Uses hidden information in strategy games to open a question about judgment and observation.
The question
What should a decision-maker infer from what cannot be seen? The essay uses competitive strategy games as an entry point into this problem.
The approach
A first-person, interpretive essay. Its public excerpt discusses AlphaStar's early demonstrations and the importance of gathering information. This record is based on that excerpt rather than a review of the complete paid post.
What remains open
A stronger empirical treatment would specify the system, version, observation interface, training conditions, and task. Distinguish an agent's limited observations from claims about the world itself, and test a capability against contemporary evidence.
What this work can establish
A historical match is not a current benchmark of AI capability, nor a demonstration of a general impossibility.
READING ACROSS THE ATLAS
A connection, with a distinction.
Orbis explores a physical interface for reflection; this essay concerns inference in an informational environment. Both raise questions about human agency, but neither is evidence for the other's proposed effects.
THE EVIDENCE TRAIL
Sources & authorship
Author of the original essay. The reading note below adds context; it does not alter the original.
- Read the original essay / public excerpt Original publication · 27 February 2026 · paid
- AlphaStar: the later primary account Google DeepMind · 30 October 2019
Kanade, T. (2026, February 27). AI's Fog of War Problem. The Frontier State. https://thefrontierstate.substack.com/p/ais-fog-of-war-problem
Reading record prepared 23 September 2026. Summary and critical reading notes were prepared with AI assistance from the linked public sources. They are distinct from the original author's work. How to read this evidence →