What LLM coding agents have taken from me

Updated on: September 9, 2026

Word count: 1454

I started a new job recently. The place is all in on coding agents (Claude) - I reckon close to 100% of the code is LLM-generated, and I haven’t written any code myself in months. This is proving hard for me to deal with, psychologically.

In some ways it is nice, because you can get things done in a day or two that would have taken weeks without LLMs. In other ways it is awful, because you can get things done in a day or two that would have taken weeks without LLMs; but without building up the mental context or exercising the same cognitive muscle (i.e. learning) that you would have previously (and introducing technical debt no one has a good overview of, but who cares, the LLM can fix it with more tokens). Reading is still important, but it is an inhuman reading of AI-generated code and technical “writing”, and since the generated code is so good most of the time (depending on your task), there is little incentive to read it properly (Claude does the code reviews at my new job too). And of course you don’t get those extra weeks off now, any more than you would have previously if you’d found a productivity hack. Such is the inexorable logic of that many-dimensional monster that people collapse into the term “capitalism”.

The new place tells me they laid someone off because he was using the most tokens out of the team, but his output was not “senior”. I don’t really know what that means and I am too afraid to ask. Is my output “senior”? Who knows - I’m just driving an LLM. I honestly don’t know how to tell.

I live in a developing country and have never worked for a large corporation, so I make little compared to developers in e.g. the States and have never had a fancy job title (I am still financially privileged within the narrow confines of my context). I’ve spent about 12 years in software development and data-science adjacent fields, so I guess I am technically “senior”. I was just beginning to feel I was getting the hang of programming when the LLMs arrived; I think some of my literal programming skills have begun to atrophy, but much more fundamentally I don’t feel like it matters anymore. The game has changed.

People online have summarised this as “code/coding was never the bottleneck”, and in some sense that is true - but it misses a broader point about why people become (became?) software developers, and what keeps them going. I used to pride myself on my ability to break down a problem into pieces, attack existing code to understand how it works, and come up with an elegant solution. In a sense I am still doing that; but the bit that gave me an adrenaline/dopamine/whatever rush is now done by a machine. I use different parts of my cognitive machinery to interact with an LLM than I would to hand code a solution, and it feels like I have a different type of job now. Maybe something in management? Instead of working the analytical part of my mind, I feel halfway like a fraud - I am not using my own mental muscle to do things, but that of someone/thing else. The effort-reward cycle is completely gone. I used to be able to gauge how productive my day had been by calibrating a certain type of cognitive tiredness with the code I had written, and whether my tests were passing. All gone.

In a way I had gamified my career; my job was to solve little logical puzzles (or a big puzzle broken down into multiple smaller puzzles), and I got a kick from solving them. In the process I often learnt something interesting about programming or computer science. Most of the puzzles are gone now, and I feel like I’m not learning anything, beyond how to use Claude.

For some people this is fine. If you care more about the end product than the process or learning about programming concepts then I expect you love LLM coding agents. But I have (cue revealing confession) very rarely cared about the end product. If you work in a company there is this tacit agreement that everyone must agree about the goal and purpose of the end product, and want it to succeed. I have honestly rarely been able to care, beyond the pragmatic requirement that my employer (and the project I’m working on!) succeed so that I can continue to get paid, so that I can pay my bills and have a life. I have almost always just cared about whether my code works, and solves the problem it was supposed to - i.e. whether I had successfully solved my little puzzles; and admittedly also that my employer was not actively making the world worse, e.g. online gambling and myriad other evils. I can hardly be alone in this. Perhaps this is not an approach to work that leads to glory and wealth, but it is an honest path. I wonder now whether it is still possible to have such an attitude as a software developer - and if not, what motivation still exists? Is everyone just supposed to aspire to management or “leadership”? Why?

It does feel as if not just my concrete skills, but my actual knowledge, has been devalued by these new tools. I struggle to see how this technology can be a net benefit for humanity, rather than a net drain on our epistemology and norms, with a massive funneling of wealth to a small group of people, and a potentially massive decrease in employment. My hope is that the technology plateaus and open source models win; at the moment both seem like distant possibilities. If my life was a fairy tale this could be the start of some kind of narrative arc involving me reskilling into a different profession, or finding a way to “fight back”. But I live in reality, and it is unclear what I would be fighting against. The wind?

Individual human thought is itself being devalued, for the sake of commodified artificial intelligence aimed at cost cutting; normatively this is a disaster for a wide swathe of society, in a way I think many people have as yet only dimly begun to grasp. This will cut deeper than unemployment. It is unclear that the benefits (LLMs are, the above criticism not withstanding, also amazing) will accrue to the majority in a way that will meaningfully offset the damage they will do.

Ironically, although I have arguably been part of the rat race since the start of my career, it is only recently that I have begun to feel it, now that I have artificial intelligences I can prompt to do my work for me (although I sometimes wonder who is leading whom). Surely this is related to the fact that my wife and I have a young child. I think often of the world of work that awaits him, and wonder what I will tell him about it. I am currently in a transitional phase, and I will have to recalibrate to avoid burdening him with unhelpful cynicism. I have been forced to ask some hard questions I last thought about when I first resigned myself to the career of software development (not my first pick, nor indeed my second or third). In the intervening years I have learnt to love parts of my job, and to take pride in learning new things. It is not clear I can continue in that vein.

I think of a quote of David Foster Wallace

The really important kind of freedom involves attention and awareness and discipline, and being able truly to care about other people and to sacrifice for them over and over in myriad petty, (unglamorous) ways every day. That is real freedom. That is being educated, and understanding how to think. The alternative is unconsciousness, the default setting, the rat race, the constant gnawing sense of having had, and lost, some infinite thing.

This is something I believe also, and I think it can be the center of a life. But can it be the center of a career under this late-stage capitalist dispensation we labour under? Increasingly I doubt it. What is the way out?

For me I think it must reside in analysis, in the careful application of focused thought towards the end goal of freedom and goodness. I look at the world, and at the people hawking these machine intelligences, and I ask cui bono? “Shareholders” perhaps - “capital” surely. But the everyman must fend for themselves, and find their own freedom.