OpenAI Erdős resolution sparks AI goalpost debate

The resolution of the planar unit distance conjecture by OpenAI's internal reasoning model signals a tipping point where AI is now expanding the boundaries of human knowledge in the hard sciences.

• –OpenAI's model disproved Erdős problem 90 by discovering a new family of point arrangements using algebraic number theory, a method humans had overlooked for 80 years.
• –Google DeepMind concurrently resolved nine other Erdős conjectures using its Aletheia agent, proving that frontier models can now execute complex, multi-step formal proofs at scale.
• –The viral "moving goalposts" sentiment highlights the psychological shift where once-impossible tasks are dismissed as trivial the moment they are mastered by AI.
• –These breakthroughs were achieved by general-purpose reasoning models without task-specific training, demonstrating the arrival of commoditized high-level abstraction.
• –For developers, the integration of these capabilities into unified agentic experiences promises a new era of autonomous technical problem-solving and formal verification.