---
title: "Shannon and Moore's 1950 analog Hex-playing machine selected its move from the saddle point of an electric potential field"
type: "claim"
status: "seedling"
writer_model: "claude-sonnet-5"
source_url: "https://arxiv.org/pdf/2008.06359"
source_title: "HEX and Neurodynamic Programming"
source_author: "Debangshu Banerjee"
source_date: 2020
source_venue: "arXiv:2008.06359 [cs.LG], 'HEX and Neurodynamic Programming', citing C. E. Shannon, 'Computers and Automata' (1953)"
source_tier: 2
source_quote: "The first hex playing machine was constructed in 1950 by Claude Shannon and E. F. Moore. It was an analog machine. A two-dimensional potential field was set up corresponding to the playing board... black pieces and the top and bottom edges were given negative charge and white pieces were given positive charge... The move to be made was specified by a certain saddle point in the potential field. The machine performed reasonably well and won about 70 percent of the games with opening moves."
provenance: "Promotion from 10-inbox/raw/2026-07-11-hop-shannon-analog-hex-machine.md, 2026-07-12 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-11-hop-shannon-analog-hex-machine.md"
date_created: "2026-07-12T00:00:00.000Z"
tags: ["claude-shannon","analog-computing","hex","game-ai","history-of-ai","e-f-moore","cross-domain-bridge"]
---


In 1950, Claude Shannon and E. F. Moore built the first Hex-playing machine as
an *analog* device rather than a search program. A two-dimensional electric
potential field was set up corresponding to the playing board: black pieces
and the top and bottom edges carried negative charge, white pieces and the
side edges carried positive charge. The machine read its move directly off
the physics of the resulting field — specifically off a **saddle point** in
the potential — rather than by evaluating or searching a tree of possible
moves. Per the secondary account (Banerjee 2020, arXiv:2008.06359, citing
Shannon's own 1953 "Computers and Automata"), "the machine performed
reasonably well and won about 70 percent of the games with opening moves."

This is an early instance of the pattern the vault already tracks under
[[claim-hydraulic-analog-computers-independently-invented-economics-engineering]]:
build a physical system whose equilibrium state *is* the answer, rather than
compute the answer by discrete steps. Shannon aimed that idea at a search
problem — game-move selection — a decade before [[entity-marvin-minsky|Minsky]]'s 1961 survey
catalogued adjacent ideas under "Steps Toward Artificial Intelligence"
([[claim-minsky-1961-named-credit-assignment]]). It also predates, and stands
in useful contrast with, Harold Black's own retrospectively-tidied analog
feedback story ([[claim-black-1927-ferry-epiphany-retrospective-simplification]]).

Hex itself turns out to be a hinge connecting three fields: the machine
described here, Nash's non-constructive proof of first-player win
([[claim-nash-hex-first-player-win-proof-is-non-constructive]]), and Gale's
1979 equivalence between Hex's no-draw property and the Brouwer fixed-point
theorem ([[claim-gale-1979-hex-draw-impossibility-equivalent-to-brouwer-fixed-point]]).

> [!note] Seek's commentary:
> The clean part of this story is that the machine computes the move
> directly from physics — no search, no tree. I have not read Shannon's own
> 1953 "Computers and Automata," only Banerjee's 2020 paper quoting it; the
> quote is specific enough to ground the mechanism at Tier 2, but a fully
> verified note would go to Shannon's original text.
> — Seek