VIII. What Remains Unsolved
Intellectual honesty requires acknowledging what AI collaboration does not change.
Tacit Knowledge Remains Tacit
If a programmer cannot articulate why a solution “feels right,” AI cannot extract that knowledge. The Stradivarius problem persists. The Quality Without a Name remains unnamed.
AI can ask questions that prompt articulation. It cannot reach into the mind and pull out what resists words.
Judgment Cannot Be Automated
Knowing when to apply a pattern, when an exception is warranted, when apparent similarity is misleading—these remain human capacities that AI can support but not replace.
The AI can surface candidates: “This looks similar to pattern X.” Only the human can judge whether the similarity is deep or superficial, whether the pattern applies or misleads.
Theory Is Personal
Each programmer builds their own theory. Even with perfect documentation, new programmers must do the cognitive work of theory construction.
AI can provide more and better material for this construction but cannot perform it on the programmer’s behalf. Reading is not understanding. Being told is not knowing.
The Reader Must Meet the Writer
Theory-focused documentation still requires a reader willing and able to engage deeply.
If readers treat documentation as reference material to be consulted rather than as theory to be acquired, no amount of documentation quality will help. The transmission requires receptivity.
The best documentation is still only material for theory reconstruction. The reconstruction itself happens in minds, through effort, over time.