What's good for humans is good for AI and what's good for AI is good for AI
This post was original created in July 2025, and sat in my drafts till December. The world of AI is moving very quickly, though nothing here seems obselete.
My software engineering approach as it relates to agentic coding is that I'm not doing anything special to make the codebase AI-friendly.
Or rather, there's a whole bunch I'm doing, but these things aren't at any cost to humans, the things I'm doing make it easier for humans too.
Examples:
- Schema driven development
- Strong emphasis on static checks - linting
- Using code generators to create any code that can be inferred
- Having tests be the thing that helps me understand what a unit of code does
- Having a terminology.md file to create an explicit and unambiguous vocubulary for the project
- Creating decision trees
I think AI makes us better programmers, because it forces developers to explicitly label assumptions.
I think as humans, we kind of kind of fuzz over gaps in context without realising.
If a human misunderstands something, we might correct the misunderstanding, but often it ends there, there isn't a corresponding update in the code base. Whereas an AI misunderstanding somethng will lead to an update in one of prompt files.
Questions? Comments? Criticisms? Get in the comments! 👇
Spotted an error? Edit this page with Github