Over the holidays, X was on fire with people realizing AI coding tools have crossed a real line. Even Andrej Karpathy, the guy who coined “vibe coding,” said: “I’ve never felt this much behind as a programmer.”
Most of the buzz was about Anthropic’s Claude Code paired with Opus 4.5. I’ve been coding with AI tools for two years now. I’ve watched them go from glorified autocomplete to something else entirely. After a few weeks with this combination, I’ll say it plainly: coding as a common profession will be dead in one to two years.
There will be edge cases. Specialized enterprise codebases, legacy systems, and niche domains will still need humans. AI also tends to do better with brownfield work, where there’s existing code to learn from, than with greenfield builds. But for common software stacks, AI is going to dominate.
To be clear: I said coding, not software engineering. Engineering will be in more demand than ever. The center of gravity just shifts away from syntax, toward system design, architecture, and managing AI agents at scale. AI can’t replace human judgment and experience. Not yet, and probably not for a while.
The people who invest the time to master these tools now will have the advantage. They’re moving faster than cloud computing did, faster than mobile did. If you’re a developer and you’re not using them every day, that’s the gap to close first. Everything else follows.