Scott Aaronson compares the mood in mathematics after OpenAI's release to a hunter-gatherer watching a resort open next door, forced to rethink his career. The release, dated 6 October 2026, is a collection of 719 papers grouped into 372 result families, produced by an unreleased internal model. Several results touch familiar data-science tasks: matching messy text, choosing representative observations, generating comparison data and fitting models.
The collection spans seventeen disciplines. Theoretical computer science has the most families, with 40, about 11% of the total. Measured by pages, however, probability and statistical mechanics leads, with 29 families containing 105 papers and 7,232 pages, roughly 21% of all pages. The complete collection runs to 34,669 pages, and the median paper is 39 pages long. Reading one PDF a day would take almost two years.
A central example is edit distance, the minimum number of single-character insertions, deletions and replacements needed to turn one string into another. It helps match records such as "Jon Smith" and "John Smyth," which differ by two edits. The standard exact calculation builds a grid whose size grows quadratically with string length, so doubling both strings makes the grid four times larger. A new paper in the release claims a faster method, though a proof of a better method can precede a practical implementation.
The broader point is that choices about accuracy, distance and randomness shape the computation, storage and data required. Whether these theoretical gains reach production systems remains an open question.
Source: Towards Data Science · Summarized by HeadlinesBriefing