work-blog/articles/published/what-makes-us-better.md

title	date	topics	related	abstract
“What Makes Us Better Engineers?”	2026-03-06	philosophy craft automation		Whether LLMs like Claude change engineering in degree or in kind — and why the Aristotelian distinction between techne and phronesis matters for answering that question.

Recently, it has been asserted that Claude will make engineers “better”. Even if we stipulate to this assertion, what immediately comes to mind for me is, “better at what, exactly?”.

The Carpenter Analogy

Let’s simplify the question by making an analogy to carpentry (we’ll take the assumption that software development is a ‘craft’ in a similar sense as carpentry). Carpenters have many tools. Some of them are categorically different from others. For example, the hammer and the sabre (jig) saw.

The hammer doesn’t merely make the carpenter “better” at producing wood-crafted artefacts. It expands what is possible to be made, because with it the carpenter can make use of metallic fasteners, whereas without it, driving a metallic spike into a piece of wood is an exercise in absurdity.

The same is not true for the sabre saw, though it may seem so on the surface. Yes, the sabre saw makes possible a whole range of new artefacts that are not possible without it. But the difference between it and the hammer, as that the sabre saw is replacing a tool that already opened that possibility up: the bow saw.

Why is this significant? Because the real categorical distinction is not what artefacts the tools make possible, but what habits the tools make necessary in the carpenter. In a nutshell, the transition from hammer and bow saw to nail gun and sabre saw is a transition to a different kind of craft. And with it, categorically different skillsets and constraints and even materials used in the production of artefacts.

So is the sabre-saw user a “better” carpenter than the bow-saw user? Only if we redefine “better” away from the original standard. He is better at rapid, repeatable, complex curves. He is worse (or at least differently skilled) at the deep tactile feedback and material intimacy that the bow-saw craftsman possessed. The artefact itself changes character: one carries the visible signature of hand and breath; the other carries the signature of machine-assisted repeatability. As Aristotle would say, the telos (ultimate end) has subtly shifted from the qualities of a unit to quantities of units.

What Does This Have To Do With Claude?

Perhaps a more direct analogy will make the case clearer. I think that the distinction between pencil-and-paper, slide-rule, calculator, and spreadsheet PC, is one of degree, while the distinction between all of those tools and the LLM, is one of kind.

• Paper-and-pencil: Pure externalisation of memory and stepwise reasoning. You still perform every logical step; the paper merely prevents forgetting and lets you see the chain at a glance.

• Slide-rule / calculator: Accelerates arithmetic and logarithmic operations, but the selection of which formula, which approximation, which interpretation belongs entirely to you. The device has no model of the problem — it is a dumb multiplier of your own causal understanding.

• Spreadsheet PC: It automates repetition and what-if analysis, but you still have to construct the model cell-by-cell. Every formula is your own thought made explicit. The spreadsheet never surprises you with a “better” formula it invented; it only executes what your techne (practical skill) already enables.

In every case the telos remains untouched: the human retains full mastery over causes, constraints, and judgment. Phronesis (practical wisdom) is actually strengthened because the tool forces clearer articulation (“if this cell changes, what must that cell do?”). The craft expands in power and scale, but the kind of skill — deliberate, sequential, fully owned reasoning — stays identical. Hammer territory. The carpenter is still the same kind of carpenter, only faster and capable of larger structures.

But in the case of an LLM, the tool is no longer an extension of your reasoning; it is a statistical simulation of reasoning itself. It does not just compute — it generates the next step, the architecture, the test suite, the explanation, sometimes even the problem statement.

With a calculator or spreadsheet we have to type the formulas ourselves; the act of typing would require at least minimal causal thought. Claude encourages us to bypass that entirely. The model produced a fluent artefact whose internal logic we never have to own. The skill required has shifted from “model the system in your head and express it mathematically” to “prompt effectively, triage obvious junk, and make sure the black box produces the right outcomes.”

The older tools left the engineer’s hexis (disposition) intact and deepened it. The LLM invites a new hexis: fluency in orchestration, comfort with probabilistic correctness, and delegation of causal responsibility. That is not “more of the same, only faster.” It is a new craft whose telos can quietly drift toward “whatever the model can ship before the next stand-up” rather than “systems whose causes I can defend in a crisis.”

Aristotle’s test is definitive: a true tool for the craftsman must serve the existing telos and enhance the virtues required to achieve it. The hammer and slide-rule pass. The sabre saw is borderline. The LLM crosses the line into a device that can substitute for the craftsman’s core activity. It is closer to giving the carpenter a robot arm that decides where to cut than to giving him a better saw.

The Bottom Line

To be clear, LLMs are not an inherently bad things. However, they are tools that are much more dangerous than anything we’ve ever used before. It is possible to make use of Claude in the slide-rule sense, but the seductive impulse to treat Claude like a substitute for genuine cognitive effort, will only at best make you better at a different kind of craft whose name we may eventually have to change – and at worst, make you a slave to your own tools.

Still, even if we stipulate that an LLM is a tool in the sense that a calculator is, tools do not inherently confer phronesis. It is something that can only be gained by continuous practice of the right activities, in the right way, under the right circumstances. It is sovereignty over the causes plus sound judgment in irreducible uncertainty. The LLM can supercharge our techne; only we ourselves can grow in phronesis, through praxis and study. When both rise together, we can indeed become “better” engineers. That is the only definition of “better” that survives Aristotle’s standard.