# Exploring a question in knot theory

- Researcher: Robert Lipshitz
- Role: Mathematician, University of Oregon (paper affiliation)
- Field: mathematics
- Tools: Claude Code (Fable 5, Opus 5, Opus 5.5, Sonnet subagents)
- Published: 2026-10-27
- Source: https://arxiv.org/html/2610.37595v1
- Written with the current flagship models (Claude Opus 5.5 and GPT-6.1 Sol) available.

## What they did

Lipshitz chose an algebraic problem about Khovanov local equivalence groups that he thought Claude had a chance of solving and that he could check line by line. He reports that Claude did much of the exploratory research. He began with Fable 5, then used Opus 5 and later Opus 5.5. Some sessions assigned coding work to Sonnet. The project kept notebooks recording progress and failed approaches.

Lipshitz says he wrote and checked the proofs by hand. He marks some computations and additional remarks as Claude’s reports and says he has left those unchecked. The paper leaves open whether the extra algebraic structure can be witnessed by knots.

## What they learned

He estimates 50 to 100 hours of his own active work on rewriting. That work changed the mathematics. When one proof struck him as surprisingly indirect, he asked Claude to look for another route. The paper includes the resulting proof.

In a [28 October update to his July Claude Code guide](https://pages.uoregon.edu/lipshitz/projects/claude/), he revised his earlier assessment that he had found these tools unhelpful for proving theorems and discovering new mathematics.

## Limits of this account

This is Lipshitz’s account in a preprint. I read the introduction, the open questions, and Section 5 on methodology and ethics. The technical proofs and computations have had no independent check for this collection. The project used several models over time, and the account gives no controlled comparison between them.
