Scattered thoughts on how doing math feels these days (dynomight.net)

In a reflective piece, a mathematician gauges the rapid evolution of AI in mathematical contexts, noting a significant shift from 2024, where AI was seen as a rudimentary tool, to 2026, where it has become indispensable for many tasks. The author reveals that while AI has dramatically improved productivity—boosting their efficiency in "doing math" by about 50%—its ability to truly understand and create mathematical abstractions still lags behind human intellect. The transition to incorporating AI in mathematical workflows prompts questions about the evolving nature of mathematics itself, suggesting that the relationship between human mathematicians and AI is nuanced and requires careful navigation. The piece highlights several key limitations of AI, particularly its struggles with writing clarity and the ability to form coherent, modular arguments. Despite its impressive speed and technical capabilities, AI's tendency towards complex, convoluted expressions often complicates rather than simplifies the understanding of mathematical proofs. The author expresses optimism, however, about the potential for AI to augment human creativity in math, envisioning a collaborative future where AI serves as a powerful assistant rather than a replacement, emphasizing the enduring value of human insight and critical thinking in mathematical pursuits.