We are entering a period where the initial “shock” of AI capabilities is fading, replaced by a ruthless reality: The learning period is well underway.

What do we do with a boundless new toolset? In 2026, the question isn’t “Can AI do this?” its: Who do you want to learn with?

Public “Oh 💩” Moments

The discourse on Twitter/X right now is a perfect barometer. You can see it in threads like this one, where the realization is the tools aren’t just toys; they are replacing the fundamental loops of software creation.

The follow-up replies indicate a lingering uncertainty and hesitation that must be confronted not with fear, but with measurement. While the industry processes that shock, the reality for the Code Shepherd is already here. Various data points support this shift, e.g.: GitClear’s 2025 AI Code Quality Research reveals that code duplication has risen 8-fold, and for the first time in history, the volume of copy-pasted code has exceeded refactored code. The “volume” is up, but the “shepherding” is the new bottleneck. This is not the same as reviews. Reviews are commodity within a quarter or 2. Evals are likely to be commoditized next.

For the unprepared, this could be a crisis. For the Code Shepherd, this is the baseline.

Continue reading “The “CEO Archetype” is the new 10x” →