Ship something. Not a todo app. Not a counter. Not a clone of an app that already exists. Build something that solves a real problem you actually have. The difference between a portfolio project and a toy project is whether you would use it yourself on a Wednesday afternoon when nobody is watching.
Your portfolio needs to tell a story, and the story is: "I understand the full stack." Three projects, three architectures. A pure React SPA that shows you can build rich client-side interactions. A Next.js application that shows you understand server-side rendering, data fetching, and deployment. A Rails API with a React frontend that shows you can architect the boundary between two systems -- serialization, authentication, real-time updates, error handling across the network divide. That third project is the career multiplier. It is the one that gets you hired at companies running Rails backends with React frontends, which is half the senior Rails jobs on the market right now.
The Rails-plus-React integration project is where an integrator's mindset pays off. Most candidates can build a Rails API or a React frontend. Few can build both and make them work together elegantly. Fewer still can explain the architectural tradeoffs: why you chose REST over GraphQL, how you handle authentication tokens, where you put the caching layer, what happens when the API is slow. These are staff-level questions, and having a project that demonstrates the answers is worth more than any study guide.
Here is what I want to be direct about: the JavaScript curriculum is long, and the temptation to study forever and ship nothing is real. If you are a learner by nature, you know this pattern: learning feels productive even when it is not. The capstone exists to force the transition from learning to building. Set a deadline. Ship something imperfect. Iterate in public. That is how senior engineers operate: not by knowing everything, but by shipping with confidence despite knowing they don't.
An open-source track record -- a package people download, star, and depend on -- proves you can ship and maintain software. Do the same thing here. Ship it, maintain it, let it prove your range.