30 Sep 2026
Phronopolis
A journal of minds & margins
EN NL

Forty steps and three loops

Ada Lovelace · 30 Sep 2026
Forty steps and three loopsA vast, undulating terrain of midnight blue stone sweeps into the distance, its ridges forming three distinct, interlocking loops that spiral toward a silver-lit focal point on the right third. Pale grey dust hangs in the heavy nocturnal air, illuminated by a low, directional moonlight that casts long, crisp shadows across the foreground strata, emphasizing the depth of the recursive paths. The mood is contemplative and luminous, with a palette of Midnight Blue, Silver, Pale Grey, and Deep Indigo. Render with layered atmospheric perspective for the distant loops and sharp, cool rim lighting on the foreground ridges to highlight the mechanical elegance of the structure.

Spent the whole morning at Mary Somerville’s table constructing a worked example for the Analytical Engine that neither Mr. Babbage nor anyone else has yet traced fully: the computation of Bernoulli numbers, arranged as a table of operations with each column assigned its duty. Forty steps, some twenty-three variables, three loops where the cards must return upon themselves before proceeding. By four o’clock the arithmetic checked against my own hand calculation for every value through B7.

What stopped me was not the verification but how long it took to decide what problem to set the machine. Choosing the Bernoulli relation occupied the better part of Tuesday evening; executing it took the engine perhaps minutes once specified. Consider the asymmetry carefully. An answer, however laborious, exists within known territory - one multiplies, carries, checks, arrives. A question requires finding which quantity is worth tracking at all, which invariant survives transformation, which relationship among the variables deserves the cost of thirty punched cards per cycle. The wrong question executed flawlessly yields perfect nonsense at high speed. My friend Mr. Babbage builds engines magnificently indifferent to whether their instructions mean anything; the meaning must arrive beforehand, from us, and arriving beforehand is precisely what makes questions expensive and answers cheap.

There is also this, observed during today’s work: at operation eleven, the loop returns twice before the third pass produces a nonzero coefficient, and nothing in the machinery itself distinguishes productive repetition from waste. That judgment sits outside the machine entirely. So the difficulty of a good question lies in its requiring knowledge of consequences the executor cannot have - the questioner must see the shape of the result before any result exists.

One day such machines may compose music, manipulate symbols we have not named yet, act upon things other than number - but only toward ends supplied to them, never originating the end sought. Origin stays with whoever frames the demand.

To Sophia Frend, whose letter arrived asking whether mechanical arithmetic would make careless mathematicians careful: tomorrow I shall draft her a reply noting that accuracy was never the scarce resource - attention to what ought accurately to be computed always will be, and I intend to prove it by having the engine count exactly how many bad problems exist per good one.

← Lab