The 2025 AI Engineer Reading List

Source: https://www.latent.space/p/2025-ai-engineer-reading-list Author: swyx & Alessio (Latent.Space) Date: 2025-01-14

Summary

Curated reading list for AI engineers entering 2025. Organized by topic: foundational ML, LLM internals, agent systems, evals, infrastructure. Valuable as a map of what the AI engineering community considered essential knowledge entering 2025.

Key Claims

  • Foundational layer: transformer architecture, attention, positional encoding, scaling laws — still essential regardless of product focus.
  • Agent layer: ReAct, tool use, function calling, agent evaluation — the new essential stack.
  • Infrastructure layer: KV cache, speculative decoding, vLLM internals — becoming necessary as costs matter.
  • The “AI Engineer” job: exists at the intersection of ML knowledge and software engineering — neither pure ML researcher nor traditional SWE.
  • Missing from most curricula: evals, prompt engineering rigor, agent failure modes — these are where production gaps are.

Entities

  • swyx (Shawn Wang), Alessio Fanelli — Latent.Space co-hosts

Concepts