The Babbage Trap: When to Stop Mining for Gold Because You Suspect There’s a Path to Diamonds

A Founder’s Live Dilemma

I am writing this essay because I am currently in the middle of it.

Charles Babbage is the ultimate paradox of the industrial age — a man who architected the digital future but could not ship his mechanical present. As the designer of the Difference Engine and the Analytical Engine, he is rightly celebrated as the father of computing. Yet his career is a study in strategic misalignment. He famously abandoned a nearly-finished, fundable machine — a “wedge” into the market of mathematical tables — to pursue a grander, general-purpose programmable architecture.

This transition was strategically brilliant but operationally disastrous. I call it the Babbage Trap: the tendency to let a higher-order vision consume the credibility, funding, and institutional patience that only a finished minimum-viable platform can provide.

As co-founder and CTO of an AI startup, I now face this question constantly. Modern AI tooling lets me hold dozens of strategic alternatives in flight at once, and the abundance of bare compute available to early-stage founders means the constraint is no longer resources — it is judgement. Should I keep mining the gold seam in front of me, or chase the geological evidence that suggests diamonds three valleys over? Babbage’s career, examined rigorously, offers a framework for that decision.

The Gold-and-Diamonds Problem

Every founder who builds for long enough hits a moment where the seam they are mining starts to feel narrower than the seam they suspect lies elsewhere. The temptation to abandon the current path is rarely about boredom. It is about a credible technical insight that the current problem is a special case of a larger one — and that solving the larger one would be transformative.

This is the only honest reason to seriously consider a pivot. But it is also the most dangerous, because it is intellectually flattering, harder to falsify than market feedback, and infinitely fundable inside one’s own head.

The discipline is to refuse the binary. The mature move is not “gold or diamonds.” It is: finish the gold mine to fund the diamond expedition, and run the diamond expedition as a falsifiable experiment whose first job is to try to kill itself on scientific grounds. Alphabet’s X program calls this “monkey first” — prove the hardest, most invalidating part of the moonshot before you build the easier parts. The point is not enthusiasm. It is rigorous, time-boxed disbelief, applied to your own most beloved idea.

The rest of this essay is the framework I use to keep that discipline.

Pivot Only When You Discover a Higher-Order Primitive

Babbage’s pivot from the Difference Engine to the Analytical Engine was not a product iteration. It was a fundamental jump in the level of abstraction. The Difference Engine was a special-purpose mechanical calculator for polynomial functions. The Analytical Engine introduced the primitives of modern computing: punched-card control, separated “store” (memory) and “mill” (processor), and general-purpose operation.

This is the only valid reason to pivot: when the current project teaches you that it is a degenerate case of a larger architecture. Babbage moved from “automating a calculation” to “making calculation itself programmable.”

The intellectual spillovers were profound. Ada Lovelace’s 1843 account framed the machine not as a numerical tool but as a symbolic device capable of manipulating music or logic. Alan Turing’s Automatic Computing Engine was a deliberate homage. John von Neumann acknowledged that the universal machine was anticipated by Babbage.

Vision Without an Artifact Becomes Mythology

The bear case against Babbage is unsparing. He possessed a concrete, high-value wedge: the crisis of human error in mathematical tables used for navigation, astronomy, and finance. By 1833 a portion of Difference Engine No. 1 had been demonstrated to great acclaim. Had he completed it, he would have created category-making legitimacy. A working machine would have made automatic calculation real, useful, and economically justified to the British government and the public.

Instead, his pivot moved him into a far more complex engineering programme before the first had delivered. By abandoning the wedge, he lost his constituency. Without an operating artifact to prove the category, his grand designs were relegated to intellectual theory — useful mythology, but mythology nonetheless — for over a century.

A pivot that abandons proof destroys the very bridge required to reach the larger vision.

Past Spending Is Sunk; Past Learning Is an Asset

In systems thinking, a pivot is not a rejection of the past but a dynamic optimisation of the current state. Strategists often invoke the sunk-cost fallacy, but Babbage’s career reveals a more nuanced truth: past spending is a sunk cost; past learning is a sunk asset.

The rigorous decision rule is a Bellman calculation: choose the action that maximises current reward plus the discounted expected future value of the resulting state.

Babbage’s state at the time of his pivot included technical know-how, manufacturing techniques, and design drawings. The proof of this asset accumulation is the Science Museum’s 1991 build of Babbage’s Difference Engine No. 2 — designed later, it required only one-third of the parts of the original because it benefited from simplified mechanical logic that Babbage developed while designing the more complex Analytical Engine.

We must, however, distinguish private value from social value. As a private project manager, Babbage destroyed project value by failing to ship. As a frontier explorer, he was peerless — increasing the social option value of the entire species by articulating the architecture of the computer. Most of us cannot afford to optimise for the second.

The Kelly Criterion: Ruin Is an Absorbing State

Positive expected value is insufficient to justify an all-in pivot. The Kelly Criterion dictates bet sizing to maximise long-run geometric growth while avoiding ruin.

For a startup, the survival bankroll is more than cash. It is team energy, funder patience, and institutional credibility. Babbage’s error was one of sizing: he staked his entire institutional bankroll on the Analytical Engine — a high-upside, high-variance bet — before establishing compounding proof.

In the Kelly framework, ruin is an absorbing state. Startups do not get to average results across parallel universes; they live on a single path where, if the bankroll hits zero, the net present value of all future options becomes zero. Because estimates of probability of success (p) and upside multiple (b) are systematically overconfident, the disciplined strategist uses fractional Kelly — betting ten to twenty-five per cent of the theoretically optimal fraction.

Translated to the gold-and-diamonds problem: allocate a small, time-boxed fraction of your bankroll to the diamond expedition while the majority of the team finishes the gold mine. If the diamond expedition’s “monkey-first” experiments don’t materially change your beliefs in a defined window, the expedition retires and the gold mine continues uninterrupted.

The Three-Layer Rule

To avoid the Babbage Trap, distinguish three layers of commitment:

A founder should not pivot away from a core mission simply because a model update commoditises a feature. Pivot the implementation while protecting the strategic primitive. Follow the bottleneck: when generation becomes cheap, provenance and workflow integration become the new high-value primitives.

The Counter-Cautionary Tale: HMS Victoria and the Cost of Being Overly Victorian

If Babbage warns us against premature abstraction, the Victorian Royal Navy warns us of the opposite failure: the refusal to deviate when the evidence in front of you is screaming.

On 22 June 1893, Vice-Admiral Sir George Tryon ordered two columns of the Mediterranean Fleet to turn inward toward each other at a distance that every officer on every bridge knew was insufficient. HMS Victoria and HMS Camperdown collided. Victoria sank within minutes. Tryon drowned with her. The disaster’s enduring lesson is not that Tryon misjudged the geometry — he certainly did — but that no one in the chain of command corrected him in time, because the institutional culture rewarded fidelity to the plan over fidelity to reality.

The Babbage Trap and the Tryon Blunder are mirror failures. Babbage abandoned a buildable engine because he saw a larger architecture and could not resist it. Tryon executed a doomed manoeuvre because he had committed to it and the institution could not bring itself to interrupt him. The discipline a founder needs is the synthesis: be ruthless enough to abandon a wedge when the evidence demands it, and also ruthless enough to abandon a beloved pivot when the evidence demands that. Most founders are biased toward one failure mode or the other. Know which one you are.

Be Babbage in Notebooks, Not in Delivery

The synthesis is a lesson in sequenced commitment. Exploration is justified when the next experiment can materially change your mind; exploitation is justified when additional information has low marginal value relative to the returns of execution.

Modern founders should strive to explore like Babbage — writing down the larger architecture in their notebooks and chasing the prophecy of symbolic manipulation — but exploit unlike Babbage by shipping the smaller engine first. A pivot is wise when it preserves compounding; it is foolish when it converts a hard but finishable product into an elegant unfinished cathedral.

So when you next sit with the temptation to chase the diamond seam, ask the rigorous question:

If you cannot answer that question with a falsifiable next experiment and a Kelly-sized budget, you are not pivoting. You are romanticising.

— Jesse M. Blum, co-founder and CTO, Numonic Labs