irreversible hash of your IP address — the IP itself isn’t stored, no cookies, no tracking; feedback is stored with the same hash, to limit spam. Your IP is checked on this server to see whether it comes from mainland China, which decides whether the page opens in Chinese or English; it’s discarded right afterward and never sent to any third party. The payment email is optional and used only for the reward: to pay you and to apply the per-person limits. It is never published with your feedback, never put on a mailing list, and shared with no one except Interac (via my bank) or PayPal to make the payment. Feedback without an email is just as welcome. I read every piece of feedback before deciding what to publish.

Where does it fall apart?
Tell me.

I’ve heard enough encouragement. What I need is the case against it: what can’t be built, which situation nobody actually has, what I’m fooling myself about. One well-aimed objection is worth more than a hundred clicks on “I want this”.

The first 100 useful critiques get a US$1 reward each.

What counts
Point at something specific that doesn’t hold up — it can’t be built, the need isn’t real, the cost doesn’t add up, someone already does it better, or there’s a safety risk I’ve missed. “It’s bad” or “this will go nowhere” doesn’t count, and neither do points this page already makes (including the two examples in the form) or ones someone else has already raised. I read and judge each one myself.
How many are left
— of 100 slots remaining (a slot is taken when I accept a critique, not when it’s submitted)

All 100 reward slots have been claimed. Criticism is still welcome and I still read every one — I’m just no longer paying or collecting emails.

Today’s feedback limit has been reached, so the form is paused for now — please come back later.

Ideas and questions are welcome here too.Only criticism earns the reward.

0 / 1000 characters

Used only for the reward: to pay you and to apply the per-person limits. Never published, never put on a mailing list, and shared with no one except Interac (via my bank) or PayPal to make the payment.

Payment method (only needed if you leave an email)