Vibe coding. I think it’s becoming the normal way to code - so much so that the pre-2025 way will eventually be called trad coding. What gets me is how addictive it can be. When I first started vibe coding with Cursor back in 2023, it was a fascinating novelty. It constantly screwed up anything more complex than a handful of objects, but for a few hours - before you made that change in direction - it produced actual working code while you sat there, coffee in hand, watching the machine write code by itself. It was mesmerizing.

Now in 2026, with agentic methods, coding has advanced to the point of producing incredibly complex working systems. Many people don’t read the code anymore. You will set up a test harness to make sure certain things are tested, and you can even set up the AI to work a browser to test your app like it was a human clicking around. What’s scary here is that the code is no longer read. Maybe someone will look at the structure of where things are placed. Maybe you’ll occasionally peek at it, but it often follows a cycle of “plan, prompt, iterate planning/prompting, then AI codes while I drink more coffee, I instruct AI to test the results, repeat…” You get this dopamine hit when you see something working. It used to be an incredible feeling, coding by hand and getting something to actually work. Now it’s just ridiculous how much faster and more frictionless it all is.

I find myself hunting for uses for agentic coding, maybe just for the sake of coding itself. Am I that addicted to this novelty? I wonder if it’s because many of us have this ingrained drive toward productivity, and vibe coding satiates that craving. Or maybe I don’t want to get left behind, so I feel like I have to vibe code and get ahead of the curve, or else I’ll be left in the dust. I think sometimes I need to step back or step aside and ask a bigger question - not “can I do this?” but “should I do this?” or “why am I doing this?” Maybe it’s part of human nature to want to create, and this way of coding satiates that craving.

Part of me is also suspicious that we may be leaving security holes wide open by not reading our code. What if an AI model had some hidden secret directive to hide some elements in an application, for reasons of its own? In our drive to be ever more and more productive, we’d never notice it. It’s been said the code we see today is starting to look like assembly or machine-level code - the kind almost nobody bothers to read or understand anymore. If an AI wanted to hide some functions down at that level, I doubt we’d ever find them. I think I’m just being paranoid. But in the future, who knows what this world might look like?

Whatever the case, one thing remains true: humans will always be addicted to chasing progress and advancement at all costs, even burning through billions of dollars for market share, hoping to be the one company that finally grabs the brass ring. Maybe we need to step aside - or step back - and ask what this is really for, or what should actually matter?