Rachel Webb argues LLMs will drastically change how she executes math research but not her metric for mathematical interest or her two humanistic reasons for doing math.

The author separates the approach to mathematical research from its execution: LLMs do not change her approach, but they will drastically affect how she executes it. She locates AI's impact in how research is done rather than why it is done or what counts as interesting, which differs from the common framing of whether AI replaces mathematicians. This is a self-reflective argument based on the author's own research experience, grounded in her first-person practice and judgment, with no experiments, samples, or statistics.

The author says the concrete mechanism of acceleration is that the machine has read all the books and knows how many standard lemmas go, which turns some lands of mathematical fantasy into worlds she can realistically start exploring. She uses an operational metaphor to state both benefit and tradeoff: taking the interstate instead of the side roads has tradeoffs, but for any given leg of the journey she can choose which route to take. This is personal experience and analogy; the text gives no specific cases, time savings, or counts of lemmas.

The author argues the metric for mathematical interest does not change with LLMs: even if a high-profile problem has a solution, the shards of understanding and the surrounding terrain remain interesting and valuable. She reframes a solved problem as terrain still worth mapping, responding to the worry that AI might take the interest out of mathematics. The argument uses Millennium-type high-profile problems as a reference point; it is an opinion-based argument without specific problems or results as evidence.

The author gives two humanistic reasons for humans to keep doing math: math is interesting to us individually, and math's shared interest creates communities such as student-teacher relationships, collaborations, conferences, colloquia, and tea-times. She frames AI risks as possibilities rather than certainties: it could tempt us toward knowing the answer over understanding the solution, and it could weaken social bonds by making scooping fears more common or making it easier to ask a machine than a colleague. The support is a personal life choice, walking seven miles round trip to church weekly, and a quotation from Wendell Berry; this is value-based argument rather than empirical research.

This article is aimed at mathematicians and people interested in how AI changes scholarly practice, and it applies at the individual level of choosing where to use LLMs to accelerate execution and where to insist on personal understanding. The author explicitly sets aside economic reasons and offers only two humanistic reasons, so the scope of its conclusions is personal motivation and community life rather than economics or policy.

Readers still need to judge for themselves how large the execution acceleration is in specific mathematical areas; whether the author's rule of using AI only in ways that make research more enjoyable and yield understanding is sustainable in a competitive research environment; and how far the worry about weakened community bonds will materialize. In addition, this is an incomplete reading scope, and the original was converted from another format using AI, so any figures or sections not included are outside this summary.