Part I - Retrospective Benchmarks

Retrospective Benchmarks for Machine Intelligence

Part I: The AGI Series

Did we meet their standard for AGI?

Six definitions of Artificial General Intelligence, spanning 1997 to 2023. Each was explicitly trying to specify what machine intelligence would require. Each yields a different verdict when applied to frontier AI systems as of December 2025.

Chapter 1: The Gubrud Benchmark (1997) Does it rival the human brain in complexity and speed?

Chapter 2: The Reinvention Benchmark (2002) Can a single system handle broad cognitive tasks?

Chapter 3: The Formalization Benchmark (2007) Does it achieve goals and learn across environments?

Chapter 4: The Corporatization Benchmark (2018) Can it autonomously outperform humans at economically valuable work?

Chapter 5: The Critique Benchmark (2019) Does it efficiently acquire skills on genuinely novel tasks?

Chapter 6: The Synthesis Benchmark (2023) Where does it fall on the spectrum from emerging to superhuman?

Evaluated December 2025. Working papers—comments welcome.