Introduction

AI-assisted coding has moved past simple autocomplete. Developers are now using large language models to design, scaffold, and ship entire applications. One engineer recently put that idea to the test by rebuilding a personal website from scratch in just 48 hours. The stack was Next.js 16, and the coding partner was Claude Code. The result wasn't just a fast launch—it was a case study in how to work with AI without losing control of the codebase.

What Happened

The project began with a clear boundary: rebuild an existing site using Next.js 16, but do it quickly and with AI handling much of the implementation. Over two days, the developer generated components, pages, and configuration files through Claude Code. Rather than accepting every suggestion, they applied strict constraints to keep the architecture manageable. Automated tests were written early to catch regressions, and changes were merged in small, reviewable increments instead of one massive diff. By the end of the second day, the site was live, but the author noted that the final line count told only part of the story. The real work happened in the decisions about what to build, what to leave out, and how to verify each piece before moving on.

Why This Matters

AI coding tools can produce code faster than most humans can read it, which creates a new risk: shipping software you don't fully understand. This experiment shows that speed is meaningless without guardrails. Constraints prevent scope creep, tests provide confidence when the AI writes unfamiliar code, and small merges make it easier to spot mistakes before they reach production. For teams adopting AI-assisted workflows, the lesson is that the tool should accelerate disciplined engineering, not replace it.

Key Takeaways

  • Next.js 16 and Claude Code can power a full site rebuild in roughly 48 hours.
  • Setting hard constraints keeps AI-generated code from becoming unmanageable.
  • Tests are essential when you didn't write every line yourself.
  • Small, incremental merges reduce risk and simplify review.
  • Deciding what to build matters more than how much code you generate.

Conclusion

Rebuilding a website in two days is an impressive feat, but the more valuable takeaway is the workflow behind it. AI can handle the typing, but humans still need to handle the judgment. By combining Next.js 16 with Claude Code—and wrapping the process in constraints, testing, and careful integration—developers can ship faster without sacrificing quality. Whether you're refreshing a personal portfolio or prototyping a product, the same principles apply: build less, test more, and merge often.