AI Surpasses Human Math Skills but Still Lags in Real-World Tasks

Artificial intelligence has surpassed human capabilities in mathematics, according to Capriole Investments founder Charles Edwards. He describes AI as the new apex predator, a shift he believes has gone largely unnoticed.

His observation follows a month where AI achieved scores matching top students at the world's most challenging math competition. However, the same AI systems continue to struggle with routine professional tasks.

Did AI Just Cross the Line in Mathematics?

Edwards, who serves on BeInCrypto's Market Intelligence Expert Council, shared his views on X on October 21.

“This is a evolutionary turning point. The last month has been perhaps the first time AI has surpassed the best human intelligence in certain fields, like mathematics. Human intelligence is now no longer the apex predator.”

At the 2026 International Mathematical Olympiad (IMO), models developed by Huawei and Xiaohongshu achieved a perfect score of 42 out of 42.

The official IMO results show that only seven out of 666 human participants matched that feat.

In early October, OpenAI provided an unreleased model with approximately 4,000 unsolved problems. The model generated 722 research manuscripts organized into 372 families of findings.

Stefano Gogioso, an Oxford lecturer on BeInCrypto's Future Tech and AI Experts Council, offered a blunt assessment.

Mathematical rigour is now free.

AI developers are even more emphatic. OpenAI CEO Sam Altman stated in July that the world is “now, like, in the singularity.”

Elon Musk told The Economist that AI could surpass the combined intelligence of all humanity by around 2031.

Who Feels the Predator First?

International Monetary Fund (IMF) Managing Director Kristalina Georgieva has cautioned that young job seekers are most vulnerable.

Payroll data supports her warning. Employment among 22- to 25-year-olds in roles highly exposed to AI is now 19% below its expected trend, according to Stanford researchers.

That gap was 15% a year earlier and is driven by reduced hiring rather than widespread layoffs.

Announced job cuts tell a clearer story. US employers listed AI as a factor in 120,136 planned layoffs through September, per Challenger, Gray & Christmas. That represents roughly 21% of all 2026 reductions.

Is Bitcoin's Math Next on the Menu?

That same development unsettled the crypto space. Ethereum Foundation researcher Justin Drake argued that mathematical superintelligence has arrived.

He advised large holders to adopt a "bunker mode," transferring funds to addresses that have never been used.

The concern is that AI-driven mathematics could compromise Bitcoin's wallet cryptography without requiring quantum computers. Every address that has made a transaction exposes its public key on the blockchain.

Ethereum co-founder Vitalik Buterin recommended against hasty migrations. Coinbase cryptographer Yehuda Lindell dismissed the warning as baseless alarm.

To date, there is no evidence that the underlying cryptography has been compromised.

Where Does the AI Apex Predator Still Fall Short?

OpenAI itself noted that not every proof has undergone independent verification. Gogioso also pointed out that many of the solved problems were already near a human solution.

Paid work presents a greater challenge. Scale AI's Remote Labor Index evaluates agents on real freelance tasks. The top-performing agent produces professional-quality work on roughly 21% of assignments.

That figure was 2.5% when the index launched a year ago. Former Meta chief AI scientist Yann LeCun continues to argue that human-level AI is still years away.

That uneven track record may explain the silence Edwards describes. AI dominates narrow areas while remaining inconsistent across the broader economy.

AI pioneer Geoffrey Hinton once suggested asking a chicken what life is like below the apex of intelligence. For graduates and Bitcoin holders, that question no longer feels theoretical.