Case study · Media & Reputation
Watchline: reputation intelligence for X, scored by Claude.
Public figures and executives are talked about on X faster than they can read. Raw notifications do not say which of the day's forty mentions becomes tomorrow's headline. Watchline turns a day of noise into a ten-minute morning check.
- Client
- Watchline
- Industry
- Media & Reputation
- Duration
- Product build
- Team
- Senior pod, LAUNCHPAD + CORTEX
The challenge
What stood in the way.
Notifications from X arrive as a flat stream. They do not show how sentiment is trending, which conversation has to be answered right now, or which mention has the audience to become a story.
Our approach
How we framed it.
Ingest everything, mentions, replies, quote tweets and indirect name references, score it with Claude, and rank by what matters: unanswered, negative, high reach. Scheduled workers on BullMQ keep ingestion and scoring off the request path; signed webhooks and per-channel alert delivery make it operable without watching a screen.
The solution
What we built.
- Multi-account ingestion from X API v2
- Claude scoring for sentiment, emotion, topic and urgency
- Needs-attention queue ranked by audience size; trend analytics
- Configurable alerts delivered in-app, by email, Slack and webhook
- AI-drafted replies
- React 18 / TypeScript / Vite front end; Node.js and Express with BullMQ on Redis; PostgreSQL with Drizzle ORM
Results
What changed.
A morning check that surfaces the handful of mentions worth answering, with a drafted reply ready, instead of a feed to scroll.
Stack
Tools used in production.
- React 18
- TypeScript
- Vite
- Tailwind
- shadcn/ui
- Recharts
- Node.js
- Express
- BullMQ
- Redis
- PostgreSQL
- Drizzle ORM
- Anthropic Claude
- X API v2
Talk to us
Have a similar problem? Bring it to a senior engineer.
A senior engineer plus the relevant department lead joins the first call. No discovery gauntlet, no junior reps.