Most of my agentic pilled colleagues get lapped by people using llms as a search tool/ conservatively.

A lot of agentic use creates a feeling of psuedo productivity.

Ive only met two devs who genuinely have supercharged their output with them, but they were also superstars before CC.

The agentic only folks are twiddling their thumbs all day messing with their 20 different meta-agentic tools, thinking that extra thing is finally going to give them those superpowers too.

I think being a "super contributor" has little to do with just having access to llms, but people will downvote this.

▲

satvikpendem 3 hours ago | parent [-]

I used to see the 20 meta agentic tool thumb twiddling last year such as with Ralph loops and plan modes, but since this year with frontier models, they just...do what you tell them, so I've seen that a lot less today. And thus the people use Claude Code are lapping manual coders, at least from what I've seen.

	▲	seanw444 3 hours ago \| parent \| next [-]
		Seeing the kinds of stuff my coworker is spitting out into merge requests on my codebase via Claude Code is sometimes baffling. It's literally spaghetti code. I won't be surprised if the advent of LLMs causes power usage to surge globally, not only because they themselves use a lot of power, but because the code they spew out is horrendously ugly and inefficient. And then the time they take to sort out the spaghetti code they spat out, to add new features on top, takes even longer and uses more power than before. It's gonna be a mess. I don't see how we can keep going down this road without huge investments into nuclear power collectively.
	▲	lifeisloving 2 hours ago \| parent \| prev [-]
		It does not just do what you tell it, the last six hours of my day are a fckn testament to that fact.