All projects

Walkthrough

The dashboard concept

The interface concept that sets the target for the build. The capture services are being built first, so these screens show the intended product with sample content rather than live data.

SignalSift

1. 01The promise, written downThe goal on one screen. A save should be confirmed in under five seconds, even offline, while workers transcribe, summarise and tag it afterwards.
2. 02One save, end to endEach capture moves through the pipeline, from captured to extracted, summarised and indexed, and only asks the user for input when extraction is blocked.
3. 03Honest about each platformFinished cards are tagged and searchable. A risk panel records which platforms have clean data access and which need user-assisted fallbacks.

Previous1 of 3Next

A design concept with sample content. The handles and posts shown are illustrative, not real saves.

1. 01 Problem

   Useful things get saved inside social apps and never seen again. Each platform keeps its own saved list, none of them are searchable, and none of them talk to each other.

2. 02 Approach

   Before building, I evaluated whether no-code tools could do it (Shortcuts, webhooks, Zapier). They can't deliver a real share target or reliable metadata, so SignalSift is a lightweight capture app plus a queue-backed ingestion service that summarises and tags each save.

3. 03 Implementation

   A TypeScript monorepo with a web app and three services (dispatcher, worker and reconciler) on Postgres. Every capture moves through an explicit state machine, and state can only change through one transition function, so retries and failures stay predictable.