AI is changing programming education and competitions through this year's International Olympiad in Informatics (IOI) in Tashkent, Uzbekistan, where 386 young programmers from 97 countries tackled six algorithmic problems. The competition tests much more than writing code: participants must understand the problem, design algorithms and data structures, implement solutions and verify that they work. AI can increasingly assist with several of these stages, forcing programmers and educators to reconsider which skills matter most.
One major change is the potential shift from writing code to spending more time on problem-solving. Chinese gold medallist Qiwen Xu noted that contestants spent more than two hours of a five-hour round turning their solutions into working code, suggesting that AI could automate part of that implementation work and leave programmers more time for reasoning. Xu is already using AI as a learning tool, studying how AI approaches difficult problems and using those approaches to inspire his own thinking.
But the article emphasizes that simply having access to AI does not eliminate differences between programmers. Swiss competitor Myriam Faltin highlights the importance of intuition—knowing which algorithm to use and when—while IOI figures argue that programming fundamentals remain essential. IOI president Tan Santek says that without a strong foundation, students cannot effectively guide or evaluate AI-generated solutions. Instead, computer-science education may need to put greater emphasis on computational thinking: understanding and formulating a problem before asking a machine to solve it.
The broader lesson is that AI may change what it means to be a good programmer rather than make programming skills irrelevant. Coding could become less of the differentiating skill, while reasoning, problem formulation, algorithmic intuition, verification and effective AI use become more important. This also has implications for schools and universities: future programmers may need to learn not only how to write software, but how to decide what should be built, evaluate what AI produces and use AI as a collaborator without surrendering their own problem-solving ability.
