Against LLM Doomerism

⁜ 2026-08-14 (last updated: 2026-10-15) by giulianopz

In this age of great technological leaps and marketing hype, one has to carefully scrutinise their beliefs and opinions to preserve a little bit of mental stability. Programmers, especially, are among the most affected (although if privileged) categories, as prosumers of this AI/LLM revolution (which started not so long ago btw, in November 2025).

One has to remember that LLMs are just another tool in the programmer’s toolbox, even if of a very different kind, and clean up their feed from both LLM haters and LLM evangelists.

The community of programmers has been sharing its thoughts to help navigate the uncharted waters of this exciting yet troublesome transition. Reading what has popped up recently can serve as a valuable guide for the future:

They all seem to (involutarily or not) echo some well-known voices from the past:

A critical view of LLMs does not downplay the scientific significance of their discovery; their success is a severe comment on how most of linguistics has ignored probabilistic approaches to the study of language (Mr. Chomsky, the eminent scholar and imaginary anarchist, is not to blame only for being friend with Epstein and Bannon).

Instead, it should protect people (programmers and non-programmers) from cognitively surrendering to a consensual hallucination that presents LLMs as a God-like Logos: an infinite wording that generates all the reality. Or a simulation that produces simulacra in a while-true loop, if you prefer.

It’s clear that we are also in the middle of a crisis of presence: I borrow the term from the Italian anthropologist Ernesto De Martino. It describes a profound existential moment in which an individual feels the acute risk of losing their place in history, society, and reality itself, effectively threatening their capacity to act as a human being and to transform the world through their actions.

This already happened in history. Thousands of years ago, in an Age of Anxiety, the Christian religion emerged, saturating the offering of meaning, order, and consolation in a time where the Roman Empire was collapsing; and it used people’s spiritual needs to effectively exercise a biopolitical force.

Everyone knows (or should know) the final outcome of this story.

And so please, stop bother us with agent consciousness when even the mind scholars don’t know what it’s about.

Knowing where to place our trust and how things work (do you remember the “build X from scratch/build your own X” frenzy raging just a few time ago?) remains relevant. LLMs are just a (powerful) tool to augment our zone of proximal development, to put it into psychological terms. But no technology is neutral. It’s up to us to make it a convivial tool that can help humanity to advance in history as a whole towards a better future.

This won’t happen until LLMs are an external service that requires a $ub$cription, high electricity consumption and computing power that only high-end hardware can support. As things stand, LLMs only risk exacerbating the conflict between capital and labour (see Will AI Crowd Out Jobs?, Paul Krugman) and increasing the power asymmetry which separates governments and corporations from civil society. Normal people (i.e. non-billionaires) will never accept the bad things (i.e. negative externalities) produced by this new technology. And AI CEOs would hardly grant us fully-automated luxury communism.

Code may be cheap now, but it is expensive to maintain, to trust and to build a community upon: you can generate all the slopforks of Postgres you want, but you will fail to understand the real value of projects like Postgres is not its code, but its community of contributors who are devoting their life to improve and maintain its code base and documentation for the rest of their life, hopefully.




Don’t Bend! Ascend!