Build

TipsyThoughts

A mobile prototype for recovering messy voice notes through local audio, AI cleanup, and next-day review.

Status

Building

My role

Product framing, Expo mobile prototype implementation, trust-model design, and iterative feature shaping

Best next step

Pick the fastest way to see the work or ask for the walkthrough that matches what you want to understand.

Ask George about this build

What it solves

Most note tools assume a user is coherent when they capture something. TipsyThoughts starts from the opposite assumption: the idea may be useful, but the moment of capture may be chaotic, tired, late-night, or otherwise messy.

Why it matters

The current prototype already records locally, preserves session history, supports playback, cleanup review, tags, search, reminders, and simple resurfacing logic.

How AI fits

AI is part of both the product concept and the build process. The app is designed to use AI for transcript cleanup and summary generation, but the stronger product decision is the trust model: the original audio remains primary, while AI output is treated as a recoverable interpretation layer rather than unquestioned truth.

What it is

TipsyThoughts is a voice-first idea-recovery build. The point is not to create a perfect transcript on the spot. The point is to save the recording, preserve the messy original, and make it easier to recover what still matters later.

What’s in the prototype now

One-tap audio capture in a mobile Expo app
Local audio save before any optional cloud processing
A review screen that separates raw recording, raw transcript, cleaned note, and summary
Lightweight tags and search across saved sessions
Morning-after reminders and a small resurfacing layer for older notes worth revisiting

Why it matters

It shows product thinking around trust, not just interface polish
It treats AI as a support layer instead of pretending the model always knows what the user meant
It is a practical example of founder-style product shaping: start with a narrow, believable loop, then build toward retention and recovery rather than novelty

Approach

Build a mobile-first Expo / React Native prototype around one-tap recording, local audio persistence, and honest review rather than polished note-taking theater.
Keep the original recording as the source of truth, then layer in raw transcript, cleaned note, and summary states so interpretation never replaces evidence.
Add lightweight tags, search, next-morning reminders, and a `worth revisiting` view so the archive becomes reusable instead of turning into a pile of ignored voice memos.
Leave cloud cleanup optional, with a local-first fallback that still preserves the recording and review flow when transcription is unavailable.

Tools

Proof points

Built a working Expo prototype with one-tap recording, local file persistence, and session history
Structured the review flow around original audio, raw transcript, cleaned note, and summary instead of flattening everything into one AI-generated output
Added lightweight tagging, in-app search, and a `worth revisiting` section to make saved ideas easier to recover later
Implemented next-morning reminder scheduling and optional cloud-processing hooks without making cloud AI a requirement for the product to remain useful