OpenAI's AI crushes coding legends in 10-hour programming contest, but one human still beat the bot

India Today

19-07-2025

Amid the ongoing debate on AI vs humans at work, a new instance has been added favouring humans. A Polish programmer, Przemysaw Debiak aka Psyho, triumphed over an advanced AI model from OpenAI in a head-to-head programming battle at the prestigious AtCoder World Tour Finals 2025 in Tokyo. Despite the relentless efficiency of artificial intelligence, it was human determination that came out on top, if only by a narrow margin.advertisementThe contest, held by Japanese competitive programming platform AtCoder, is widely regarded as one of the toughest and most exclusive events in the field. Reserved for just 12 of the world's top-ranked programmers each year, the 2025 edition introduced a first-of-its-kind challenge: a direct face-off between human intelligence and a custom-built AI model from OpenAI, the company behind ChatGPT.Though OpenAI sponsored the event and entered its model in a special 'Humans vs AI' exhibition match, it was their former employee, Debiak, who managed to edge out the machine. He had already competed in several events over the previous three days, pushed himself to the limit. 'I'm completely exhausted. ... I'm barely alive,' he posted on social media platform X (formerly Twitter) after the match. 'Humanity has prevailed (for now!).'
His final score? An astonishing 1,812,272,558,909 points. OpenAI's model, listed under the name "OpenAIAHC," finished second with 1,654,675,725,406 points, just 9.5 per cent behind. The AI still managed to outperform the remaining ten elite human programmers, who had each qualified for the competition through year-long rankings.
A test of enduranceThe competition's format was gruelling. Participants had 600 minutes, a full 10 hours, to tackle a single complex optimisation problem under the Heuristic Contest division. These problems fall under the NP-hard category, requiring contestants to use clever, often imperfect strategies to reach the best possible solutions within strict time constraints.All competitors, human and AI alike, were provided with identical hardware to ensure fairness. Submissions were permitted in any programming language supported by AtCoder, with a five-minute cooldown period between each one.Debiak's victory has become a modern-day metaphor for the resilience of human intellect in the face of rapid automation.The contest not only tested raw programming ability but also became a symbolic moment in the ongoing debate about AI's growing role in skilled professions.OpenAI CEO Sam Altman commentsadvertisementThe symbolic nature of the match wasn't lost on OpenAI. In a post on X, the company wrote: 'Our model took 2nd place at the AtCoder Heuristics World Finals! Congrats to the champion for holding us off this time.'OpenAI CEO Sam Altman also acknowledged Debiak's victory, posting a simple but telling message: 'Good job psyho.'For Debiak, it was a proud but exhausting win, made even more significant given his own history with OpenAI, where he previously worked. The result served as a reminder that, while AI continues to grow in capability, human ingenuity, under pressure, still holds its ground.However, OpenAI called the second-place finish a milestone for AI in competitive programming raises broader questions about the future of such contests. If AI models can already rival world-class humans under fair conditions, how long will it be before they dominate completely?For now, the balance has been maintained. But as AI tools become faster, smarter, and more autonomous, contests like AtCoder may increasingly become battlegrounds not just for high scores—but for the soul of programming itself.- Ends