// Progress

The idea is proven. The product is being earned.

Frank is no longer waiting for an end-to-end proof. Idumet is turning a working Android assistant into a faster, safer, more contextual and more dependable product through measured development and private testing.

01

The starting idea

Hands-free visual assistance

Frank began as a smart-glasses concept: a camera, natural questions and spoken answers for everyday visual access.

02

The first proof

Raspberry Pi prototype

A standalone camera, microphone, wake word, cloud intelligence and spoken response proved that the full loop could work. It also exposed the cost, bulk and reliability burden of dedicated hardware.

03

The decisive pivot

Software first on Android

Moving to a phone removed wires and component cost while adding mature accessibility APIs, connectivity, identity, location and sensors. Raspberry Pi is now historical, not actively supported.

04

Frank v0

The end-to-end assistant works

The native app listens, captures visual context, reasons, speaks, continues conversations and stores useful memory. Diagnostics identified model response time—not memory or database work—as the main speed bottleneck.

05

Frank V1 now

Preparing for controlled testing

V1 adds protected server access, request-level tracing, fresh camera capture, structured context and product-data contracts, accessible feedback, guided setup and private Google Play distribution.

Early wearable Frank prototype
Smart-glasses exploration
Raspberry Pi camera and voice Frank prototype
Working hardware proof
Frank Android listening interface
Android product direction

// What comes next

Improve what matters in the order it matters.

This is a development sequence, not a public release promise. Each step must pass accessibility, privacy, reliability and safety evidence before the next claim is made.

Now

Speed, reliability and control

Reduce slow model responses, stream useful audio earlier, complete per-request diagnostics and strengthen recovery, memory controls and camera freshness.

Next

Secure data and Context Snapshot

Work with Chivora on the intended scalable data foundation, then validate location, movement and journey-aware assistance that works from a pocket.

Then

Offline resilience and wider pilots

Add a deliberately narrower on-device voice, text and context path; polish accessibility; run a controlled blind and low-vision pilot; explore iOS and later communities only with evidence.