Discussion
working on the water in block game is the perfect example of why agentic coding can't actually work
i could've just gone full steam ahead without thinking and started banging out hundreds maybe even a couple thousand lines of code to get the feature in and even get the water looking good
but instead i spent the day thinking about how to best approach it, slot it into existing structures in a way with minimal extra code, complexity, and performance overhead
agentic AI simply *cannot* do this
this is also, by the way, why coders who are forced to crunch write terrible code. trying to cram too many features into too little runway leads to that "how do i best approach this" period getting snuffed out so you just start banging on your keyboard and yeah you get your features and you get them fast but the code and performance suffer as a result
ask me how i know
@eniko I think "we", as in the software engineering community, have been crunching far too much for far too long. There is way too much technical debt across the whole stack, because everyone is pushing or being pushed for features over tooling and code cleanup.
And that's why agentic coding is so appealing to many, because it temporarily papers over this. Let the agent deal with the mess. Except of course, it'll just end up making an even bigger mess and making everything worse. But it works... for a while... and people fall for it.
anyone who thinks it can is a fool being taken for a ride by their own brain's reward centers because oh how alluring the prospect of getting something for nothing
but there ain't no such thing as a free lunch