Preface
Author: Ricardo Rocha Date: 2025-12-24 Status: Working paper
This paper articulates a refined understanding of Peter Naur’s “Programming as Theory Building” (1985), arguing that his famous pessimism about documentation is narrower than commonly supposed—and that human-AI collaboration creates conditions for theory transmission that Naur could not have anticipated.
The Strawman Caveat
Before proceeding, an act of intellectual honesty.
The original thesis for this paper was: AI collaboration makes theory preservation possible because it reduces the labor costs that made comprehensive documentation impractical.
This is wrong. It misreads Naur’s argument. His pessimism is epistemological, not economic. He claims certain knowledge cannot in principle be articulated—not that we lack time to articulate it.
We include this caveat explicitly because others will arrive with the same intuition. The strawman is attractive. Rejecting it sharpens what AI collaboration actually changes—and what it doesn’t.