Each new generation of LLMs delivers a smaller leap than the one before. Improvements do come, but they’re increasingly incremental. Most of the progress we are seeing now is around tooling - coding agents like Codex and Claude Code are changing the way much (if not most) software is being built today. But that’s an entirely different thing than the AGI that was just around the corner a couple of years ago, and is supposedly still just around the corner today.
I use Claude every day and find it genuinely useful. But my lived experience of what it can and can’t do is almost entirely divorced from what I read in the news or on LinkedIn. LLMs are a really good tool for a lot of things, but they consistently fall short of the grand promises. Anyone willing to be honest about it has noticed the gap. The headlines about what GenAI is and will be capable of are a combination of tech company executives fueling the hype machine, and news outlets looking for the juiciest headlines, even if they are out of context.
There is a governance problem nobody is talking about#
If you take the AGI narrative at face value, we are catastrophically unprepared. I’m not aware of a single government seriously working through what happens if a large share of white-collar work gets handed off to a handful of tech companies over the course of a decade. What does structural unemployment at that scale do to a tax base, a healthcare system, a social contract built on the assumption that most people work?
We’ve had about five years now to watch how governments, markets, and other institutions respond to the notion that AI will replace large swathes of the workforce. This is useful data. Just because we haven’t seen it come to fruition yet, it doesn’t mean that it can’t or won’t in the future. This is a genuine opportunity to be proactive instead of reactive.
There are some that believe everyone must work for no other reason than “because”, and they will argue there’s always service work or gig work to fall back on. I’d say: in a world full of Uber Eats drivers, who is going to be ordering the Uber Eats?
UBI becomes a requirement#
Given our current course, I don’t see how Universal Basic Income doesn’t become part of the solution. I don’t present this as an ideological stance, but a practical one derived from a basic set of requirements: if we continue to automate everything we possibly can, but still want low-crime societies full of happy, educated, and well-fed people, I don’t see a way to do that other than relaxing the expectation that everyone derives their livelihood from working.
If one genuinely believes that large-scale displacement is coming - AGI, sufficiently advanced LLMs, or otherwise - then choosing to not figure out how to make society thrive in such a world is naive, negligent, or outright misanthropic.
We have an opportunity to prepare now. Make it clear to your elected representatives they need to be working on this. If they spend any time talking about how AI will change the world but spend no time on this problem, they are doing you a disservice.