AI doomer or bloomer? Most people land in between

The AI debate has its own vocabulary, and most of it sorts people into teams: doomers and bloomers, accelerationists and decels. Search for one of these words and you mostly find threads, memes and arguments about who belongs where.

This post is the plain version. For each word, it gives a short definition, where the word came from, and whether it’s mostly used as a put-down. It also makes one argument: the words mix up separate questions, mainly how well AI turns out and how much it changes the world, and most people land somewhere between the ends of both.

We build Doom or Bloom, a short interview that places your view of AI’s future on those two questions, so that’s the map we’ll use.

The words on one map

Across, the map runs from Doom to Bloom: how well or badly someone expects AI to turn out for people. Up and down, it runs from incremental to civilizational change: how much they expect AI to change the world, for better or worse.

Placed this way, the labels crowd the edges. AI doomer and x-risk sit in the top left corner, e/acc and zoomer in the top right. The words for what AI might become, AGI, superintelligence and the singularity, sit on the center line, because they say how much changes, not whether that’s good. Only gloomer and skeptic sit in the bottom half: the debate has few words for people who expect AI to change less than the headlines say, and none at all for the bottom right, hopeful about AI but expecting modest change.

Some common words aren’t on the map, because they answer other questions. Foom and fast takeoff are about how quickly AI improves. Decel and d/acc are about what we should do. They’re covered below.

Doomer, gloomer, zoomer, bloomer

What is an AI doomer?

An AI doomer is someone who thinks advanced AI is likely to end in catastrophe for humanity, such as human extinction or a permanent loss of control over our future.

“Doomer” is older than the AI debate. It was used for people expecting peak oil or climate collapse, and from 2018 for a hopeless twentysomething in a popular meme.⁠^[1] Applied to AI, it’s mostly a label from the outside. aisafety.info, a site that explains AI risk, notes that it’s often used pejoratively,⁠^[2] and Palladium says it was coined by the view’s opponents, though “some have adopted it themselves for lack of a better alternative”.⁠^[3] Many of the people it’s applied to would rather be called advocates for AI safety.⁠^[4]

The word bundles two beliefs: that AI will become extremely powerful, and that this will go badly for us. You can hold either one without the other.

What is an AI bloomer?

An AI bloomer is someone who is hopeful about AI and wants to push ahead while managing its risks. In AI, the word comes from Superagency (2025), a book by Reid Hoffman and Greg Beato that sorts reactions to AI into four groups. Hoffman summed them up on X:⁠^[5]

• Doomers: “Won’t stop predicting an AGI-becomes-terminator future”
• Gloomers: “Think we’re worse off with AI”
• Zoomers: “Accelerate at all costs w/o considering risk”
• Bloomers: “Accelerate towards a bright future, manage risk”

These are the definitions of someone who counts himself a bloomer: in an interview with TechCrunch, Hoffman describes a bloomer as “inherently technology optimistic”.⁠^[6] So read his other three with that in mind. Before the book, “bloomer” was a meme character too, the cheerful, go-getting opposite of the hopeless doomer.⁠^[7]

Gloomers and zoomers

The other two groups fill in what doomer and bloomer leave out. A gloomer thinks we’ll be worse off with AI, without necessarily expecting catastrophe. A zoomer wants AI built as fast as possible, without weighing the risks, and has nothing to do with Gen Z.

The part of the map closest to the gloomer view, worried but not expecting catastrophe, is where the largest share of our participants land.

Most people land in between

So how many people are doomers or bloomers? Among everyone who had finished Doom or Bloom’s interview as of October 3, 2026, few land at either end of either axis.

About one in eight sit at the doom end of the outlook axis, the part of the map closest to what “doomer” describes, and about one in twelve at the bloom end. The other four in five land in between, and the largest group, more than a third, leans toward doom without reaching the end. Scale works the same way. About one in eight sit at the top edge, where AI changes civilization itself; more than a quarter sit in the lower half, where its effects stay modest; most land in between.

These are people who chose to take an AI interview, most of them from two launch-week crowds that landed far apart on the map, so they aren’t a sample of the public. But polls of the public point the same way. When a poll offers a middle answer, many people take it: 37% of Americans are equally concerned and excited about AI,⁠^[8] and 52% say it does equal amounts of harm and good.⁠^[9] Anthropic’s 81,000 AI-led interviews found that hope and alarm “didn’t divide people into camps, so much as coexist as tensions within each person”.⁠^[10]

What’s the opposite of an AI doomer?

It depends which part of “doomer” you flip.

• Flip the outcome and you get an optimist: in Hoffman’s terms, a bloomer. Wiktionary gives “boomer” as the opposite, a variant you’ll also see⁠^[1]
• Flip the stakes and you get a skeptic, who expects AI to change the world far less than doomers and optimists both assume
• Change the question and you get an accelerationist. Doomers and accelerationists are often set against each other, but one word is about how AI turns out and the other is about what to do

Here are the labels rated against four questions they might answer: how well AI turns out, how much it changes the world, how quickly it improves, and what we should do about it.

Read across the rows and the familiar pairings stop lining up. e/acc supporters often call their opponents doomers or decels,⁠^[11] but the two words answer different questions: one is about how AI turns out, the other only about what to do. Someone can expect catastrophe and doubt that slowing down would help, or expect a bright future and still want a careful path to it. The real opposites answer the same question: doomer and bloomer on how AI turns out, doomer and skeptic on how much it changes, e/acc and decel on what to do.

Accelerationists and decels

What does e/acc mean?

e/acc, short for effective accelerationism, is the view that technological progress should be sped up, not slowed, because growth driven by technology and markets is good and can’t be stopped anyway. It was named in a July 2022 newsletter post by two pseudonymous X accounts, Beff Jezos and bayeslord, which sums up the idea in a few lines, among them “You cannot stop the acceleration” and “You might as well embrace it”.⁠^[12]

In December 2023, Guillaume Verdon confirmed to Forbes that he is behind the Beff Jezos account.⁠^[13]

What is a decel?

Decel, short for decelerationist, is what e/acc supporters call people who want to slow technology down, from AI safety advocates to regulators.⁠^[13] It’s a put-down: Wikipedia notes that e/acc supporters “often label their opponents derogatorily as ‘doomers’ or ‘decels’”.⁠^[11] The word says nothing about how someone expects AI to turn out, only that they want it to arrive more slowly.

e/acc vs d/acc

d/acc is Vitalik Buterin’s alternative to e/acc, from his November 2023 essay “My techno-optimism”. He writes that the “d” can stand for many things, “particularly, defense, decentralization, democracy and differential”: speed up technology, especially the kinds that help people defend themselves and keep power spread out. To e/acc readers, he describes d/acc as “a subspecies of e/acc”, one that is “much more selective and intentional”.⁠^[14]

e/acc vs EA

The “effective” in e/acc comes from effective altruism (EA), a movement that aims to find the best ways to help others and put them into practice, and counts helping to create the field of AI alignment research among its projects.⁠^[15] The e/acc founding post presented itself as “a contrast to anti-AGI factions of EA”.⁠^[12] So the names sound alike, but e/acc began partly as a reaction against EA.⁠^[13]

How fast and how far

Foom and fast takeoff

Foom is slang for an AI that improves itself so quickly that it goes from about human level to superintelligence in a matter of minutes, days or months.⁠^[16] The word took off with a 2008 online debate between the economist Robin Hanson and the AI theorist Eliezer Yudkowsky about whether a generally intelligent AI could improve its own capabilities very quickly;⁠^[17] one of Hanson’s posts in it is titled “AI Go Foom”.⁠^[18] The idea is older: in 1965, the statistician I. J. Good called it an “intelligence explosion”.⁠^[19]

Fast takeoff, or hard takeoff, is the less playful name for the same idea. In Nick Bostrom’s 2014 book Superintelligence, a fast takeoff happens in minutes to days and a slow one over decades or centuries, as Katja Grace summarizes it.⁠^[20] Paul Christiano framed the argument as whether AGI looks more like “a breakthrough within a small group” or “a continuous acceleration distributed across the broader economy”.⁠^[21]

Neither word says whether the result is good, and neither is about how much AI eventually changes the world, only how fast it gets there. That’s why they’re not on the map.

What is AGI?

AGI, artificial general intelligence, is AI that can do most of what people can. The details depend on who you ask. OpenAI’s charter defines it as “highly autonomous systems that outperform humans at most economically valuable work”,⁠^[22] while a 2023 Google DeepMind paper describes AI “at least as capable as a human at most tasks” and proposes levels of AGI rather than a single line.⁠^[23] When people argue about whether AGI is here, they’re often using different definitions.

Superintelligence and the singularity

Superintelligence is AI that far surpasses people: in Nick Bostrom’s 1998 definition, an intellect “much smarter than the best human brains in practically every field”.⁠^[24]

The singularity is the point where superhuman intelligence changes the world so fast that what comes after can’t be predicted from here. The science fiction writer Vernor Vinge popularized it in 1993, writing that within thirty years we would have the means to create superhuman intelligence, and “Shortly after, the human era will be ended”.⁠^[25]

How bad it could get

X-risk and existential risk

X-risk is short for existential risk: in Nick Bostrom’s 2002 definition, a risk whose adverse outcome would “annihilate Earth-originating intelligent life or permanently and drastically curtail its potential”.⁠^[26] AI is one candidate among several, alongside risks like engineered pandemics and nuclear war. “AI x-risk” is the chance that AI causes one.

AI alignment

AI alignment is the work of getting AI systems to do what people intend. In Wikipedia’s words, it “aims to steer AI systems toward a person’s or group’s intended goals, preferences, or ethical principles”.⁠^[27] Much of the x-risk debate is about whether alignment can keep up as AI gets more capable.

P(doom)

P(doom) is the probability a person assigns to advanced AI causing an existential catastrophe for humanity, such as human extinction or a permanent loss of control over our future.

P(doom) puts a number on the doomer question, but it isn’t only for doomers: people all across the map give one, and the same number can mean very different things. For the numbers thought leaders have stated, with sources, see the P(doom) page.

Ask, don’t label

Labels are handy shorthand inside a group and blunt outside one. Several of the words above were coined by one side to describe the other, and the people they get pinned on often describe themselves differently.

Two questions get further than any label: how well do you expect AI to turn out, and how much do you expect it to change the world? Then a third, if you like: what’s your P(doom)? That’s what Doom or Bloom asks. Take the interview (about 3 minutes, no sign-up) to see where you land, and which simulated thought leaders land near you.

How we put this together

• Definitions: each from its original source where we could find one, read on October 6, 2026, with quotation marks only around exact words. For the takeoff speeds in Bostrom’s Superintelligence, we used a published summary rather than the book
• Placement: where each word sits on the map, and how the chart rates the questions it answers, is our editorial reading of its definition, not a measurement. The map places words, never people
• Participant numbers: Doom or Bloom’s production aggregates as of October 3, 2026: the earliest result of each person, leaving out simulated users, forks and our own test accounts, with every number describing at least 10 people. The rows are ranges of map position, not survey answers
• Polls: Pew’s June 2026 survey and Gallup’s 2026 survey of US adults, as reported in Why polls on AI disagree
• Thought leaders: the people on Doom or Bloom are simulated from their public writing and interviews