OpenAI is facing industry-wide ridicule after executives falsely claimed GPT-5 had solved previously unsolved mathematical problems. The controversy erupted when VP Kevin Weil's now-deleted tweet claiming breakthrough discoveries was exposed as a fundamental misunderstanding of what constitutes an 'unsolved' problem, prompting harsh criticism from competitors and mathematicians alike.
OpenAI just handed its competitors the perfect ammunition. The company's latest embarrassment started when VP Kevin Weil posted what he thought was a victory lap on social media, claiming GPT-5 had "found solutions to 10 (!) previously unsolved Erdős problems and made progress on 11 others."
The post, which Weil has since deleted, sent ripples through the AI community - but not the kind OpenAI was hoping for. Meta's Chief AI Scientist Yann LeCun quickly pounced, describing the situation as being "hoisted by their own GPTards," while Google DeepMind CEO Demis Hassabis simply called it "embarrassing."
The reality behind the claims is far more mundane than revolutionary. Mathematician Thomas Bloom, who maintains the authoritative Erdős Problems website, explained that the problems weren't actually "unsolved" in the traditional sense. When Bloom lists a problem as "open" on his site, it simply means "I personally am unaware of a paper which solves it" - not that the mathematical community considers it an active research frontier.
"GPT-5 found references, which solved these problems, that I personally was unaware of," Bloom clarified, delivering what amounts to a devastating correction to OpenAI's narrative. In other words, the AI didn't crack any mathematical mysteries - it just did some literature searches that turned up papers Bloom hadn't catalogued yet.
The situation became more awkward when OpenAI researcher Sebastien Bubeck tried to salvage the claims. After acknowledging that "only solutions in the literature were found," he attempted to spin this as still noteworthy, arguing "I know how hard it is to search the literature." The defense fell flat with critics who noted that sophisticated literature search is hardly the breakthrough capability OpenAI has been promising with its next-generation models.
This incident reflects a broader pattern in the AI industry where companies race to announce capabilities that sound more impressive than they actually are. The Erdős problems, named after prolific mathematician Paul Erdős, represent genuine mathematical challenges - making the false claims particularly damaging to scientific credibility.
