Yes. It could just be a coincidence. But I don't think it's just a coincidence. I think the lack of critical thinking and lack of good norms you need to let a cult take over your movement is a bad sign, and may predict things to come. Particularly if there isn't really an adequately large post-mortem, an attempt to reckon with such a catastrophic mistake, such a catastrophic failure, and learn deep lessons from it.

This risks going down a rabbit hole, but I think one of the systemic problems with the EA movement is a consistent failure to do good post-mortems. (And while we're at it, good pre-mortems would probably help too.) Many movements, organizations, groups, individuals, etc. experience big failures. Sometimes they learn from them and turn things around. Sometimes they don't. There are several instances such as the Leverage Research infiltration, the Manifest racism scandal, and even the collapse of FTX where I think EA had opportunities to learn hard lessons and do things better in the future, but didn't.

But, of course, I can't prove any of this. It's all just my opinion.

But longtermism doesn't necessarily include AI safety, since many advocates of AI safety are not longtermists. If you think there's a 50%+ chance of superintelligence within a decade and a 5%+ chance of human extinction if superintelligence is created, you don't have to be concerned at all with anything that might happen 1,000+ years from now to treat that as an urgent priority.

In 2021, Will MacAskill, who coined the term longtermism, defined longtermism like this:

Longtermism is the view that positively influencing the longterm future is a key moral priority of our time.

In the introduction to the 2025 anthology Essays on Longtermism, Hillary Greaves, Jacob Barrett, and David Thorstad cite that definition from Will MacAskill. They also characterize longtermism like this:

A cluster of ideas going under the label ‘longtermism’ hold that considerations of the far future—on timescales of thousands, millions, or even billions of years—are highly significant for today’s decision-making.

Not all AI safety advocates are longtermists. Some are concerned with what will happen within the next 100 years and don't really think or care that much about the future 1,000+ years from now.

To accurately break down the EA 1.0 vs. EA 2.0 distinction by cause area, it would have to be something convoluted like: global health and development + animal welfare vs. AI safety + longtermism.

The citations are inline. When a word or phrase is hyperlinked, that's a citation. You can open the links and see the evidence for yourself.

For example, I cited this comment from Oliver Habryka, who worked for the Centre for Effective Altruism during the 2010s:

I will again remind people that Leverage at some point had approximately succeeded at a corporate takeover of CEA, placing both the CEO and their second-in-command in the organization. They really were not very peripheral to EA, they were just covert about it.

I cited Zoe Curzi's account of her time at Leverage Research, which supports its characterization as a cult. I also cited the Centre for Effective Altruism's webpage which documents some of its historical relationship with Leverage Research, and notes that Leverage Research organized the EA Summit conferences and the Pareto Fellowship. An additional citation in the post is an EA Forum comment (with a reply corroborating it) that describes cult-like behaviour during the interview process for the Pareto Fellowship.

I've actually never heard anyone in EA either a) deny that Leverage Research is a cult or b) deny that it was deeply involved in EA and the CEA, although maybe some people do deny one or both of those things. I don't necessarily have my finger on the pulse.

By in EA's orbit, I mean, e.g., that if Stop AI's co-founder had committed a mass shooting at OpenAI (like he said he wanted to do) some of the Stop AI money — which the co-founder wanted to use to buy high-powered weapons and ammo, but was prevented from doing by other people at Stop AI — that bought those guns and those bullets that massacred people in the OpenAI offices might have been donated by people in the EA community who thought they were donating to an EA-aligned organization in an EA cause area (AI safety). For example, here's a post on the EA Forum that debates whether to donate to Stop AI. It ultimately decides against, but not for the reason that Stop AI is likely to be too extremist or might end up killing people.

The Zizians did not come from EA directly, but they came from the LessWrong community. There is so much overlap between EA and the LessWrong community these days that the distinction between the two is porous. The lead sentence of the Wikipedia article on the Zizians describes them like this:

The Zizians are an informal group of rationalists allegedly involved in six violent deaths in the United States, three in 2022 and three in 2025.

Whether that's "in EA's orbit" or what the phrase "in EA's orbit" means is something you can feel free to disagree with me on. Would it have been better if I said "adjacent to EA" or "only one or two degrees separated from EA"? The core point is that Leverage Research was not just a totally random, out-of-the-blue fluke. New groups with extreme views and violent behaviours are still not that far separated from EA. This is not normal and doesn't typically happen. 

Okay, so I'm advocating two things. The first is a new piece of terminology. The second is online discussion spaces oriented around EA 1.0 (and not EA 2.0).

If people can articulate the distinction better by having this terminology, it might mean people talk past each other less, which might mean they have fewer frustrating discussions where neither person feels like they're getting their point across. In that way, making the distinction could help people get along better rather than worse.

I'm totally against echo chambers, but the EA Forum is mostly an EA 2.0 echo chamber and that's unlikely to change anytime soon. I think there should be EA 1.0 spaces for discussion that make room for all the pro-EA 1.0 and anti-EA 2.0 conversations that can't happen on the EA Forum. You risking winding up in just another echo chamber if you do that, so it will be up to whoever gets involved to make sure that doesn't happen. And if people who prefer EA 1.0 want to engage with EA 2.0 discourse, the EA Forum will still be around for them to do it.

The philosopher David Thorstad has an exemplary post on peer review, with strong evidence and arguments for its effectiveness — both as a means of increasing research quality and as a means of persuasion of expert communities. 

Invitation for bets

I’m willing to bet that Anthropic’s revenue growth over the next year will be slower than its revenue growth over the last 3 years. I proposed a specific bet here. Anyone who wants can offer to take the other side of that bet. Or you can make a counteroffer.

I’m also willing to make a longer-term bet that the AI industry is in a bubble. I proposed a specific bet for that, too, here. Feel free to offer to take the other side of that bet or make a counteroffer.

I’d also be open to other bets. It seems pointless to bet about whether AGI or transformative AI will be deployed within the next 5-10 years, yet, for the heck of it, I would agree to a bet against that, too. (I’ll make bets for small, nominal amounts of money to be donated to the winner’s charity of choice, since the practical and legal problems with betting are too large otherwise.)

I’d also bet against the deployment of 100,000+ SAE Level 5 fully autonomous vehicles in North America within the next 3 years, if anyone has a strong opinion on that. I’d make a similar bet against the deployment of autonomous humanoid robots in North American households, although we’d have to come up with some specific resolution criteria.

Similarly, I’d bet against any significant level of near-term labour automation by LLMs or generative AI. Or against LLMs becoming capable of performing all sorts of specific tasks well.

On any of these topics, I’m also open to invitations for a public dialogue. (More on that topic here.)

Why did the EA organizations find it disappointing? I’m afraid you’re going to say they didn’t like that peer reviewers didn’t agree with them, and therefore they decided the peer reviewers were wrong, and peer review is a waste of time.

Not all EA organizations are consistently producing high-quality work. That’s part of the problem. For instance, the problems with the METR time horizons graph are numerous and severe. Many of them were entirely avoidable, and should have at least been better disclosed. I can’t get over that most of the longer tasks, on which the 2025 segment of the graph depends, don’t have empirically measured human baselines. The baselines are just guesses by the authors. Surely if you don’t even bother to measure data that doesn’t qualify as high-quality? This also wasn’t disclosed until 2026 — a major omission.

What I would recommend to people at this point is to not believe any of METR’s claims, research, or analysis going forward unless and until it can be independently verified by a reliable source. You don’t know if METR’s data is data or just guesstimates. You don’t know that the typical best practices of scientific research have been followed. You don’t know that flaws or shortcomings or limitations that METR is aware of will be disclosed with sufficient emphasis, consistently across all communications.

Very few people outside of EA consider EA’s idiosyncratic ideas to be serious and credible. What is the strategy for gaining credibility outside the EA echo chamber? Right now, it seems to be a media strategy that counts on people not fact checking EA’s messaging. This could work — a lot of misinformation misleads a lot of people a lot of the time — but it also might rightly damage EA’s reputation if people eventually learn EA is not telling them the truth. It’s a risky strategy that depends on being able to fool people, rather than intellectually convince them.

80,000 Hours’ abysmal video on AI 2027 is an example of this. It misinforms its audience about AI experts’ views and insinuates there is a consensus in support of AI 2027’s core claims that doesn’t exist. Either 80,000 Hours knew this and misled its audience anyway, or it didn’t do a proper fact check of its script before producing the video. I was a lifelong fan of 80,000 Hours until that video. Now I no longer trust 80,000 Hours about anything. Not even career advice. I was in the top 1% or 0.1% of biggest supporters of 80,000 Hours. Now I’ve been completely polarized in the opposite direction. This is anecdotal, but, also, most people become angry when they feel as if they’ve been misled. It’s not a stretch to think this strategy could really blow up in an ugly way.

In my opinion, EA is aggressively burning down its reputation and risks being correctly labelled as a purveyor of misinformation. Steps should be taken to at least stop the bleeding.

I don’t know for sure that peer review would help move idiosyncratic EA ideas outside the EA echo chamber. I also don’t know that there isn’t a better strategy for doing so. It just seems like a good idea to me.

The economist Tyler Cowen was actually the first person who I heard suggest this. I believe he was talking about AGI/AGI safety. It was on a podcast, either his or someone else’s. I remember he said: publish, publish, publish.

The provider I originally mentioned in this post definitely looks like a shady company that I definitely wouldn’t recommend. I was wrong to mention that company and, in retrospect, the signs were obvious that it wasn’t a trustworthy company. I only gave it a few cursory glances. I’m grateful to Clara for giving it a second look and realizing that both Google Gemini 3.1 Pro (with “Extended thinking”) and I had been duped by some devious SEO.

The trustworthiness of that provider — or indeed any similar company offering a convenient, off-the-shelf service — is beside the point of whether peer review is a good idea or not. There are scam companies selling fake Ozempic online. That has nothing to do with whether genuine Ozempic is a good drug or not.

Reuters recently published a fantastic article on Tesla’s efforts to deploy robotaxis: https://www.reuters.com/investigations/why-teslas-ai-trainers-dont-trust-its-self-driving-tech-or-its-safety-stats-2026-05-28/

The article argues that Tesla’s actual technical progress is far behind what the company claims.

I sense this post has insightful and important things to say, but it’s written in a way that’s so abstract that I find it hard to follow. I’ve read several posts like this on the EA Forum. Posts like this desperately need lots of concrete, specific examples described in plain English. The best would be telling stories (anecdotes) that illustrate a problem you’re trying to talk about.

This part I think I can understand and agree with:

EA community building seems less about helping people have an impact by their own lights, and more about having an impact by the lights of the community builders.

Concretely, I infer this means whether it’s 80,000 Hours’ career advising, EA groups at universities doing outreach, interactions at EA conferences, or whatever else, the EA veterans are ultimately just trying to evangelize to the EA newbies. Which is not what the EA newbies want them to do, nor what the EA veterans are leading the newbies to believe is going to happen. Am I right?

This too:

FTX didn't just happen to EA - it also happened within EA. SBF was supported, celebrated and elevated by many parts of EA leadership. The damage wasn't primarily to EA's brand; it was to the trust people had placed in EA - and that loss of trust was at least partially warranted.

To quote the writer Matt Yglesias:

I understand the various reasons why people can't speak in detail about certain things would I would love to see the issues discussed in "EA and the current funding situation" get revisited in light of the fact that the funding situation has changed a lot. It seems to me that when money became more abundant that didn't just make it easier for certain grant proposals to get approved, but was seen as having big-picture strategic implications. When I go back and read the post, the arguments seem sound to me. But beyond not reflecting knowledge of wrongdoing at FTX (hard to find blame here) they seem to not reflect awareness of Bankman-Fried's overall attitude toward risk which I think should have been knowable. 

I’d add it wasn’t just Sam Bankman-Fried’s attitude toward risk that was knowable. It was also the riskiness of crypto. Shouldn’t people have been expecting there was a decent chance FTX would blow up, even without any financial crimes?

Besides community building and FTX, the third major example in this post is the Centre for Effective Altruism’s communications strategy or brand strategy. The Centre for Effective Altruism describes this in general, abstract terms that leave me guessing what they mean, concretely and specifically. This post responds to the CEA’s description of its strategy in terms that are also abstract and general. 

Who is right and who is wrong? What is the disagreement even about? Is there an actual disagreement or are people just talking past each other? I don’t know. We’d need specific examples of what it would look like to have “honest and positive” messaging versus neutral and purely informative messaging.

I do agree that it feels like EA is dying, but, for all I know, my reasons for feeling that way might be the exact opposite of yours.