Genuine question. Why does almost every AI training setup still feel extremely engineer-focused? Most tools I’ve tried expect people to already understand things like: CUDA VRAM LoRA settings Docker dependency issues quantization optimizers terminal co…

Hot take: A lot of what’s being called “AI engineering” right now feels like: prompt tweaking chaining APIs adding retries/guardrails Not actually building models or understanding them deeply. Don’t get me wrong—there’s real skill in making these syste…