Do you remember this famous picture from Tim Urban’s piece in the New York Times on the path to here and the path from here? I was thinking about it over the weekend while watching Indy Johar’s Long Now talk on civilisational optionality.
Broadly, Johar argues that the first-class asset for any civilisation is the degrees of freedom available to it – the manoeuvre space to respond not just to named uncertainties but to the deep unknowns as well. Optionality, in that sense, is what lets a society recompose itself without becoming brittle. Lose it, and the future narrows into a tunnel where each crisis reduces your ability to meet the next.
Which made me wonder what widespread LLM use might be doing to that adjacent possibility space. (It is a little ironic that I used ChatGPT to help generate the images below, but here we are.)
To wit: is there a universe in which, at the individual level, there is a genuine expansion of what feels like the adjacent possible (because you can do things you couldn’t before, write code faster, draft faster, think faster), and the subjective experience is real, but at the collective level, it is convergence? A reduction of the adjacent possible at the level of the planet?
Aggregated across billions of users, a shared model becomes a shared attractor, and the outlier paths that made the adjacent possible genuinely wide get paved over. The paths that used to exist, kept alive by people walking them, stop being walked.
Are we, at some level, exchanging collective resilience for individual productivity? And it isn’t a symmetric trade because once the people who maintain the weird paths stop practising the weirdness, you cannot synthesise it back from a model trained on its absence.
Johar calls this optionality collapse through externality recoupling. The engine that expanded our optionality starts consuming the optionality it produced.
I would love to be wrong about this.
Originally written for LinkedIn on 24 April 2026. View original →
Gautam John 
Leave a comment