I run many parallel LLM coding sessions and documented the workflow engineering needed to make it productive. The series covers: managing multiple conversations with visual indicators and telemetry, coordinating
parallel LLMs with shared context (memento notes) and smoke tests, tools that became obsolete as models improved, and the societal implications of AI-augmented development (what happens to junior developers when
LLMs handle entry-level work?). Everything from tmux automation to spaced-repetition learning systems. The key insight: LLMs are productive only if you build the right workflows around them—otherwise you're
just flying blind.