When Olympic coding champions grade AI homework, the results are sobering: best models solve 53% of medium problems, 0% of hard ones. Turns out those "AI beats human coders" headlines might need a reality check! 📊

How Do Olympiad Medalists Judge LLMs in Competitive Programming? - Slashdot
A new benchmark assembled by a team of International Olympiad medalists suggests the hype about large language models beating elite human coders is premature. LiveCodeBench Pro, unveiled in a 584-problem study [PDF] drawn from Codeforces, ICPC and IOI contests, shows the best frontier model clears j...