A doomer is someone who thinks advanced AI has a high probability of ending in catastrophe, at the extreme human extinction, and that its development should therefore be slowed, paused or heavily regulated (p(doom)). The word comes from internet culture (a 2018 4chan meme about hopeless young men) and, depending on who uses it, is a self-description or an insult. Its canonical recent text is If Anyone Builds It, Everyone Dies: if anyone, anywhere, builds a superintelligence with anything like current techniques, everyone dies (Yudkowsky & Soares, 2025).
Opposite them stand the effective accelerationists (e/acc), a movement born in 2022 in Substack newsletters and pseudonymous Twitter accounts such as @BasedBeffJezos, whom Forbes identified in 2023 as the physicist Guillaume Verdon. The name parodies effective altruism, the cradle of much doomerism. For e/acc, technical progress is almost a thermodynamic imperative and slowing it down is the real danger; Marc Andreessen gave it an investor's version in his Techno-Optimist Manifesto (2023). The song condenses it into one line: "And you're optimizing, accelerating" (osmarks, n.d.).
Santelli (2026) proposes calling the enthusiastic camp, broadly understood, hyperos (from hype, with a Spanish suffix): from those who expect a post-scarcity utopia to the "infantry" (infantería) of AI bros selling courses. Doomers and hyperos, he argues, share the premise that AI is either a god or a monster; against both, Narayanan and Kapoor propose treating it as a "normal technology," powerful but manageable with ordinary institutions (Narayanan & Kapoor, 2025). A third group is neither: the critics behind the stochastic parrot, for whom talk of extinction is one more way of inflating the hype, since an AI able to exterminate us would be all-powerful (Santelli, 2026). Buckner describes a similar pendulum in philosophy: depending on whom you ask, deep networks are on the verge of superintelligence or merely "bludgeon" problems with statistics (Buckner, 2024, p. 4).