Enshittinomics
When inefficiency is valued, all bets are off.

You might have heard recently about an exponential rise in confirmed vulnerabilities, or CVEs, related to the Linux kernel in the past few years. As a reality check, I'm here to tell you that desktop Linux users should never read or care about CVEs. Seriously.

Update your system every once in a while and never run untrusted code. 99% of them are privilege escalations, which means an authenticated user of your computer could become root. If you share your computer with your kid or partner, make sure you trust them not to be a secret agent.

Thank the security researchers for doing their thing, but do not let it concern you greatly. Linux did not become meaningfully more vulnerable since 2024, rather, it has stayed on the trajectory of being more actively targeted as it grows in popularity, but the attack vectors are still almost entirely brute force, not novel kernel vulnerabilities.

The focus on these reports is a guerilla marketing campaign for ai bots to make you think that, if for nothing else, ai is here to stay for vulnerability research.

The truth is the compute required to come up with these dwarfs any tried and true method, and nobody is paying the full cost yet.

The other hard truth is that a particular model will come up with the same vulnerabilities despite having thousands of users. The public release of a model marks the public release of all vulnerabilities that model is capable of generating because the knowledge the large language model contains is entirely surface-level. All new capabilities can be fully explored after only a few days of prompting, leaving AI security researchers with only scraps before long.

The benefit of these models is concentrated entirely in the hands of the model trainers in that they can offer exclusive access to new vulnerabilities, and of course that's exactly what the main AI labs have done. In effect, they've created a product that becomes less useful the more people use it, but still costs just as much.

My point, so far, is to tell everyone that has been instinctively grossed out by llms but who fears they are inevitable in some forms of software development or research: hold out, just a little longer. Vibe coding of all sorts is a passing fad, one that will unfortunately take with it the respect of some of the open source community's brightest minds and most promising projects.

I am, of course, pleased that ai companies have invested billions of dollars, through direct research grants and subsidized compute, in cleaning up the kernel's codebase, but I remain unimpressed, and more sure than ever that those grants and that money won't last forever, and neither will the labs themselves.

Fuck, Apple's case against OpenAI is so dead tight, that alone will force liquidation eventually. OpenAI just straight up ignored all negotiation attempts by Apple, they know delaying is their only tactic.

How investors read Apple's account and think of anything except running as fast as they can is insane to me. How can an organization still, today, decide to not only use but to institutionally rely on OpenAI internally, sending them every trade secret they have, while OpenAI being sued for stealing trade secrets potentially worth trillions?!

It is easy to assume that there must be rational actors at the top: irrationality usually doesn't pay the bills. The fact that this insanity has pervaded so many corporations is a sign that the economy itself is irrational.

Everyone can tell you how the economy is doing, barely anyone can define what an economy is. There was even this terrible commercial I saw once where a kid asked a bunch of grownups complaining, 'what is the economy?', to a bunch of crickets and flustered glances. It is so simple to define though: the economy is the distribution of goods and services.

All economic value is generated through efficiency gains. New tools make services and products cheaper, better fuels reduce transport costs, lean/jit manufacturing reduces storage costs. The GDP growth rate is actually an indirect measure of efficiency: it's the effect of lubrication added each year to make the economy have higher, or lower, throughputs of capital, goods, and services.

That conception of the economy is no longer true, in my opinion, because the relationship between profit and value has been perverted. Profit is economic friction, it is the opposite of efficiency, of economic value. Profit is necessary in order to save and spend on up-front investments in efficiency projects with a long road to maturity. It is not meant to be used as collateral for loans, sitting in a corporate bank account that is legally prohibited from being drawn, and it is not meant to buy back company stock, which is the corporate equivalent of eating yourself on a full stomach.

Money in a bank account is money that is not part of the throughput the defines whether an economy is 'good' or 'bad'. But as the investors' and corporate bank accounts grew over the last three decades, and their political power grew with it, profit has become the valued thing itself, flipping the notion of efficiency on its head. Large corporations are now on a race to the bottom to maximize profit at the expense of their economic value. They are adding inefficiencies and instabilities into their products and services knowing full well it will increase friction.

You can think of this like being in a boat in a series of locks separating two elevations of water. For a boat to travel toward the higher side, you could either drain the next lock or fill the one you're in. Either will bring you closer, but that's not really the purpose of locks, they are meant to elevate. In a sufficiently large lock, it will be difficult to tell if the water is truly rising around you or if your future is being drained for the appearance of forward progress.

We need to raise the floor, not lower the ceiling, and that will take systemic change, it will require legal change, accountability, and justice at the highest levels, it will involve structuring corporate taxes to account for the externalities that are currently being brunted by individuals and government, and it will require that we all accept a 'worse' economy, as I'm sure CNBC would call it, that thinks more about sustainable growth than taking moonshots and squandering value.

The path we are on is a dead end; the chasm between our potential and our reality is growing every day; and, sooner or later, we will drift into that final lock, with only a thin wall holding back a tsunami of economic despair.