"So there is a machine, right now, substituting me after 32 years of coding. It’s just doing what I’ve done—so much faster. Can I go home now?" — from my video.
The Moment I Realized AI is Coming for My Job
I’ve been coding for over thirty years, writing, refactoring, and maintaining more code than I can count. Started with BASIC on C64 and transitioned into the web world. Yet here I was, sitting in front of my computer, watching Claude Code migrate my 90,000-line Next.js 14 codebase to Next.js 15 while I just pressed Enter.
This was not just some gimmick. This was real. The AI wasn’t just suggesting code; it was executing migrations, fixing errors, even reasoning through problems. It wasn’t perfect, but damn—was it fast.
For the first time, I questioned:
If AI can do this now, where does that leave us, the engineers?
The Challenge – Why I Even Tried This
"I put Claude Code to the test—can it really migrate my 90K LOC Next.js 14 codebase to Next.js 15 without breaking everything?"
We run a B2B SaaS platform with a large frontend built in Next.js 14. Migrating to Next.js 15 meant switching from Page Router to App Router, updating use client directives, fixing TypeScript compatibility, and running tests.
Doing this manually would take hours—if not days.
It’s tedious, error-prone, and frankly, boring.
So, I turned to AI. I let Claude Code do the heavy lifting.
And the results? Well, let’s just say… I wasn’t ready.
AI vs. Developer – Who Wins?
1. Setup Was Easy—Too Easy
Installing Claude Code was straightforward. Within minutes, it was analyzing my entire codebase. It wasn’t just looking at files; it was understanding patterns, naming conventions, linting rules, and architecture.
Then came the first real test:
Me: “Update all my components and add the use client directive where necessary.”
Claude: “Sure. Here’s a batch script to automate it.”
Wait, what? Instead of manually modifying 900 files, Claude generated a script that did it in seconds. When I ran it, boom. It was done.
2. AI Didn’t Just Follow Orders—It Thought Ahead
At first, I expected Claude to simply follow instructions. But no—it started anticipating problems.
Type errors? It refactored props to remove deprecated defaults.
Import mismatches? It auto-fixed them and ran a build check.
Test failures? It debugged and reran failing Cypress tests.
It wasn’t perfect, and sometimes, it overlooked nuances or applied brute-force solutions, but it worked at breakneck speed.
3. I Became a Passive Observer
"I just pressed Enter all day long. AI iterated, debugged, and even verified its own fixes. It felt… unsettling."
The scariest moment was when I realized I was no longer coding. I was just approving changes, like a manager reviewing work rather than doing the work myself.
Was I really needed?
Key Takeaways – What I Learned
AI is no longer just an assistant; it’s an engineer-in-training. It doesn’t just suggest—it executes, iterates, and verifies.
The role of a developer is shifting. We’re moving from writing code to reviewing AI-written code.
AI won’t replace engineers, but it will replace hands-on programming. Engineers will remain valuable in architecture, decision-making, and deep problem-solving.
Trust AI, but verify. AI makes mistakes—sometimes big ones. But it also catches issues we might overlook. (Hint: Same as a human engineer)
Test coverage is more critical than ever. If you let AI refactor a large codebase, tests are your safety net.
The Future of Software Development
We’re at the start of something huge. AI isn’t replacing us yet but fundamentally changing how we work.
If you’re still skeptical, I get it. But I’ll leave you with this:
If AI can migrate entire Next.js apps in minutes, what will it do in 3 years?
Let me know in the comments: Would you trust AI to refactor your production code?
Next Video: Migration Page to App Router in Next.js 15
I will upload the second part of the video in my paid section. It’s about an hour of uncut material with more information than the free one.
YouTube upload:
Share this post