01. Preface: This is a long topic, but easy to read
Working with existing code is a vast topic
—Part of CSS Primer for the AI Era — Working with Existing Code
Working with existing code is where AI makes
the most subtle mistakes.
So, we are going to repeat principles under different common scenarios.
Nothing to memorize, just go through them and recognize each as they are,
based on your own experience.
We are not talking about dramatic errors
or broken syntax,
not even missing semicolons.
The mistakes here are quieter:
- choosing the wrong pattern
- adding complexity where none is needed
- breaking boundaries that were invisible
- inventing features that never existed
- misreading the goal
- explaining too much
- overlooking the edge case that matters most
These mistakes are small in appearance
but large in consequence.
They are the kinds of mistakes that slip through reviews,
accumulate over time,
and slowly distort the architecture.
A simple truth:
AI rarely fails loudly in existing code.
It fails quietly
— and quietly is where systems drift.
This section is long because we must learn to
recognize these quiet shifts from many angles.
Each chapter teaches a different way the work can drift,
and a different way we bring it back.
We encounter these errors at every level of programming —
and in every form of AI‑assisted work.
So, this chapter is worth reading,
even for those who do not write code.
- The repetition is intentional.
- The variety is intentional.
- The length is intentional.
But we can read it calmely because:
We are not teaching rules.
We are teaching instincts.
And instincts require practice from multiple directions.
- Suggested Next Reading: → How AI Sees Code (and What It Cannot See)
Tony de Araujo —New York | Lisbon