In the felted but formidable universe of code competitions, the machines gain ground with each edition. Artificial intelligence models, now capable of solving complex problems in record time, impose themselves in the rankings and modify the most experienced benchmarks. However, during a finalized final in Tokyo, a Polish programmer signed a human victory against AI, by pushing his limits to the extreme to surpass a model designed by Openai in one of the most technical events in the world.
Among the competitors were Przemysław Dębiak, alias “Psyho”, a Polish accustomed to the podiums. The test? Solve a complex optimization problem in ten hours, without external help. As Tom's Hardware reports, this competition is distinguished by his requirement. It is not a question of finding the perfect solution, but of approaching the best possible score in a limited time.
A human victory against High Tension
The test required participants to lead groups of robots through a labyrinth of mobile walls, using as many orders as possible. This heuristic category rewards the strategy, intuition and the ability to adjust its choices as time passes.
After ten hours of extreme concentration, Dębiak ended in the lead, with a score 9.5% higher than that of AI. Ars Technica compares this test to the legend of John Henry, this worker who had beaten a steam engine with the strength of his arms, at the cost of his energy.
Faced with him, the Openaiahc model, from Openai laboratories, has however dominated other human competitors. Only Dębiak managed to overcome it. This victory was also publicly praised by Sam Altman, CEO of Openai.
What this duel reveals about the future of algorithmic genius
This human victory in front of AI does not in any way guarantee the sustainability of such a feat. On the contrary, it is part of a context where the AI progress at full speed. Dębiak could be one of the last humans to beat a machine in such a framework.
The figures speak for themselves. Stanford's 2025 report on artificial intelligence shows that in just one year, the models went from 4 to more than 70% success on certain programming tests. Github, for its part, indicates that more than 90% of developers already use programming assistance tools like Copilot.
However, intuition, adaptability and mental resistance remain human forces. Where the machine applies powerful but rigid methods, humans can still surprise, improvise and adjust in real time. As long as these qualities have value, AI will still have a rival to their measure.
With an unwavering passion for local news, Christopher leads our editorial team with integrity and dedication. With over 20 years’ experience, he is the backbone of Wouldsayso, ensuring that we stay true to our mission to inform.
