Excellnt point on needing specialists as gatekeepers. The issue with AI replicating destructive patterns is understated in most converstions. At my last project we noticed the codebase quality dropping after 3 months of heavy AI use because nobody caught the repeated anti-patterns accumulating across pull requests.
I'm not saying models won't improve, but the progress will be gradual. Just like with all tools before it, we’re approaching the end of the S-curve for LLMs.
Excellnt point on needing specialists as gatekeepers. The issue with AI replicating destructive patterns is understated in most converstions. At my last project we noticed the codebase quality dropping after 3 months of heavy AI use because nobody caught the repeated anti-patterns accumulating across pull requests.
"AI Slop" isn't limited to just pictures and video!
"My post assumes that AI won't improve much beyond its current state, which seems like a safe prediction"
How on earth is that a safe prediction? I would rate it as the least safe prediction.
I'm not saying models won't improve, but the progress will be gradual. Just like with all tools before it, we’re approaching the end of the S-curve for LLMs.
It feels like the optimal number of team members is trending towards something less than 6. Everyone has so much leverage now due to new tools
"1" feels like an incorrect answer though