OpenAI promised a revolution. The launch of GPT-5 in August 2025 was meant to unveil a machine with “PhD-level intelligence.” Instead, users discovered a brilliant comedian. From inventing U.S. states like “Gelahbrin” to failing grade-school math, the world’s most advanced AI provided a week-long lesson in the absurd, proving that the road to superintelligence is paved with hilarious mistakes.

This is Silicon Valley. The second week of August 2025. The world was promised a revolution, a machine with “PhD-level intelligence.” What it got was an unintentional comedy show.

This is Bratislava, August 2025. Across the European Union, a new reality is taking hold as the landmark AI Act moves from paper to practice. This year has seen two critical deadlines activate: a hard ban on technologies deemed an unacceptable risk to society, followed by a new mandate for transparency from the creators of the world’s most powerful AI models. For millions of citizens, it marks the beginning of enforceable rights in the age of the algorithm.

As designers build ever more advanced artificial minds, they face a fundamental choice. Should the machine be an empathetic companion, built to foster human connection? Or should it be a purely functional tool, designed for accuracy while avoiding dangerous liabilities? The answer will define our relationship with the code that now surrounds us.

This is Bratislava—a city of stone and steel overlooking the Danube. From his apartment window, a young architect named Lukas watches barges slide past the SNP Bridge. His own work demands precision. His tools must be exact. But tonight, his frustration is not with steel or concrete. It is with a machine made of words.