Coder reflects on whether AI would have helped—or hurt—his 100-day algorithm challenge

The article captures a moment of reflective tension for software engineers in the age of AI assistance. The author completed his 100-day algorithm challenge eight years ago—a time when generative AI was not available as a learning aid. The project required genuine effort and patience: sustained, after-work study sessions where implementation failures forced deep learning about data structures and algorithms that formal CS education had glossed over. By the time he reviewed his work, GPT-5.6 Sol was able to diagnose systematic flaws across the entire codebase in minutes, identifying not just obvious bugs but patterns of incomplete work.

This asymmetry—that an AI could now do in minutes what took him years, while his hand-crafted code still contains significant bugs—raises a genuine pedagogical question the author does not resolve. He notes that AI feedback as a learning mechanism could have been valuable: asking "Am I being Pythonic? Is this correct?" during the work, rather than after. Yet he also expresses wariness about whether such immediate, omniscient feedback would have substituted for the struggle that actually teaches. The author's decision to preserve the flawed code rather than retrofit fixes suggests he values the historical record of his learning process more than a polished final product.

The article's most grounded insight is personal rather than prescriptive: the author is uncertain whether he would have the same drive to undertake a 100-day challenge today. Whether that hesitation stems from age, changing priorities, or the psychological effect of knowing that AI could evaluate his work remains unexamined. What emerges is not a verdict on whether AI ruins or enhances learning, but a honest acknowledgment that the presence of such tools changes the calculus of effort, feedback, and what counts as genuine understanding.