It's not very often that I get feedback on the blog posts I write. Some of my friends have the distinct privilege (or curse?) of hashing out my ideas over the course of days, weeks, months, sometimes years before they ever become posts, if they do at all. This one has been boiling slowly for a couple of years now, a reaction to a shift in how the work I do is perceived, and in how I perceive myself doing it. So, let's try this again, a little less polemic this time.

The output matters.

The biggest criticism I received for the last post was that my argument that the output didn't matter was fundamentally flawed. As someone who proposes to care deeply about what they do, who is spending time to think critically about it, and then engaging in Socratic debate about the merit of that position; the output obviously matters.

And to my friends who have tacitly argued this, I must acquiesce, they are correct. The output matters greatly to me, but not for the reason we're focused on.

When a craftsman engages with their craft the intent and purpose of their labor is to create something beautiful. If that craftsman is disengaged, divested of the responsibility of what they're performing, uncaring in how they do what they do; then the output will not be great. These two things are inseparably intertwined from an Aristotelian perspective. But it's deeper than just the act of creating something. How did the craftsman engage in the creative process? What shortcuts were taken? What habits were perpetuated? What was the intent of the craftsman choice to practice their craft? The output is the goal, but it is also a byproduct of the process of craft. But focusing solely on the goal isn't how we reach it as people.

This is my fundamental argument against LLM usage, it short circuits the steps in between the creative process and the goal of creating that output. And if we iterate over that process our skills will atrophy as we fail to habituate them. Aristotle describes this more eloquently than I can. The output exists due to the combination of poesis (the creative act) informed by techne (craft skill) and phronesis (wisdom to understand why, how, and when to) perform a craft. If we outsource the creativity to the LLM, brainstorm using it, then we short circuit poesis. If we outsource our techne, agentic work, then we don't practice our craft. If we choose to stop practicing either poesis or techne then we either fail to build phronesis or accept that we are willing to miss out on developing the wisdom to more deeply understand a component of our craft. For me, personally, compromising this is a fatal flaw that results in the atrophy of skill.

But why does the skill atrophy? The counterargument to my entire statement is that there are, without a doubt, some subset of labor that is not conducive of my time and effort to perform. Drafting an email, writing a script, documenting something; all of these in theory sound viable, but what is the macro impact delegating these tasks to an LLM? I find very little benefit in using an LLM to organize and draft prose, I'm plenty capable of generating extremely verbose and intellectual prose myself, further if I outsource a human touch point I deprive myself of being able to fully engaging in an empathetic manner with that person. I want to be respected enough to hear your voice, flaws and all, and will return mine to you in turn. A script is a good idea in theory, quick and dirty work that's short lived and limitedly technical. But unfortunately the LLM doesn't learn, a junior admin however does. I'd rather delegate technical work to people, so that they have an opportunity to learn, so that I have an opportunity to engage with them and teach. That's how I learned, I was given enough rope to hang myself with or get down. I broke SO many things, and I had to fix them myself, but the grizzled admins I learned from were there to help. The LLM isn't really a steward in the same sense, even if I can reach it instantly. And documentation is a synthesis of understanding, it's a product that someone will consume, how can I produce quality functional documentation without engaging in the process? The tasks themselves could be executed sufficiently by an LLM, but how do I grow in so using them? How do I provide opportunities for others to grow as well? This then leaves LLMs in a precarious position between work I should be doing, or work I should be delegating, and it doesn't fully resolve the tensions I have with either action.

This isn't a cut and dry "LLMs bad" take, I think they can be useful tools. I don't think the environmental impact of their use justifies that, but from a purely technical standpoint I am comfortable admitting that they are useful. What part of my poesis or techne should I sacrifice to that tool then? I can make it generate ansible playbooks, or terraform modules, or salt states. But at some point I have to own that output, but I don't have the muscle memory or mental model built up from actually implementing it. I can tell you the shape of those things, what they're supposed to do, but at a high level, not intimately. In the process of divesting myself of creating or thinking critically about them (in tandem!) I lose my wisdom about them. All of this results in akrasia (essentially ethical/moral failure). If as a principal level engineer I implement a system that I cannot fully explain from a technical standpoint, if the only level of engagement I've given it is the shape and review of what it does, then I've fundamentally failed to execute my craft.

I don't like working like that.

There is no joy in that process.

I want the ownership, the technical knowledge, the deep damn near philosophical zen of understanding the stack from the programs we compile to the CI system that builds it to the infrastructure it rides on and the name of the person who consumes the product it delivers. That's the degree of ownership I want from the execution of my craft. Because it's what makes engaging with it worth while.

I don't think I fully expressed this in my last post, I was too busy trying to make a ridiculous nod to how my former AP Art History teacher explained Socrates metaphysics with chairs, and so I shoehorned a bunch of emotions into a metaphor about chairs that didn't really meaningfully convey the message I was after.

My friend Eli got the sense of what I was trying to say, and I really encourage everyone to read his discourse on the topic as an extension of my own, and this continuance.

This is a constitutive claim rather than a teleological one. Value doesn’t accumulate at the end of a process, it exists in the process or it doesn’t exist at all. Gerund over noun.

My criticism of Ikea furniture and by comparison the elevation of Thos Moser furniture isn't that either company makes an objectively better "chair". They both make chairs, each company has an exceptionally well documented history of respect for their craftsman and the work that they produce. Both of them serve different markets, where Ikea's intent was to create democratized furniture, and Moser's was to create luxury furniture. Neither chair is objectively better or worse in its chairness, but the process in which they were created is fundamentally different.

Ikea produces at a scale that necessitates a complex global supply chain to manufacture the massive catalog of goods in the quantities they sell them. Thos Moser produces relatively little. They're both still making chairs, but in the process of making a Thos Moser chair there was a real human presence. Someone was home, contemplating the action from start to finish.

The Thos Moser craftsman signing the chair isn’t certifying that the chair achieved its proper form, they’re marking that a person was present and caring throughout the making of it.

I perceive my craft skill like that of a Thos Moser carpenter. I sign my real honest to god name on my open source work. I put my heart, and soul, into the stewardship I provide to open source communities I work with; and I feel the same way about the work I do professionally that enables me to do that. Eli drew on Emma Goldman to frame this, like him, I find her articulation to be delightfully amusing, and incredibly precise.

Goldman insists that the living cannot be deferred to some achieved future state. Like, if I can’t dance, it’s not my revolution instead of “we’ll dance when we win.” The dancing is the revolution.

For a lot of people engaging with LLMs it's just another tool in a toolkit that they can use to augment what they do. I respect that. I am struggling to apply it to myself, not because it can't, but because it absolutely emphatically can. Go ask Claude to package something for Alpine, it can. Or maybe put together a salt extension, totally viable. I can choose tomorrow to take my name off of everything that I do, but I don't want to. I get only one life to live, so I insist that I must fill it with meaningful things. This is one of them, for me.

Ergo the shape of my argument is actually closer to what Wendell Berry describes in "The Unsettling of America". The process of automating and outsourcing the Aristotelian craft process from the output alienates people from the very skill they're practicing. Berry's framework is agricultural, so he's mostly concerned with large scale farming divesting people of an understanding of how to respect nature and be good stewards of the land, but that's not all that much different than what I do. When I package something for Alpine I'm not just throwing it over the fence to be abandoned. People depend on that, I maintain projects like cloud-init and salt that literally touch an unfathomable number of servers. That requires stewardship, I can't just outsource it to an automation system and hope it's good.

But that's just my perspective, in this moment. I'm currently reading through a number of books to try and better understand how to think about these things. In a sense it's the same first principles approach to stewardship I'm describing, but from a philosophy standpoint. I want to know how to reason and analyze my own opinions so that I can grow and learn. And I don't want to close the door entirely on this technology, it's rapidly transforming our industry and will continue to do so. I just want to engage with it authentically and I worry about my ability to do so without losing myself in the process.

For the curious

These are the books I'm reading as I'm contemplating everything. I'd love additional suggestions if anyone has them.

The Republic, Plato
Nicomachean Ethics, Aristotle
The Myth of Sisyphus, Camus
The Rebel, Camus
The Ambiguity of Ethics, Beauvoir
The Unsettling of America, Berry
What are People for?, Berry
The Question Concerning Technology, Heidegger
Labor and Monopoly Capital, Braverman