tl;dr: pgrust v0.2 claims 300x faster analytics than Postgres (beating ClickHouse on ClickBench) by rewriting the query engine to overcome Postgres's row-at-a-time Volcano model. The post demonstrates three key optimizations on a SUM query: batching (processing 1024 rows at a time with stack-allocated buffers), operator fusion (combining scan and aggregate into one node to eliminate copies), and SIMD (parallel float operations that compilers won't auto-generate due to associativity). Combined, these took a toy query engine from 1.3s to 135ms—10x faster—with JIT compilation teased for a follow-up.