I started coding in 2019, but I mostly stepped away after entering university. I remember working with VS Code, React, TypeScript, and the usual web-development tools. Now I'm trying to return and feel out of the loop: AI assistants seem to be built into almost every coding product, many developers recommend tools like Claude or OpenAI, and I recently learned that Create React App is no longer the recommended starting point. What has changed in web development over the past few years, and which languages, frameworks, and tools would make sense for someone getting back into it? I'd also like to understand how necessary AI assistance actually is.
4 Answers
You can get back up to speed without trying to learn the entire ecosystem at once. Build a small project with HTML, CSS, JavaScript, TypeScript, React, Git, and browser developer tools. Then add testing, a backend, and deployment as the project requires. Framework names and build tools change frequently, while fundamentals like HTTP, accessibility, state management, version control, and debugging remain valuable.
The industry is noisy right now because AI companies are promoting large language models as if every developer workflow needs them. The more balanced view is that they can save time on repetitive tasks, but they still produce incorrect, outdated, or insecure code. Learn to read and test code yourself first, then use AI selectively rather than choosing a stack solely because an influencer recommends it.
Create React App being deprecated does not mean React is obsolete. For a new project, the React documentation generally points people toward a framework such as Next.js or a modern build tool such as Vite, depending on what they are building. A practical current stack would be TypeScript, React, Vite for a client-side app, and a straightforward package manager such as npm or pnpm. If you want routing, server rendering, and full-stack features, look at Next.js after refreshing your React basics.
A lot has changed, but the AI emphasis is partly a genuine productivity trend and partly aggressive company marketing. You do not need an AI assistant to become a competent developer. Treat it like autocomplete or a second pair of eyes: useful for explanations, boilerplate, and debugging, but verify everything it produces and keep practicing the underlying concepts.
That is also a good way to avoid getting trapped in tutorial cycles. Pick a project you would genuinely finish, use the official documentation, and only add libraries when a real problem calls for them.