Memory was the first problem to solve
On 21 June 1948, the University of Manchester's Small-Scale Experimental Machine, or SSEM, successfully ran a program stored in electronic memory. It later became widely known as the Baby. The team of Frederic Williams, Tom Kilburn and Geoff Tootill was not primarily trying to perform large calculations for users, but to test whether a high-speed storage device would operate reliably during real computation.
Electronic circuits could already calculate rapidly, but without equally rapid access to data and instructions, input and control would still slow the whole machine. Drawing on cathode-ray-tube technology developed in radar research, the team attempted to turn charge states on a screen into readable and writable binary memory. Kilburn's 1947 progress report described the principles and plans; operation of the complete machine in 1948 advanced a component demonstration into a continuous program-controlled test.
Screen charges were not ordinary pictures
The Williams–Kilburn tube used an electron beam to establish different charge distributions on a cathode-ray-tube screen, representing binary states. Reading depended on electrical responses, not an operator visually recognizing displayed characters. Because charges decayed, control circuits repeatedly read and regenerated them, turning transient physical traces into information that could be retained.
The electron beam could select storage positions for rapid access to a specified word. The Baby's main memory held thirty-two 32-bit words, totalling 1,024 bits, into which both program and data had to fit. A dedicated display tube helped staff inspect the contents, but visible bright points were only an aid to debugging. The engineering requirement was to coordinate reading, writing, refreshing and computation without destroying information already stored elsewhere.
Fetching instructions, calculating and changing sequence
The Baby processed information serially in binary form, used an accumulator for values during computation, and recorded the next instruction's location in its control section. Programmers entered encoded instructions and initial data into memory through switches. During execution, the machine fetched an instruction, interpreted its operation and address, and performed actions such as subtraction, storage or changes to the control sequence. Results could affect the next choice, allowing repetition and conditional decisions within the machine.
The important change was that a new task could be implemented by changing stored contents rather than rewiring an entire control system for each calculation. Instructions and ordinary numbers both had binary representations, with execution determining a word's role. The stored-program concept had already been discussed and designed, however. The Baby's contribution was continuous operation of the approach in actual electronic hardware, not the origin of every underlying idea on one day or through one inventor.
References: [2]
Finding a factor as an endurance test
Kilburn's first successful program sought an integer's largest proper factor, testing candidates through repeated subtraction. The laborious algorithm did not imply ignorance of faster mathematics: repeated execution subjected memory to many reads, rewrites and conditional transfers, making it useful for exposing intermittent faults. Lavington's computer history records a run of approximately fifty-two minutes, involving millions of operations and producing the correct result.
In their short report published in Nature on 25 September 1948, Williams and Kilburn explicitly described a small, primarily experimental machine. It demonstrated the principle of changing programs but did not yet provide routine scientific computing services. Their detailed joint paper with Tootill in 1951 further explained circuits and operation. Publication followed the test, so 1951 must not be mistaken for the year the Baby first worked.
From an experimental machine to a usable system
After the Baby's success, the Manchester team expanded the machine. The Manchester Mark 1 of 1949 increased storage, incorporated a magnetic drum and developed mechanisms such as indexing, while input, output and conditions for program use improved. Ferranti subsequently converted university research into an engineered product, delivering the Ferranti Mark 1 in 1951. Testing a principle, establishing a university computing facility and manufacturing a machine for supply were related but distinct steps.
Other teams were also exploring stored-program computers. Systems including Cambridge's Electronic Delay Storage Automatic Calculator used different memory technologies, with mercury delay lines and cathode-ray tubes involving different engineering trade-offs. The Baby did not mean all later computers would retain its components. Its value was demonstrating that electronic memory could hold both data and a modifiable program while remaining consistent through repeated execution. This testable whole-machine result made memory central to the organization of computation rather than merely an accessory.