I still remember the day ChatGPT launched.
I was using GPT-3.5 to write a bunch of SEO articles, and one of them is still the best-performing SEO article at the company I worked for.
Back then, the critics said the language had "no soul." But it worked. Better than most human writers for that specific job.
Fast forward to today: I’m building startups using Cursor and agentic workflows. And the criticism is exactly the same.
- "Vibe coding is shit."
- "It’s insecure."
- "It’s just for experienced devs."
- "AI is just autocomplete."
But here’s the reality:
In 2026, there’s no such thing as "vibe coding" anymore. It’s just coding.
The only people still fighting it are the tech dinosaurs who refuse to adapt. If you're building in public and using natural language to ship software, you aren't "vibing"—you're just working with the most powerful tool ever created.
Stop fighting the future and start building it. 🚀
Top comments (1)
I like the pushback on the term, the risky part is when “vibe coding” becomes “skip the feedback loops.” In my experience, the moment you bring in a lightweight harness (tests for critical paths, linting, a couple of invariants) the AI output gets dramatically easier to trust and iterate on. Also +1 on doing small diffs, it keeps code review and rollback sane. Curious if you have a go-to checklist for the first hour of hardening after a prototype starts getting real users?