Ep82: AI and Code Generation + Code Reviews in a Changing Landscape

Recap of my week, with several different topics where I expanded my knowledge or surfaced some recurring patterns in my efforts. Highlights: • Repeated prompt iteration can cause LLMs to get stuck in unhelpful loops or produce slightly different code on reruns. Restoring to a checkpoint helps reset context and regenerate cleaner output. • A personal AI workflow shortcut: using chat models to brain dump and organize ideas into structure before writing. • Pet projects. Alongside expanding a Goodreads project (https://github.com/JMHReif/springai-g...) , there's a potential new personal improvement app idea • Other tasks: session review for NODES conference (https://neo4j.com/nodes/) , upcoming refresh work on a Knowledge Graph e-book (https://dev.neo4j.com/jmhreif-kg-ebook) , and plans for short-form video content. • Article: “Read Less, Steer More (https://blog.ezyang.com/2026/03/read-...) ” by Ezyang on mentoring code reading, using LLMs to explain code, avoiding auto-accept edits, and writing code by hand when learning fundamentals.