Skim Web
Building the Credibility Layer of the Internet: The Skim Web Platform.
Skim gets you to the point of any YouTube video or news article. The web platform is the full product surface: infinite article and video feeds, per-brand credibility pages for dozens of news outlets, and analysis pages where every extracted key point carries its own timestamp, impact rating and structured arguments for and against.
The work happens in a Turborepo monorepo that shares a component library, Zod schemas and types with the mobile app. On the frontend that means Next.js and React with TypeScript, TanStack Query driving the paginated feeds, Socket.io behind in-app notifications, and encrypted cookie sessions supporting a login flow that spans Google, Apple and X OAuth plus magic links. The work has not stopped at the API boundary either: I have implemented changes across the backend services the platform runs on.
Because every analysis lives at a permanent URL, the site also generates its own Open Graph images per article and video, so a shared key moment previews properly wherever it lands.
- The problem
- AI summarisers usually hand you one fixed block of text and no way to judge whether the source deserved your trust in the first place.
- My approach
- I work across the stack on the public web platform: a Next.js and React frontend with infinite-paginated feeds, per-publisher credibility pages and analysis views where every key point is deep-linkable and carries its own timestamp, impact rating and for-and-against breakdown, plus changes and features implemented in the backend services that feed it. Sessions are shared with the browser extension, so signing in once covers both.
- The outcome
- skim.plus is live and free to use, and every analysis is a permanent, shareable URL rather than something that disappears when you close the tab.