My current thoughts on agentic coding

I find GitHub Copilot very useful for asking questions. It's a lot more efficient than Googling. I know a lot of people are hyped about agentic programming, but I have yet to make the leap for the following reasons.

I work in a front-end React app. Let's say I need to implement a new feature. The first step is reasoning. This happens in my brain. Billions of neurons process over the years accumulated knowledge about architecture, past code bases, human preference, code longevity and probably a whole lot other things. This is amazingly efficient. At least I think so. Maybe someone can calculate my calories spent on brain activity vs. data center electricity. Now, someone might have Clod Code do this, but I’m not sure why. As I understand it, in order for agents to do this well, I am to first set up guardrails for agents in my code base to do this. Seems like a lot of work in order to not have to work. A bit of a paradox.

After I've come up with a plan, implementing the feature is really not a big deal. Boilerplate? That's been solved 20+ years ago with the advent of IDE's.

Occasionally I need help, not so much with React, but perhaps I need to change some groovy code on a Jenkins file. That's where a chat bot can be super helpful. Any free model is fine.

Have you fundamentally changed the way you write software with agents? Are you accomplishing more?

PS. My whole career, I've been a slow adapter. This is quite beneficial in he software industry, as you avoid a lot of hype, while others get burned.