Problem
Code, logs, docs, books, and articles are frozen at a single written abstraction level. You can’t zoom out for overview or zoom in for details — you’re constrained to whatever resolution the author chose.
Pattern
Generative AI makes text elastic. You control the level of detail by how you ask and how you steer it. You can expand, collapse, and shift abstraction levels on demand.
Zoom Out — ask for high-level synthesis
- "Give me the high-level architecture of this codebase"
- "Summarize the main components and how they interact"
- "Make this paragraph much shorter"
Zoom In — interrogate specifics interactively
- "How does authentication work here?"
- "Show me the implementation of this flow"
- "What edge cases does this function handle?"
Examples
Exploring codebases: Opening Kafka repository: First ask for high-level architecture overview with ASCII diagrams. Once oriented, zoom into specific areas: "How does the consumer group rebalancing work?" AI researches and explains the details.
Research papers: Start with chapter-level summary, then zoom into one experiment's methodology. Ask questions when something is unclear.
Refactoring: Extremely powerful as one of the steps during refactoring:
- ask a coding agent to explain your code in English
- force to the right level of abstraction by zooming in or out and tuning the details out until English is extremely clear
- then ask AI to align code with the English
