By Robert Wiblin | Watch on Youtube | Listen on Spotify

As we now know, Anthropic has built an AI that can break into almost any computer on Earth. That AI has already found thousands of unknown security vulnerabilities in every major operating system and every major browser. And Anthropic has decided it's too dangerous to release to the public; it would just cause too much harm.

Here are just a few of the things that AI accomplished during testing:

• It found a 27-year-old flaw in the world's most security-hardened operating system that would in effect let it crash all kinds of essential infrastructure.
• Engineers at the company with no particular security training asked it to find vulnerabilities overnight and woke up to working exploits of critical security flaws that could be used to cause real harm.
• It managed to figure out how to build web pages that, when visited by fully updated, fully patched computers, would allow it to write to the operating system kernel — the most important and protected layer of any computer.

We know all this because Anthropic has released hundreds of pages of documentation about this model, which they’ve called Claude Mythos.

[...]

---

Outline:

(01:35) Why people are panicking about computer security

(05:12) Mythos could break out of containment

(07:22) Anthropic is losing billions in revenue by not releasing Mythos

(09:09) Mythos is actually the most aligned model to date, except...

(11:44) Mythos knows when its being tested

(13:50) Mythos can hide its thoughts

(16:24) Mythos cant be trusted about whether its untrustworthy

(20:00) Does Mythos advance automated AI R&D?

(22:28) Mythos scares Anthropic

(24:53) Learn more

---

First published:

April 10th, 2026

Source:

https://forum.effectivealtruism.org/posts/SyJx8Mbvi2ft78esn/how-scary-is-claude-mythos-303-pages-in-21-minutes

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Crosspost of a blog post.

A recent investigation uncovered that the leading insect farming company, Ynsect, was a horrifying disaster. While company leadership soaked up millions in subsidies with sunny rhetoric about their commercial success, the dark reality inside the farm was like nothing anyone imagined.

Investigators found filth clogging the machines, endangering workers. Larval bins would routinely break open, leading to repeated showers of larvae. Insects crawled all over the place, wholly uncontained. Airborne larval food floated haphazardly through the air, making workers sick. Birds wandered freely throughout the facility, their feces coating the machines. The whole area was a deluge of filth, feces, disease, and pollution.

The photos above show what the place looked like. In person, it was even more of a horror show. Investigators described1:

And here, in the photos and videos, you don't have the smells. In reality, it makes you dizzy. It's impossible to breathe. Even with a mask, it makes you feel sick.

The accounts from workers were even more harrowing.

There was no dust extraction, nothing at all. I started having a runny nose, itchy eyes, a scratchy throat, my nose running, my eyes burning, and difficulty breathing. Rashes on [...]

The original text contained 1 footnote which was omitted from this narration.

---

First published:

April 10th, 2026

Source:

https://forum.effectivealtruism.org/posts/m22GCyaKGd8gKBwFX/leading-insect-farm-took-millions-in-subsidies-then

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

I wrote a book! It's called Obsolete: The AI Industry's Trillion-Dollar Race to Replace You—and How to Stop It, and it’ll be available in May if you preorder through my publisher (OR Books+The Nation).

With it, I took on a dual mission: get anyone up to speed on the state of AI and its risks (especially skeptics and lefties) AND write something people working in the space will enjoy and learn from. An unbiased panel of early reviewers (i.e. people I know) think it succeeds. For instance, here's a review from a skeptical friend:

It felt like I was being drawn into this profound vision I had managed to keep out of my head for years… I feel this deep, visceral distrust of AI companies, and chatbots, for reasons that you were able to lay out in various places. I think I came to this like a very non-SF, non-tech, ‘normie’ who is reacting with their gut to seeing the proliferation of ai... But I also had never directly faced the story of AI, how the tech worked, and whether my skepticism was in fact, ill-founded. This hit that, and I thought it did a great job balancing [...]

---

Outline:

(06:37) Ways to help

(08:22) Excerpt - The Obsoleting Project

---

First published:

April 9th, 2026

Source:

https://forum.effectivealtruism.org/posts/ekShYQpaScoZLqQyi/help-me-launch-obsolete-a-book-aimed-at-building-a-new

---

Narrated by TYPE III AUDIO.

We're excited to share that we’ve overhauled our job board! It now has a much better interface and email alerts, to help people find impactful roles.

We’re still fixing a bunch of bugs, but we’ve opted to launch and iterate on the live version rather than wait until everything is perfect, because we think the board is now at a state where we expect this to be better for users. If you spot any issues or have suggestions, please let us know!

Moving forward, we’ll continue improving the board, though we’ll mainly shift our focus back to adding more roles. We’ve recently hit a bottleneck on this, because we’ve exceeded the limits of our previous (MVPish) systems as we’ve scaled the board. This should no longer be an issue, and we’re also planning to make further improvements to our backend, which mostly won’t be visible, but will help us continue to scale.

Ultimately, our goal is to continue improving and scaling the board, which has already grown more than 5x in the last 12 months compared to the previous ones (getting well over 100K clicks on impactful roles), and is still growing rapidly.

In the meantime, here are a [...]

---

First published:

April 8th, 2026

Source:

https://forum.effectivealtruism.org/posts/hZ8Xep8aHP5BeKQeL/launching-a-massively-improved-version-of-our-job-board

---

Narrated by TYPE III AUDIO.

This is a link post.

[Subtitle.] The utilitarians were right!

This is a crosspost for Infinite Dust Specks Are Worse Than One Torture by Bentham's Bulldog, which was originally published on Bentham's Newsletter on 20 March 2026. Bentham's Bulldog published a post 2 days later responding to comments on the 1st post.

1 Introduction

You hear the torture vs dust specks example come up a lot when discussing the alleged vices of utilitarianism. Emile Torres once wrote:

On another occasion, Yudkowsky argued that in a forced-choice situation you should prefer that a single person is tortured relentlessly for 50 years than for some unfathomable number of people to suffer the almost imperceptible discomfort of having a single speck of dust in their eyes. Just do the moral arithmetic — or, as he puts it, “Shut up and multiply!” Suffice it to say that most philosophers would vehemently object to this conclusion.

This is extremely misleading. If one bothers to read Eliezer's piece on the subject (rather than opportunistically scanning it for things that sound bad out of context and then gleefully spreading them across the internet), they will see he has arguments for biting the bullet. He doesn’t just instruct [...]

---

Outline:

(00:33) 1. Introduction

(02:26) 2. The opposite of love on the spectrum

(12:36) 3. Risk

(16:02) 4. The extremely simple argument

(18:32) 5. Scope neglect

(20:00) 6. Conclusion

---

First published:

April 7th, 2026

Source:

https://forum.effectivealtruism.org/posts/yRD24v9ebsC46xWgs/infinite-dust-specks-are-worse-than-one-torture

Linkpost URL:
https://benthams.substack.com/p/infinite-dust-specks-are-worse-than

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Note: This post was crossposted from Planned Obsolescence by the Forum team, with the author's permission. The author may not see or respond to comments on this post.

Subtitle: What can AI do on its own, and when?

The closer you think you are to your “powerful AI” milestone of choice, the more important it becomes to define it precisely. If your timelines are in the few-year range, changing the particular definition of powerful AI can easily double or triple your median forecast. Take my definition of “full automation of AI R&D” from my predictions post earlier this year:

[F]or whichever AI company is in the lead [at the relevant point in time], if you fired all its members of technical staff, its rate of technical progress on relevant benchmarks would be slowed down by less than 25% (this is a pretty arbitrary threshold, you can make the milestone more or less extreme by choosing a smaller or bigger number).

There's a lot that could be made more precise about this definition, but let's zoom in on that arbitrary quantitative threshold.

In every major sector of economic activity, the productivity hit you take from removing humans is and [...]

---

First published:

April 3rd, 2026

Source:

https://forum.effectivealtruism.org/posts/ff7rtk44cfpaFbpPT/six-milestones-for-ai-automation

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

I once ran the monitoring and evaluation department of a large nonprofit in rural Tanzania. We had 25 full-time data collectors who ran surveys with farmers on their crop yields and planting practices. My job was to oversee the team, analyze their data and produce reports with actionable recommendations.

The field team was incredibly productive and we were collecting tons of data. We were also producing very few actual reports and horrendously behind on everything. This seemed to me like a “time to work harder” problem but reading a business professor's novel about manufacturing helped me see it differently.

The correct solution was to tell my 25 field staff to stop working for two months.

The most engrossing book I've read since Harry Potter

The Goal by Eliyahu Goldratt is a novel about a manager given 6 months to turn around his factory or face closure, and the tactics he uses to get production up. It should not be good. It is very good. I’m not claiming it's great literature but it contains one of those ideas that, once seen, cannot be unseen.

A system moves at the speed of its slowest component, and nothing you do to [...]

---

Outline:

(00:55) The most engrossing book Ive read since Harry Potter

(03:54) Rory, the bottleneck

(07:09) Bottles as bottlenecks

(08:23) Some observations on bottlenecks

(09:39) Conclusion

---

First published:

April 8th, 2026

Source:

https://forum.effectivealtruism.org/posts/7excF9Zsv3gKvAdAZ/the-outsized-benefits-of-removing-bottlenecks-some-personal

---

Narrated by TYPE III AUDIO.

This is a link post.

Rethink Priorities has recently come out with the Digital Consciousness Model, a groundbreaking synthesis of the evidence for and against the consciousness of large language models and chickens. I think the Digital Consciousness Model is some of the most exciting work coming out of consciousness studies recently, and I want to walk you through what we’ve learned from it.

I’m going to start by talking about the model's findings. Then I’m going to get into the nitty-gritty details, which are kind of boring (I don’t want you guys to bounce off the post because you’re trying to understand Bayesian hierarchical models).

What We Learn From The Digital Consciousness Model

The Digital Consciousness Model doesn’t offer a new theory of consciousness or purport to say once and for all which entities are conscious. Instead, it organizes the information we already have about consciousness in a structured way that makes it easier for everyone to think more clearly about consciousness—regardless of their viewpoints about the matter.

On a very high level, the Digital Consciousness Model looks at thirteen “stances”, or ideas about what evidence we should use to figure out if an entity is conscious. Some of [...]

---

Outline:

(00:45) What We Learn From The Digital Consciousness Model

(11:41) How The Digital Consciousness Model Works

---

First published:

April 2nd, 2026

Source:

https://forum.effectivealtruism.org/posts/2T95JPJ6NzTCS63sd/summarizing-the-evidence-on-ai-and-chicken-consciousness

Linkpost URL:
https://thingofthings.substack.com/p/summarizing-the-evidence-on-ai-and

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

I sit down with Lee Jae-jeong, one of the 21 parliamentarians on South Korea's Foreign Affairs and Unification Committee. The National Assembly budget audit looms, and our conversation could impact the next year of Korea's health aid. I walk through a 5-minute presentation (stopping intermittently for translation) about Gavi's new malaria vaccines and why we at Coefficient Giving think so highly of them. She furrows her brow while she flips through the printed leave-behind I brought. What does she want to know?

I’ve had dozens of similar conversations with Japanese and Korean lawmakers as a representative of Coefficient Giving's Global Aid Policy program. Our team's goals are to grow the scale of wealthy countries’ development assistance, and direct this money to more cost-effective aid programs. Given our lean staffing model, we’re limited in the amount of direct advocacy we ourselves can do. However, after funding local aid advocacy groups in Japan and Korea, we learned that meeting with policymakers in person helps to build relationships, strengthen educational advocacy, and provide international validation for our grantees’ work.

At a time of deep cuts across major aid donors, it can be easy to imagine appropriators as heartless gatekeepers who withhold [...]

---

Outline:

(02:05) Geopolitics

(03:51) Values

(04:38) Pragmatism

---

First published:

April 6th, 2026

Source:

https://forum.effectivealtruism.org/posts/hPKBApDDPDzZEHhDg/what-questions-do-policymakers-ask-about-foreign-aid

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

TL;DR

Between 2022 and 2023, GiveWell and Open Philanthropy committed over $70 million to Evidence Action's Dispensers for Safe Water programme. An independent survey in 2025 found that in Uganda and Malawi the programme was reaching approximately 70% fewer people with detectable chlorine than its own monitoring had indicated (approximately 40% fewer in Kenya).

This piece argues that this case-study reflects a structural gap in how sophisticated funders allocate analytical rigour: extraordinary attention is paid to whether an intervention works under controlled conditions with far less to whether its implementation is actually delivering impact. I propose six practical steps funders can take to evaluate implementation fidelity with the same seriousness they apply to academic evidence, and I believe funders have a clear responsibility to do so.

Disclosure

I am the CEO of Fortify Health, an organisation that has received funding from GiveWell. I also sit on the Governance and Advisory Boards of organisations that have received funding from GiveWell. This article reflects my personal opinions and not those of Fortify Health or other associated organisations.

I shared a draft of this article with GiveWell and Evidence Action prior to publication. Both organisations generously shared comments and clarifications; however [...]

---

First published:

April 8th, 2026

Source:

https://forum.effectivealtruism.org/posts/XuNBJL3EHqhFKLEGn/impact-requires-implementation-what-dispensers-for-safe

---

Narrated by TYPE III AUDIO.