Deep Blue's Victory — How Chess Mastery Became AI's Mirage

OpenAI announced GPT-6 Astra this week and, almost as an aside, that 'the AGI era' has begun. The gap between those two claims — a new model and a new historical epoch — should feel too wide to cross in a single breath. It isn't, and that tells you everything about how AI companies manufacture the moments that move capital and attention.

When IBM's Deep Blue defeated Garry Kasparov in 1997, the company had proven something genuine: a machine could search a game tree faster than a human could calculate variations. What it had not proven was that machine intelligence had 'arrived,' yet within eighteen months the framing had calcified. Deep Blue became evidence of arrival because a corporation said so. The statement had the weight of institutional certainty behind it.

Research funding followed the narrative, not the evidence. For nearly a decade, AI labs chased benchmark victories. Problems with clear win conditions, measurable performance, publishable results. Symbolic reasoning, the painstaking work of teaching machines to understand logical structure and causal relationship, hemorrhaged support. Why pursue a ten-year research program in knowledge representation when you could demonstrate superiority in a game and call it proof that the field had moved forward?

The problem is not that declarations are false. It is that 'AGI' has no agreed definition, which means OpenAI gets to fill that void. When a milestone can mean anything, when success criteria exist only in retrospect and can be rewritten to fit whatever shipped, the incentive structure doesn't reward solving hard problems. It rewards announcing that the hard problems have been solved. When a single victory can be redefined as proof that intelligence itself has arrived, the company making the declaration gets to shape what counts as victory next.

This matters because you see it everywhere: the institute claiming breakthrough, the nonprofit announcing impact, the team saying the problem is solved when what they mean is 'we have built something we can measure and a story we can tell.' The checklist gets filled. The metric gets hit. And the people who had spent years on work that doesn't fit the newly declared paradigm simply move to what funding follows. That is not how you solve hard problems. That is how you make hard problems invisible.