antirez argues programming is fundamentally shifting from code review to controlling the ideas behind software. As a Redis creator rejoining the project and building Dwarf Star for local LLM inference, he observes three reasons: LLMs generate too much code to review daily (5k lines); they excel at local optimization but struggle with big-picture design; and 8 hours spent reading code trades off against crucial design thinking and QA.

He references Mythical Man Month's wisdom on conceptual integrity, noting pre-AI software was already "rotten." AI helps navigate complex domains like inference engines where subtle errors accumulate. When Matteo Collina questioned his AI code review for Redis, antirez admitted it's increasingly pointless post-GPT 5.5 and GPT 5.6 Sol—AI catches more errors than humans. Instead, he'd invest time in DESIGN.md files capturing data structures, implementation tricks, and mental models.

For most projects, reviewing code no longer makes sense. Focus on quality, testing, and the software vision. Young programmers should build fundamentals (interpreters, databases) rather than review LLM output for routine tasks.