Axiom Math's Axplorer solves complex mathematical problems in hours on single machines, marking a shift from AI as research assistant to autonomous problem solver.1 The tool addresses mathematician frustration with DeepMind's AlphaEvolve, which requires special access and manual problem submission.1
OpenAI is building AI systems capable of working indefinitely in a coherent way, functioning as automated research interns.2 Chief Scientist Jakub Pachocki expects these systems won't match human intelligence across all dimensions by 2028, but emphasizes they don't need to. "You don't need to be as smart as people in all their ways in order to be very transformative," he said.2
The technology targets low-hanging fruit in research. "There are tons of problems that are open because nobody looked at them, and it's easy to find a few gems you can solve," said François Charton, highlighting opportunities in neglected research areas.1
Automated research systems promise knock-on effects across technology development and financial markets as discovery timelines compress. Drug development, materials science, and algorithm optimization could accelerate as AI handles iterative testing and hypothesis generation that currently consumes researcher time.
Current limitations remain in truly novel ideation. The systems excel at solving defined problems and exploring established problem spaces, but generating breakthrough research questions still requires human insight. This positions AI as workflow automation rather than complete researcher replacement in the near term.
The shift raises policy questions governments must address, Pachocki noted, particularly around research access, intellectual property, and the changing role of human researchers.2 As single-machine solutions like Axplorer democratize access to advanced mathematical tools, the barriers between cutting-edge research capabilities and everyday practitioners continue falling.
The technology's commercial trajectory suggests a two-tier market: open tools like Axplorer competing against closed systems requiring institutional access. This mirrors broader AI commercialization patterns where deployment speed and accessibility increasingly determine market winners.
Sources:
1 François Charton, MIT Technology Review, March 25, 2026
2 Jakub Pachocki, MIT Technology Review, March 20, 2026