There is so much anxiety, ambivalence and, yes, awe around the current conversations on AI and large language models. For sure, there is a technology story here — extraordinary progress in software, architecture and hardware, meeting available capital and human opportunity. But I’m beginning to wonder whether the particular moment in which all of this is happening matters too. Because, at the same time, our lived realities — our physical, planetary experience — are becoming more uncertain. Not just risky, but uncertain. Risk is something we can broadly model and manage. Uncertainty is harder, because we may not even know the possible outcomes, let alone assign probabilities to them.
And I’m increasingly unsure whether uncertainty can be managed as much as lived through. That is incredibly expensive for an individual. Likely impossible to do for very long. It needs spaces, other people, and the ability to live and work inside the not-knowing together.
And yet, in moments of great uncertainty, our most natural instinct is to seek more certainty. It’s not like we haven’t seen this. Our politics over the last decade has shown us that when the world becomes harder to understand, simple explanations become more attractive. Answers become more attractive. People who offer certainty become more attractive. Why wouldn’t that be true here as well? Because if there is one thing large language models help us do, it is get to answers very, very quickly. It is tempting to think of that as access to information. But I wonder if it is also the machine helping us manage the discomfort of uncertainty.
And I wonder what we lose when that period of not-knowing gets shortened. Because this is often where the magic happens: time under load, time under pressure, bake time, however you want to call it. That period is a feature. It isn’t a flaw to manage away. Shortcutting it may have its own consequences, not least by prematurely closing what is still emergent and possible.
So perhaps the counterfactual bet in this moment is to increase our tolerance for uncertainty — and not just individually. Collectively. To build spaces, practices and methods that allow us to work inside uncertainty, and perhaps build greater epistemic endurance.
And this is where I keep coming back to civil society organisations and their leadership. Many have been doing exactly this for a very long time: navigating uncertainty on the side of communities, where reality rarely behaves as planned, and on the other side too, where institutions, politics and capital can be fickle. To act without knowing enough. Adapt without knowing what comes next. Continue to work on problems whose resolution is not one project away, but perhaps one lifetime away. I think there is knowledge there that matters right now. Not just what civil society might learn from technology, but what technology might learn from civil society — and what human technology we need in this particular moment.
Originally written for LinkedIn on 15 September 2026. View original →