wtf is happening： A researcher says he used GPT-5.6 Sol Ultra in Codex to produce proposed solutions to six （！） Erdős problems， and published the full research trail.
He says GPT-5.6 Sol Ultra in Codex produced proposed proofs for six Erdős problems.
His prompts explicitly reject special cases， circular reductions and unsupported lemmas. They preserve multiple incompatible approaches and use independent agents to attack every candidate proof.
Codex then continues without further interaction：
"attempt -> failure -> diagnosis -> new approach -> proof draft -> adversarial audit -> repair."
science is accelerating at unprecedented speed.
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 ...