The Compiler in the Room

Somewhere between 2021 and 2026, the fundamental unit of software work quietly changed. It stopped being the line of code a person typed and became the proposal — a block generated by a statistical model, handed to an engineer who didn't write it and often couldn't immediately say whether it was right. This book argues that AI coding assistants performed the same act of absorption that compilers did in the 1950s, taking over a layer of human labor so completely that no one misses it. But with one decisive difference: a compiler's output is deterministic. An assistant's output is a plausible guess — the first mainstream abstraction that could be confidently wrong. That single property, traced mechanism by mechanism, explains the benchmark arms race, the security anxieties, the collapse of Stack Overflow's answer economy, the reorganization of code review, and the still-unresolved argument over whether any of it actually made teams faster. Built entirely from primary sources — vendor announcements, peer-reviewed studies, court filings, statutory text — this is a rigorous, evidence-first account of the biggest change to software engineering in a generation, one that names its uncertainties instead of papering over them. Read the record. Judge the disruption for yourself.