Columbia PhD student Shouqiao Wang says he solved six open Erdős problems in five days with GPT-5.6 Sol in Codex, and publishes his workflow
Awaiting review
The takeaway
On 22 July 2026 Shouqiao Wang, a Columbia PhD candidate, posted “I solved 6 open Erdős problems in 5 days, using @OpenAI GPT-5.6 Sol”, with a thread giving his Codex workflow and prompts.
Status
- Claim
Awaiting review
- Our reporting
- Medium confidence
- Verification
- Expert-checked
- Importance
- 3 of 5
- Last verified
- 19 October 2026
Your AI and this story
- GPT-6 Astra83 days after its cutoff
- Claude Opus 5.522 days after its cutoff
- Gemini 4 Argon113 days after its cutoff
- Grok 4.752 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 22 days before it.
Key facts
- Wang on X (22 July 2026, ~1.59M views by Oct 19, fxtwitter): ‘I solved 6 open Erdős problems in 5 days, using @OpenAI GPT-5.6 Sol. I have a math background, but the Codex workflow I used does not require deep mathematical knowledge. Here’s exactly how I approached it, including my prompts’
- Per 36Kr (24 July): about 13 problems attempted, 46% success rate; one problem ran continuously for 32 hours; process of attempt, diagnosis, draft proof and adversarial audit; one solved problem was a topic Terence Tao had previously studied (no Tao statement quoted)
- Method (per coverage of his thread): pick problems mathematicians already discuss, define in advance what counts as a complete solution, list weaker results and trap cases up front, then run multi-hour searches for counterexamples before drafting proofs
- Not established: which Erdős problem numbers, whether erdosproblems.com marked them solved, and any expert review; the thread replies were not read (fxtwitter returns only the first post)
What happened
Wang posted his result and a step-by-step thread on how he drove GPT-5.6 Sol through Codex on Erdős problems. The thread drew wide attention for its claim that the workflow needs little mathematical expertise. We read only the first post (fxtwitter) and 36Kr’s summary.
Why it matters
It is an early public recipe for turning a coding agent into a problem-solving loop on open maths problems, published months before the Erdős-problem wave of September–October 2026. Its claims still need checking against erdosproblems.com.
What is disputed or not yet verified
| Verification | Unverified claim (X thread); no database update or expert check found |
|---|
Sources
2 sources from 2 sites. Numbers match the chips in the text.
2 sources: 1 press, 1 reaction
Press
- 36Kr: GPT-5.6 disproves 30-year graph theory conjecture; Peking University alumnus solves six Erdős problems in five days (24 July)eu.36kr.com, press
Reactions
Changes
- Filed. Late from the 2026-10-19 lead (Fireship video); X post read via fxtwitter