In May 1997 Garry Kasparov paced and sweated on a global stage after a chess computer made a move so strange he assumed it saw something beyond human comprehension. It was a bug. This episode treats IBM’s Deep Blue as a biographical subject, from a Carnegie Mellon doctoral project called ChipTest in 1985, through Deep Thought’s 1989 defeat by Kasparov, to IBM’s mandate to build a machine that could beat the reigning world champion.

The episode explains how Deep Blue worked as an expert system: an evaluation function split into 8,000 parts tuned with grandmasters like Joel Benjamin, an opening library of over 4,000 positions, a database of 700,000 games, and 480 custom VLSI chess chips that let it evaluate 200 million positions per second. It covers Kasparov’s 1996 comeback through closed positions, the 1997 rematch, the game one random move that came from a failsafe in an endless counting loop, the cheating accusations, IBM’s refusal of a rematch, the dismantling of the machine, and why later systems like AlphaZero made Deep Blue’s architecture a dead end.

• Why early chess computers struggled with closed positions and human long-term strategy
• How human chess knowledge was translated into decimal scores the machine could tally
• What hardwiring chess logic into silicon did for search depth and ELO strength
• How a software glitch became accidental psychological warfare
• Why Kasparov later compared Deep Blue to an alarm clock, and where its racks sit today