Project 06 / wearables · ai · productAssembled prototype

A pocket-sized approval surface for an AI agent team

Isometric CAD view of an orange pocket-sized handheld AI module with display, button, and attachment loop.
Project summary

An assembled pocket-sized ESP32 device with a display, physical control, battery, and attachment loop for reviewing agent actions away from a laptop.

Why I built it

To make the AI Agent Team easier to use away from a laptop: the device could show a pending action and let me approve or reject it with a physical control.

From the initial question to a tested result

The assembled module brings a readable approval prompt, tactile control, compact electronics, and a carryable enclosure into one device.

01
01 / Front interaction

Make the next action legible at a glance.

I gave the display most of the front face so an approval request can show its essential context. One tactile control beneath it makes a deliberate response possible while holding the assembled device in one hand.

Front CAD view of the handheld AI module showing a large screen and one button.
Prioritize context, then confirmationA large screen reserves space for the action and its context; one physical control keeps approval intentional and reachable with one hand.
02
02 / Internal packaging

Fit electronics without hiding service access.

I stacked the display, controller, charging hardware, and battery within a shallow enclosure while keeping the internal modules accessible for assembly and revision. Runtime and thermal behavior remain to be measured.

Rear CAD view of the handheld AI module with internal modules and battery packaging.
Stack components for accessThe shallow stack places the battery and controller behind the display while retaining access for assembly, charging, and future revisions.
03
03 / Carry and enclosure

Design for availability, not a desk.

I integrated a carry loop and accessible charging opening into the enclosure so the assembled module can stay close at hand when an agent requests approval. Their durability still needs repeated-use testing.

Isometric CAD view of the orange handheld AI module enclosure.
Build carrying into the enclosureThe integrated loop makes the module available away from a desk, while the charging opening and button remain accessible.
2–3 intarget height~2 indisplay500 mAhbattery

Three decisions that shaped the project

Each decision connects a technical constraint to the choice I made, the analysis behind it, and the tradeoff that followed.

01
Key decision

How should an agent request approval away from a laptop?

What I chose

Pair a readable display with one tactile control

Why

An agent approval is useful only if the user can see the context and respond intentionally. I allocated most of the front area to the display and placed the physical control beneath it for one-handed use.

The tradeoff

A single control keeps the device small and the decision deliberate, but complex edits and longer explanations still belong on a larger screen.

Analysis and evidence / 01 / Front interaction

Make the next action legible at a glance.

I gave the display most of the front face so an approval request can show its essential context. One tactile control beneath it makes a deliberate response possible while holding the assembled device in one hand.

  • Prioritize readable decision context
  • Separate viewing from physical confirmation
  • Keep the control reachable when held in one hand
Front CAD view of the handheld AI module showing a large screen and one button.
Prioritize context, then confirmationA large screen reserves space for the action and its context; one physical control keeps approval intentional and reachable with one hand.
02
Key decision

How should the electronics fit at pocket scale?

What I chose

Stack the display, controller, battery, and charging hardware inside a serviceable enclosure

Why

The screen, battery, controller, and charging hardware share a shallow shell. Keeping those parts accessible made assembly and later revisions practical.

The tradeoff

The compact package makes the device easy to carry but limits battery capacity, thermal margin, and space for wiring and connectors.

Analysis and evidence / 02 / Internal packaging

Fit electronics without hiding service access.

I stacked the display, controller, charging hardware, and battery within a shallow enclosure while keeping the internal modules accessible for assembly and revision. Runtime and thermal behavior remain to be measured.

  • Stack display and electronics within the enclosure depth
  • Reserve space for battery and charging hardware
  • Keep assembly and revision practical
Rear CAD view of the handheld AI module with internal modules and battery packaging.
Stack components for accessThe shallow stack places the battery and controller behind the display while retaining access for assembly, charging, and future revisions.
03
Key decision

How can the device stay available throughout the day?

What I chose

Add an integrated attachment loop and accessible charging opening

Why

The top attachment loop and charging opening let the assembled module travel with the user. Their stiffness, clearance, and durability remain repeated-use test questions.

The tradeoff

The loop adds material and creates a load path that needs testing, while the charging opening must remain accessible without weakening the shell.

Analysis and evidence / 03 / Carry and enclosure

Design for availability, not a desk.

I integrated a carry loop and accessible charging opening into the enclosure so the assembled module can stay close at hand when an agent requests approval. Their durability still needs repeated-use testing.

  • Integrate the carry loop into the shell
  • Balance screen area with enclosure strength
  • Check charging access and button clearance
Isometric CAD view of the orange handheld AI module enclosure.
Build carrying into the enclosureThe integrated loop makes the module available away from a desk, while the charging opening and button remain accessible.

The module has been assembled, with the display, control, battery, and enclosure integrated. The next step is to measure runtime and test the approval interaction with the AI Agent Team.