Mathematics as a Non-Renewable Resource

Terence Tao has noted that the collection of open mathematical problems is being treated as a non-renewable resource. It is currently being mined at an accelerating pace by AI-powered efforts. This marks a transition away from the traditional, episodic model of human discovery toward a landscape of continuous, algorithmic extraction. When a research thread is publicized, AI-assisted researchers frequently attempt to resolve it before the original investigator can advance their project. This effectively converts open academic inquiry into a race against the machine.

For researchers, this creates an incentive to withhold promising directions from the public domain to prevent others from claiming the resolution prematurely. The risk is that the most fruitful, unsolved problems will be pulled from public view, stalling the collaborative ecosystem that relies on shared momentum. Human research produces tools, metaphors, and new mathematical language. Automated harvesting produces only the final state of a solved theorem, leaving the surrounding mathematical landscape uncultivated.

Consider a conjecture that might take a mathematician a decade to approach by building a new bridge between algebraic geometry and number theory. A human researcher would present preliminary findings in a seminar, inviting critique and co-development. An AI-powered effort has no interest in that process. It focuses on the formal constraints, treating the problem as a target to be hit rather than a structure to be understood. The result is achieved, but the ten years of intellectual growth that would have been stimulated by that problem are lost. The output is a finished proof, yet it lacks the connective tissue—the lemmas, the wrong turns, and the intuitive breakthroughs—that actually advance the field. We are left with the answer, but the underlying work that makes mathematics a generative discipline remains invisible and unshared. It is still unknown how much of our historical progress relied on this specific, slower human process of iteration.