Al Engineering 101 with Chip Huyen (Nvidia, Stanford, Netflix)

Chip Huyen, author of the book AI Engineering, argues that teams building AI applications spend effort on the wrong things. On Lenny's Podcast she discussed a table she posted contrasting what people expect to improve an AI application with what she says does.

The expected fixes on her list are following the latest AI news, adopting the newest agent framework, agonising over vector databases, constantly comparing which model is smartest, and fine-tuning. What she says works is talking to users, building more reliable platforms, preparing better data, optimizing end-to-end workflows and writing better prompts. Her reasoning on technology choices is practical: she asks teams how much improvement an optimal choice would bring over an adequate one, and how hard an untested technology would be to switch away from later. If the gain is small and switching is costly, she says, the debate is not worth the time.

This is practitioner advice from her work with companies, not a controlled study.