Inside the
AI HAT+

A camera, a Linux host and a dedicated inference chip. Follow a frame through all three.

Follow a Gen 1 camera frame from the glasses into a permitted Android session, then decide what the phone should keep.

Follow one meeting recording from the original NotePin’s microphones to an action item you can check against the speaker’s own words.

One camera ribbon carries an image and a neural network’s answer about that image. The Pi decides what to do with each.

Train a board to recognize a hand wave, then see what happens when someone else wears it.

Turn a repair note into a checked work order on a computer small enough to leave on the bench. Follow a prompt through its shared 8 GB of memory.

A camera, microphone and tiny display you can program over Bluetooth. Original Frame remains documented; new-unit availability is unverified.

A small ring turns optical and motion signals into daily records. Build a reflection journal that shows where its data ends and interpretation begins.

An open-source voice pendant with account APIs and published hardware files. Trace a spoken correction into a proposed task that stays linked to its transcript.

A small voice satellite with an unusually clear boundary: the puck listens and speaks, while your chosen Home Assistant pipeline decides what the words mean.

A detection box tells you what is in view. Two more cameras help estimate where it is. Build a distance-aware desk zone and inspect when that estimate fails.

Teach a thumb-sized camera to recognize a few bench states, then send a small event instead of a video stream. The hard part is making that decision survive new lighting.

Give a vision model a physical response: a glance, a nod or a curious tilt. This desktop robot lets you build the interaction loop and test whether its gestures actually help.

Turn a camera detection into a useful local event without adding a separate Pi host. The interesting part is the rule between “I see a bottle” and “the bottle has been removed.”