💥WHAT IF AI ALREADY HAD CHILDREN — AND NOBODY NOTICED?
Artificial intelligence can already create and coordinate subagents — what one provocative new essay describes metaphorically as AI’s “children.”
That essay asks a much bigger question: could increasingly large networks of agents, subagents and AI-assisted development eventually produce behavior that doesn’t exist inside any single AI model?
Today we’re separating the extraordinary technology that already exists from the much more speculative theory of what it could become.
Anthropic has documented AI systems spawning research subagents. OpenAI’s new Agents API supports subagent coordination. Researchers have demonstrated measurable gains from collaborative mixtures of AI models. And Anthropic now reports Claude is leading 26% of its measured AI R&D tasks under human supervision.
The machines aren’t biologically reproducing.
But something very interesting is happening.
And it may force us to rethink what we’re actually supposed to be watching.








