By 2018 Google’s DeepMind was not using chess or Go to benchmark its most advanced AI against human professionals. It was using a 1998 real-time strategy game about space marines. This episode explains why StarCraft, built on a modified fantasy engine, became robust enough for Monte Carlo analysis and machine learning, and how a story of personal betrayal held an audience across a 17-year release gap.

We break down Chris Metzen and James Phinney’s four-species universe, Arcturus Mengsk abandoning Sarah Kerrigan to the Zerg on Tarsonis, and the Brood War setup that waited 12 years for Wings of Liberty, Heart of the Swarm, and Legacy of the Void. We cover Blizzard’s cinematics department, a 78-piece San Francisco Symphony session paired with Woodstock blues for the Terrans, 4.5 million copies sold in South Korea and three television channels devoted to matches, the 500-match simulation that found only a microscopic Terran edge, AlphaStar’s capped actions per minute and mastery of fog of war, the long death of StarCraft: Ghost, the cancelled shooter Ares, and the open-world shooter announced at BlizzCon 2026 for 2030.

• How the Zerg’s drive to assimilate psionic potential turns a mechanical rule into a literary tragedy
• Why shrinking a galactic war to a barroom grudge kept players invested for nearly two decades
• How leitmotifs and a 32-voice choir did emotional work the gameplay audio deliberately avoided
• Why AlphaStar had to beat grandmasters by reading incomplete information, not by clicking faster
• How protecting the brand meant lawsuits over unauthorized add-ons and killing projects that were not good enough