Modern Document Understanding & IXP — Teaching Robots How to Read 2/2

Closing session of Teaching Robots How to Read — moves from Classic DU into the newer Modern Document Understanding and IXP. Where part 1 was grounded in trained models per document type, this one explores the LLM-backed extraction shape: zero-shot prompts, less training upfront, different failure modes to plan for.

Topics covered: what Modern DU and IXP actually do under the hood, the practical differences from Classic DU (cost, latency, accuracy ceiling, tuning surface), where to reach for which approach, and how to mix them in a single pipeline so each component runs where it's strongest.

Best watched after part 1 (AI Center & Classic Document Understanding) for the full IDP arc.

Date is a placeholder — replace with the real session date.