Researcher Solves 6 Erdős Problems in 5 Days Using GPT-5.6 Sol
Researcher Shouqiao Wang, who has a background in mathematics, reported that he used OpenAI's GPT-5.6 Sol model and a Codex workflow to successfully solve 6 open Erdős mathematical problems in just 5 days. The event has attracted widespread attention as strong evidence of advanced mathematical reasoning capabilities in large language models.
Confirmed
According to the original post, Shouqiao Wang has a mathematical background, but he emphasized that the Codex workflow used does not actually require deep mathematical expertise from the user. Out of his total attempts, he solved 6 difficult problems and plans to share more details later. Furthermore, as relayed by Dan Robinson, OpenAI sponsored the user's exploration to observe how the model performs when pushed to its limits.
Why it matters
Erdős problems are notoriously difficult, making the achievement of solving 6 of them in 5 days highly impressive. This not only demonstrates the robust capability of GPT-5.6 Sol in complex mathematical deduction but also suggests that through specific Codex workflows or prompt engineering, even non-top mathematicians might leverage AI to conquer academic challenges. As Dan Robinson pointed out, the true differentiator might simply be prompt engineering and workflow optimization.
2026-07-23 ~ 2026-07-24 · 6 related posts
- Episode 1: GPT-5.6 Solves Long-Standing Math Problems(2026-07-15, 4 posts)
- Episode 2: Researcher Solves 6 Erdős Problems in 5 Days Using GPT-5.6 Sol(2026-07-23, 6 posts)
Primary sources
- [source] OpenAI GPT-5.6 Sol helped solve 6 open Erdős problems in 5 days — Charuru · 2026-07-23
- A researcher says GPT-5.6 Sol helped crack 6 Erdős problems in 5 days — rohanpaul_ai · 2026-07-23
- Prompting plus Codex workflow helps solve six Erdős problems in a week — schwentker · 2026-07-24
- Polymarket Says a Math PhD Solved Six Erdős Problems in Five Days With GPT-5.6 Sol — Polymarket · 2026-07-24
2 near-duplicate retellings: arampell · QuanquanGu