Manchester Baby: What the 1948 Test Proved

Two records with different dates

Williams and Kilburn’s September 1948 Nature letter describes a small experimental machine built primarily to test its storage principle. It says that a program could be changed without altering mechanical or electromechanical circuits. This is a contemporary account, not a later claim about a commercial product.

The fuller paper by Williams, Kilburn and Tootill, published in 1951, explains the experimental purpose and identifies completion of the machine in June 1948. Publication date and machine-completion date are therefore different facts.

Sources for this section: [1] · [2]

Three questions to ask of a demonstration

First ask whether information survives the operations needed by a computation. A display that can show a pattern once is weaker evidence than storage that can be read and changed while calculation proceeds. Second ask whether changing the stored instructions changes the sequence of operations without rebuilding the machine. Third ask whether ordinary users can prepare, submit and recover useful work.

Those questions are connected but do not entail one another. Imagine a museum demonstrator that reliably runs two short sequences. It may establish changeable stored control while offering no printer, scheduling system, documentation or means of loading a large user calculation. The absence of those facilities does not invalidate the experiment; it limits the conclusion.

How a minimal test can be powerful

A deliberately small machine can expose repeated interactions between storage, arithmetic and control. If a proposed memory fails after certain reads or writes, the designer needs that failure before constructing a much larger installation. The test’s value lies in exercising the mechanism under relevant conditions, not in solving the largest possible mathematical problem.

For an invented demonstration log, record the initial stored pattern, the requested operations, the final pattern and whether the run can be repeated. Then distinguish a wrong program from a storage failure. A correct final answer alone may not tell you which mechanism produced it; a trace or independent check can be more informative. This is a suggested reading protocol, not a recovered 1948 laboratory notebook.

From Baby to a computing service

The 1951 paper explains the machine as a basis for understanding larger systems, while noting the trivial scale of its calculations. Its modest scale is part of the evidence rather than an embarrassment to be edited away.

A careful caption can therefore say “demonstrated an operating stored-program machine” without implying that an office could immediately order the same system and run its business. Experimental success, a larger operational machine and a manufacturer’s product need separate dates and records. That distinction remains useful whenever an invention is described through a single dramatic first.

Sources for this section: [2]