The attack on competence | deadSimpleTech

In happy news, my mentoring project has been having some success of late, with a few students having retained me to teach them statistics. I don't know why everyone's so keen to learn statistics and probability theory, but I've been having a lot of fun with it and have been, over time, accumulating notes on the subject. Seeing as I a) have some difficulty tutoring people outside of particular time zones and b) have a need to pay for rent, bills, and food, I've made a start on reworking the notes for a wider audience and making them available online.

I will eventually set up monthly billing for them, but for now, if you make a donation and send me an email via my website letting me know that you've done so, I'm happy to give people permanent access to the early stages of the resource section of my website. The statistics notes are pretty good, the engineering fundamentals ones are a lot rougher around the edges, but they will all be steadily expanded and new notes will be uploaded weekly. Of course, the option of learning from me in-person is also very much there.

New Zealand has recently suffered through our current government's latest budget week. There were, of course, a whole lot of awful things in the budget, but one of the most aggravating ones was what they've chosen to do to the public service: they've elected to cut 8,700 jobs in the public sector, and when asked how the shortfall in services would be managed, Nicola Willis, our finance minister, came up with the brilliant idea that "AI" would be used to fix the shortfall. This is, of course, obviously and transparently stupid and isn't going to work no matter how you spin it. New Zealand's taxation and investment policy is likewise stupid: the aggregate effect of these tax and public sector cuts will be to sharply reduce state revenue, reducing the ability of our government to do, not to put too fine a point on it, anything at all. This has been a consistent pattern with the government: they do things that, practically or legally, simply don't work, and then blame literally anything else when, despite the best efforts of the public sector whose job it is to carry out their diktats (and which they've been cutting to the bone), the whole idea turns out to be incompetent and falls apart at the seams.

The inevitable conclusion of witnessing the last two-and-a-half years of what's passed for governance is quite simply that the people leading our country are incompetent. And yet they were elected. And it isn't as though their incompetence was hidden: it was plain to see from the get-go. People loved it. The pattern is similar in the tech industry: whatever people might say about what they want from people, revealed preference seems to suggest that they love incompetent people and hate competent ones. It's difficult, really, to read our current moment as anything other than an attack on the very concept of competence.

Competence isn't quite what you think it is

To discuss the phenomenon of widespread attacks on competence, we need to first understand what competence is. Colloquially, the word simply means clearing some bar of being "good enough" at something to be able to do it well: a superficially rather simple concept. Thinking about this, though, there are some odd edge cases. First of all, it's quite possible to be an expert in certain fields without actually being a competent practitioner in them. You meet quite a lot of them in tech: the kind of person who knows a whole lot of technical details and knows exactly how code should be written, and yet somehow everything that they produce turns out to be shit. Secondly, in a lot of fields you see people who, for one reason or another we know are good enough not producing competent work: they're burnt out, exhausted, checked out and just doing the bare minimum to get along. Competence, then, is clearly more complex than that.

A large part of the complexity in our edge cases seems to come down to desire or motivation. The main difference between the two edge cases we've described elucidates this: neither the incompetent expert nor the competent-but-checked-out person are currently producing competent work. We know that both of them know enough to produce competent work were they so inclined to, and that they have the practical skills to do so (though this might sometimes come into question in the first instance). And yet we can usually tell that one of these people is capable of doing competent work while the other isn't. The difference, in practice, comes down to why they aren't producing competent work. The incompetent expert has little care for what they produce: they don't give a shit about whether what they make creates good outcomes, bad outcomes or, in fact, any outcomes at all. The checked-out person, but contrast, cares about what they produce and knows it's bad, but thanks to compulsion, feels as though they can't not make it: they want to do competent work but are frustrated in their desire to do so. Competence, it seems, has a great deal to do with what you want.

A good working model of competence in a given field, then, might consist of three elements:

• Knowing enough about the field to know what good things and bad things are;
• Genuinely wanting the good things rather than the bad things;
• Being able to execute at least one method for getting the good things rather than the bad ones.

Each element here sits in a different category: the first has to do with theoretical knowledge and the third with the practical skills needed to apply said knowledge. The second, though, is a little different: it's about desire. In this model, a very large part of being competent is wanting the right things.

The most recent US entanglement in Iran brings this into stark relief. From the perspectives of statecraft and international relations, there was no good reason for the USA to attack Iran. Iran, while a hostile power with considerable influence, really had little ability to directly harm the USA outside of its ability to develop control over the Strait of Hormuz. It's also a large, populous and well-educated power with power structures such that any realistic attempt to subdue it would necessarily involve a massive land invasion far more costly in lives and money than it would be worth. So long as Iran could be prevailed on to keep the Straits open to shipping, then, the appropriate way to deal with Iran was diplomatic: it achieved more of the US goals in any circumstances than an attack would be.

Given this, there was in a very real sense no competent way to execute an attack on Iran, because wanting to do it given everything that we know was itself a sign of incompetence. You could know everything possible about the likely outcomes of this (and it seems that many people in the US armed forces did know it), you could have all of the theoretical knowledge that you needed to plan a maximally effective attack and you could have all the equipment, troops and training that you needed to execute on an attack, but if you thought it was a good idea to the point where you commanded it to happen, you would immediately be clearly incompetent.

In general, wanting the right things is integral to being a competent practitioner in any field. It's not enough, for example, to know that having automated tests for a codebase is good and to have the practical knowledge needed to write tests: you have to, if not to want to write tests (because it is boring), to want the codebase to have comprehensive tests that are kept up to date, and you have to want that enough to put some effort into it happening. Without that, you aren't competent no matter how much you know. Similarly, to be a competent statistician you have to really want to get the most accurate parameter estimates you want and to give your clients an analysis that will let them make the best decision they can on the basis of the information. If you're willing to p-hack to get the results that a manager wants, you are not competent, no matter how well you know how to do statistics.

This is a formulation of competence that is fairly stringent: it requires you, when developing competence in a field, to actually reshape your desires rather than simply learning skills. You have to reorient what you want and what you'll work at to focus on what the field thinks is good, to develop a level of taste and judgement appropriate to the field, and to defend that against threats when the need arises. None of this really sits well with power, and in its more aggressive forms, competence can represent an active threat to power. And thus, when power gets foolish enough, it tends to relentlessly attack competence with everything it has.

What's behind the attack on competence

A strange phenomenon arises in the modern workplace when it comes to the use of LLM tools, especially in the discussion of coding agents. One can use a coding agent in a competent or an incompetent way, to produce a more or less competent product. A competent engineer, when faced with the need or desire to use these tools, will carefully constrain them. They'll write a careful description of exactly what functionality the code needs to have before generating anything. They will write extensive tests, configure linters and prettiers and make many small commits between bursts of code generation. They will make sure to review the generated code manually and with considerable care, and they'll manually test the software they've written to make sure that they meet acceptance criteria. An incompetent engineer, by contrast, will just vibe together something that looks vaguely right, push it out and show management how cool their new thing is with little thought to the damage they've done to the codebase, which someone else will have to correct.

The competent LLM-using engineer will, in general, find themselves treated by management as though they're exactly as bad as the people who refuse to use LLMs at all. The engineers who push out pull request after pull request and line upon line of code without any thought given to maintainability, codebase conventions or such trifles as whether the fucking thing works at all are lauded, while the competent, careful people who ship carefully considered, working code (with or without LLM assistance) and clean up the messes find themselves sidelined or redundant. The reasons are often as stupid as the fact that a competent engineer, even using an LLM, will tend to prompt less and consume fewer tokens than the incompetent ones. The fact that the product breaks goes unmentioned. One can quite easily draw the conclusion that, far more than competence merely being an indifferent to management, incompetence is actively wanted and competence actively punished.

While the current massive push for LLM use everywhere can be productively understood through a number of lenses, understanding the technology as being a way to devalue and punish competence has real value. To understand why, we need to look at the investor class that currently monopolises the bulk of power in the world (inasmuch as anyone can in a system as economically complex as ours).

The investor class is obviously extremely varied across the world: any class that encompasses both Elon Musk and a petty landlord in New Zealand obviously has to be. The one commonality that they have is that they make the bulk of income from the collection of economic rents: from controlling access to a scarce resource and charging for it. As much as a silicon valley venture capitalist might claim that they aren't in the business of collecting rents, that is precisely what they do: they control access to investment capital and collect rents in the form of partial ownership of a company in exchange for making it available to entrepreneurs who want it. Naturally, as with all renting arrangements, this means that they find themselves with an awful lot of arbitrary power over the people whom they choose to make it available to, which they consistently abuse. Now, the thing about collecting rents is that it requires very little in the way of skill, talent or competence: it simply requires that the state apparatus of violence colludes with you in making people pay you money. Beyond a certain capacity for violence and cruelty, all that matters is that you can buy, inherit or steal the assets that you then choose to rent. It's no surprise, then, that rent-seekers tend to be stupid, narrow-minded and unimaginative.

The main ideological justification for wealth in the world in 2026, however, is that of innovation. The story that tends to get told is that people such as Bill Gates or Jeff Bezos have accumulated their obscene wealth because, in one way or another, they were much cleverer than everyone else and were thus able to build technologies that were beyond the rest of us, and that they then worked very, very hard to bring those technologies to the rest of us. The exact formation of this myth is complex (it'd make for a good essay in its own right), but in short, the idea is very much cemented in our heads that the wealthy of the world are wealthy and powerful because they're smart.

This leaves the investor class and those who, in exchange for access to some wealth and power, align themselves socially and culturally with them, in a bit of a double-bind. They are, as previously stated, quite stupid and unimaginative. However, the wealth that they accumulate is justified with reference to their intelligence, and they must thus be seen publicly to be intelligent to succeed (in much the same way as a mediaeval noble built legitimacy through skill at arms, especially in mounted horseback combat, or a member of the Chinese literati built legitimacy through the Imperial Examination and a deep knowledge of the classics). As a result, the investor class and the people in management positions who align with them in exchange for a share of the spoils (I do not call them managers because they do not manage) engage in constant legitimacy-building activity, which we call "thought leadership". Posting on LinkedIn, speaking about technology at conferences, the interminable meetings about innovation and new tech initiatives in general can all be thought of as precisely this form of legitimacy-building. What's important to understand here is that in this form of legitimacy-building, speaking about technology and the world in general is held to have the force of truth: they speak and reality conforms. They innovate, and innovation has happened. Any challenge to the factuality or truth of their statements is, in essence, an accusation of illegitimacy and a challenge.

Competence is naturally a threat to this form of legitimacy-building, as it means that any lowly engineer at a company can, in theory, tell the CTO that they're full of shit and have it be taken seriously. It's as though the emperor is sharing his royal court with a whole lot of people who constantly tell him that he's naked, and he can't have them executed because they run all the affairs of state. While on an individual level this might be a marginal threat, in any real concentration competent people form a competing centre of power in an organisation, with values and desires quite different from those of the investor class. They don't usually desire to be seen as intelligent, they don't desire to maximise wealth, they don't usually desire the kind of unrestrained power that a landlord has over a tenant and which the investor class seem to really enjoy. Moreover, much like early Christianity in Rome, they tend to think that these values and desires should be universal and often will not shut up about how bad the value system of the investor class is. The investor class has been trying to crush concentrations of competent people wherever they were since at least the 1970s, but they haven't been able to because, in addition to being a threat and a competing power centre, they were instrumental to the ability of the investor class to hold power at all: after all, without at least enough competent people to hold things together, the systems of government that allow the investor class to hold their disproportionate level of power fail and then the investor class winds up on the end of sharp objects.

The equilibrium that this created until recently was as follows: the most competent in any given field were generally excluded directly from positions where they'd work with or for the investor class, but sufficient opportunities were usually created that allowed them to live well enough and have some status without being a problem for said class directly. In exchange, they would train and provide material for somewhat competent people to work to do maintenance for the investor class, kinda badly and inadequately, but enough that they held together. This class of somewhat competent professionals was generally paid less well than they could be and expected to pander to the whims of the investor class, but were given some leeway to push back on the dumbest stuff or otherwise ignore the worst directives. This was obviously a terrible system during which everything slowly degraded and over time the investor class and their cronies, increasingly insulated from the consequences of their beliefs and actions (these consequences being increasingly assumed by the general mass of the population) pushed more and more for their legitimacy-building talk to be treated as factual reality, but it held together well enough that the decline was manageable.

And then Sam Altman unleashed ChatGPT on us.

The situation at present

Understood in this framework, the LLM is simply the latest in a series of technologies that have promised the investor class that this time, finally, they can eliminate the need for competence and not have to consider or respect any value system other than their own. This is the fundamental conflict that competence engenders: rather than being a question of economic jealousy, a simple status conflict (it can involve status conflicts, but they tend to be more complex than the regular kind we're used to) or a question of being corrected on the facts (not fun, but everyone has some reality they eventually collide with and have to make their peace with), resentment towards or a desire to punish competence stems from the discomfort that comes from being forced into regular, serious conflict with people with a value system that differs drastically from yours and that you have, on some level, to acknowledge is superior to yours because it does things that yours can't. The competent engineer's value system can produce working, effective and maintainable software products. The competent medical researcher can produce mRNA vaccines that save lives. The competent public servant can arrange government projects in such a way that they get completed effectively and provide useful public services to everyone in a country. If your value system can't do any of that, regular contact with people with these value systems will breed resentment and a desire to "put the experts in their place" or otherwise reassert your superiority over them by making them feel and seem inferior to you: by degrading, devaluing and punishing the competent, in other words.

Unfortunately, much of the resilience that our systems had against this kind of thing has been worn down by previous waves of this kind of thing, and with the current wave of right-wing populism at its height, our ability to mitigate the damage of the tool has largely been lost. This time, unlike previous times, things are finally just breaking beyond the ability of even the most competent of us to keep them alive. There, however, we run into the unique affordance of the LLM: you can talk to it. The ELIZA effect being what it is, we thus have a situation where, while things are finally truly breaking beyond repair, the investor class are able to absolutely cook their brains dealing with these things to the point where they genuinely believe that they've forever eliminated the need for competent people and can thus degrade us or eliminate us as they'd like.

I don't know what the long-run consequences of this are going to be, and I'd rather not find out, but barring drastic action, we're all going to learn rather soon.