← Back to the catalogue

WK·005

WatstheStory

A personalised audio briefing system for WhatsApp, built because a short spoken update sounded better than opening ten feeds before breakfast.

StatusPreserved
Period2023-2024
DemoBroadcast assembly
ThemesAI audio · personalisation · messaging
commit orbit · WK·005

29 commits · 13 active weeks
latest 33e3c2d · 16 Jul 2024

Overview

WatstheStory was a personalised audio briefing system delivered through WhatsApp. A listener chose topics, a voice, and a delivery time. The system collected relevant updates, turned them into a short briefing, converted that into audio, and delivered it as a message.

Steve and I built it because we wanted it ourselves, especially for cybersecurity and technology news. It was the kind of product idea that starts with a very small selfish need and then becomes more interesting when other people want it too.

The reconstruction

Choose what the brief is about and who reads it, then assemble it. One finite, spoken catch-up that you can actually play, merged in your browser from the parts you picked.

Topics
Voice

the assembly · in your browser

web audio · assembly
idle · assemble the brief to begin

the delivery · on the listener's phone

Pick a few topics and a voice, then assemble today's brief. It is voiced and merged into one message, like a catch-up from a friend.

One finite brief, only the chosen topics, voiced and merged into a single message.

WK·005 · invented sample stories, no real source, number, or delivery · the voice is real audio, synthesised from the text and merged in your browser from the parts you chose

Context

The bet was that news did not need to be another app. For a morning briefing, the best interface might be the one already on your phone. You receive the update, listen, and move on. No feed. No endless scroll. No pretending that checking five sources before coffee is a healthy information strategy.

We got around 100 signups, which was enough to prove interest and enough to make us look carefully at the operational and commercial reality.

What changed

The constraint

The pipeline worked. The channel was the problem. WhatsApp was a good delivery surface for a prototype because it was familiar and low friction, but the official terms and commercial constraints made the version we wanted to build unsafe to grow.

The lesson

A borrowed distribution channel is useful until it becomes the business risk. We could control the briefing pipeline. We could not control the rules of the platform carrying it. That is fine for an experiment and dangerous for a company.

The build

The architecture was a content pipeline: collect source material by topic, generate a concise script, synthesize the chosen voice, assemble the segments into a single audio brief, and deliver it through WhatsApp at the selected time. The moving parts were not exotic, but making them feel like one daily product was the challenge.

The archive uses sanitized sample audio and reconstructed flows. The source material is not published because the original project included real user and operational data.

Notes from the archive

I still like the idea. A finite audio brief is a good answer to infinite feeds, especially for technical news where the goal is to get oriented, not disappear into a scroll.

I am also still proud of the name. It is a cheeky take on the Irish slang "What's the story?", which is exactly the kind of casual check-in the product was trying to become.

We shut it down because the commercial path depended on WhatsApp rules we did not own. That was the right call. Slightly disappointing, but much better than building a business on top of a policy problem and finding out later with more users and less sleep.

Evidence

repoPrivate repository, reviewed for this archiverepo-reviewed
commitLatest reviewed commit · 33e3c2d2024-07-16
noteTopics, per-listener voice, scheduled delivery by time zone, and WhatsApp deliveryarchitecture

Technical detail

StackPython · RSS ingestion · LLM story generation · Text to speech · WhatsApp delivery · Scheduled jobs
Confidencerepo-reviewed
ConstraintPublic archive copy is paraphrased from evidence. Source repo files and private data are not quoted.

Other exhibits

WK·001AlphAn open source CLI for MCP setup, built because editing the same JSON in five different places is a poor use of anyone's evening.Terminal
WK·002AskHumanA hosted MCP loop for agents that need a human decision instead of another confident guess.Decision loop
WK·003RhyaA live wellness site for a real business, built close enough to the stakeholder that vague feedback was never going to survive long.Capture
WK·004GradienceAn assisted grading workflow that was probably right about the direction, but early for the market and the models.Timeline