Frontier AI has broken the open CTF format

Kabir, a top CTF player since 2021, argues that frontier AI models have rendered open online Capture The Flag competitions meaningless. He describes how GPT-4 first allowed one-shotting medium challenges, Claude Opus 4.5 extended this to hard challenges, and GPT-5.5 Pro can now solve Insane heap pwn challenges on HackTheBox. The CTFTime leaderboard no longer reflects human skill, as teams using AI orchestration dominate. Legendary teams like TheHackersCrew have withdrawn or struggle to place in the top 10. Challenge authors lose motivation to craft puzzles that are solved by agents in minutes. Beginners are pushed toward AI before building foundational instincts, breaking the learning ladder. Organizers cannot enforce anti-AI rules, and making challenges agent-hostile degrades the experience for humans. Kabir compares the situation to chess engines being allowed during matches. He notes that elite finals like DEF CON still resist AI, but qualifiers are automated, filtering out qualified humans. The community should pivot to learning platforms like picoGym and HackTheBox, and maintain social connections through meetups and conferences.