Meet Balaj Saleem - Software Engineer at Amazon working on AI adoption at IMDb, with experience from both startups and big tech, in a conversation about AI-Assisted Enterprise Coding.
Highlights from the conversation:
The 80/20 rule for AI coding - spend 80% of time planning and providing context to the LLM, only 20% implementing. The cost of fixing bad code later far outweighs upfront planning.
Repository-level markdown files are game-changers - create template files with business context, system assumptions, and coding guidelines that LLMs can reference for every prompt in that codebase.
Question-answer sessions before coding - have 15-20 critical questions with the AI to understand the problem fully before writing any code. Often you're not 100% clear on what you want to build.
LLMs excel at writing tests - they're incredibly good at test generation for existing code, making this a low-risk, high-value starting point for AI adoption.
Enjoy!