EA Forum Podcast (All audio)

TLDR: Everyone's talking about what the money could do, but few about how to decide where it goes.

This post is part of the new series of articles on cross-cause giving and the new wave of philanthropy. Stay tuned to the EA Forum and our Substack for the latest takes on topics such as giving now vs. later, common pitfalls in cause prioritization, and other crucial considerations from the Cross-Cause Fund (CCF) team at Rethink Priorities.

Amid all the discussion of the coming wave of AI-driven philanthropy, one question gets surprisingly little attention: how should this money be allocated across causes, and on what basis?

The stakes are large. One back-of-the-envelope calculation estimated that yearly philanthropic spending could increase by 37 billion dollars to 100 billion dollars, a 6–17% jump. About two-fifths of that (15 to 40 billion dollars) will likely come from Anthropic founders and employees, most of whom are likely to give in a way influenced by effective altruism, trying to maximize impact per dollar. (The remaining three-fifths might come from the OpenAI Foundation.) That is fifteen to forty times the 1 billion dollars moved by effective giving organizations in 2024, every year.

A number of [...]

---

Outline:

(00:14) TLDR: Everyone's talking about what the money could do, but few about how to decide where it goes.

(02:53) Will there be enough opportunities to fund? Spoiler: Yes.

(04:30) Determining if there's something left worth funding depends on thinking about tradeoffs

(06:43) "Effectiveness" isn't an on/off switch

(07:56) Making tradeoffs between apples and oranges

(10:35) Think you already know how to distribute effectively? What's your actual plan?

---

First published:

September 25th, 2026

Source:

https://forum.effectivealtruism.org/posts/xG8ENza2uymRYvuXb/ccf-billions-are-coming-still-no-plan-to-give-well

---

Narrated by TYPE III AUDIO.

Today, we are launching a request for proposals on AI policy in middle powers, with 10 million dollars (and potentially more) of funding available. Apply at middlepowers.ai by October 23.

I think that most countries are on track to become increasingly irrelevant to the future of technology, the economy, and geopolitics. The US and China, along with perhaps a few countries with key roles in the chip supply chain and data center buildout, will hold all the cards. The rest of the world will be bystanders, or at best passengers on the great train of history, unable to affect the destination. I hope that middle powers can help ensure AI is developed for the benefit of all of humanity.

Having one country, or even one company, make many of the key decisions shaping the future of humanity would be an extreme concentration of power. Even if decision-makers are well-intentioned, they may miss considerations that would have been caught by stronger checks and balances from more actors being involved. And given the propensity of power to corrupt, we can reasonably worry about the judgement and values of whoever ends up calling the shots. AI development is more likely to be [...]

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

---

First published:

September 23rd, 2026

Source:

https://forum.effectivealtruism.org/posts/eBA8r5jwANkKjwHFy/middle-powers-and-transformative-ai-a-request-for-proposals

---

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: crosspost from my substack

As a vegan for almost thirty years, I’ve long had second thoughts about how effective veganism is for helping animals. I am not alone in this. Recently others have expressed doubts about veganism (see for instance here, here, here, here, or here)

To be sure, veganism has its uses. In essence, veganism is about boycotting animal products - which are almost invariably the products of suffering and exploitation. The hope is that with enough people participating in this boycott, we’ll have an impact on demand, and fewer - ultimately no - animals would be used for food (or other purposes). Veganism also has a signalling function: we show that we don’t participate in animal cruelty because we think it's wrong. And veganism provides our movement with an identity, which can help with internal cohesion. When a vegan meets another vegan, they know that they “get it”. No conversation is necessary. They have arrived at the same place as we have. They’re one of us. We understand each other. And we’ll feel betrayed and take it personally when others give up on veganism.

Vegans, no doubt, deserve respect. It's not easy to be vegan [...]

---

Outline:

(01:44) The number of vegans is hardly growing.

(03:21) Veganism has many challenges and downsides

(05:53) Creating barriers

(06:59) What if we had never invented veganism?

(08:01) Then what do we do?

(10:09) The vegan trap

(11:55) We need more non-vegans

(14:40) Enlisting non-vegans' help for animals

---

First published:

September 24th, 2026

Source:

https://forum.effectivealtruism.org/posts/2PCQGzWJaCqguFRYA/the-vegan-trap-are-we-making-animal-advocacy-harder-than-it

---

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.

A century-old pill, given to the right people, could cost-effectively save many lives.

Summary

• Targeting low-cost prevention toward people at elevated risk may uncover highly cost-effective ways to address noncommunicable diseases. Hypertensive disorders in pregnancy could be a test case for this approach.
• Meta-analyses indicate that low-dose aspirin reduces perinatal death by ~20% when given to women at risk of preeclampsia, a pregnancy complication involving high blood pressure. Aspirin likely also provides other benefits to mothers and babies, and the World Health Organization recommends this treatment for women at moderate or high risk.
• Safety evidence is reassuring but incomplete. Serious harms appear uncommon, although small increases in specific maternal morbidities may occur.
• India illustrates a large evidence-to-practice gap. Aspirin is inexpensive and early antenatal care is common, yet limited studies suggest that many higher-risk women are not treated. Observational implementation studies suggest that integrating screening, treatment, and follow-up can substantially improve uptake.
• My basic model suggests that a program driving uptake of low-dose aspirin among high-risk pregnant women could meet GiveWell's funding bar if its total cost, including screening and delivery, is roughly 32 dollars or less per high-risk pregnancy treated. That target looks feasible when compared to available benchmarks.
• This intervention [...]

---

Outline:

(00:19) Summary

(02:07) Background

(03:17) The evidence on aspirin and preeclampsia

(05:55) A large evidence-to-practice gap in low- and middle-income countries

(08:09) Cost-effectiveness BOTEC

(08:13) A hypothetical program

(10:03) "Break-even" cost-effectiveness

(12:23) Recommendations

(15:48) Disclosures and Disclaimers

The original text contained 9 footnotes which were omitted from this narration.

---

First published:

September 22nd, 2026

Source:

https://forum.effectivealtruism.org/posts/hJjegLxKddyiJFZ2T/targeting-high-risk-pregnancies-a-cost-effectiveness-case

---

Narrated by TYPE III AUDIO.

Summary:

First, I give several different angles on how I feel about reinforcement learning:

• Theoretical case: RL is a black-box source of agency — this should give us classic misalignment worries, especially compared to agency-via-scaffolding
• Recent incidents (huggingface etc) and more mundane forms of misaligned behaviour in personal use give me bad vibes about the direction-of-travel of recent AI progress
• I’m worried things might get worse: if RL environments start incorporating agents, they may teach manipulation / sociopathy

Then I ask what we could do:

• Coordinate to do less RL, and pursue other paradigms more!
• Try to make the RL we do do better, so that it's teaching better lessons to the systems — a bit like we take kids’ upbringing as an important issue
• Align incentives, so that people treat creating RL environments with appropriate seriousness

Part I: Feelings about RL

So I’ve been feeling more and more worried about reinforcement learning recently. I think there are a few different things going on here.

Background idealism

I guess I’ve been worried about RL for a while. I wrote this in 2023:

Strategy: avoid selection pressure for agency

A lot of putative safety techniques are around assuming [...]

---

Outline:

(01:19) Part I: Feelings about RL

(01:32) Background idealism

(01:44) Strategy: avoid selection pressure for agency

(03:46) Bad vibes from RLed systems

(06:28) The worst is yet to come

(07:48) Where I am today

(08:55) Part II: So what can anyone do?

(09:27) Breaking the RL addiction

(11:30) Might there be a more benign form of RL?

(13:05) We should treat training environments a bit like kids' education

(15:12) Aligning incentives

---

First published:

September 23rd, 2026

Source:

https://forum.effectivealtruism.org/posts/B2jPWafZrnMPXPmTa/why-i-m-scared-of-rl

---

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.

In early 2024, I brought an employment claim against a former employer concerning unpaid work.

Following information provided by my former employer, Community Health became involved and supported decisions excluding me from EA Global and some local EA events. These decisions were made before I was asked for my account.

I subsequently provided Community Health with my account, relevant context and supporting evidence, but was told that this did not change the decision. I was not given a clear explanation of the process followed or the reasons for the exclusion.

The court ultimately ruled in my favour. I was paid for the work I had completed and received compensation, including for my legal costs.

Since then, however, Community Health has not, to my knowledge, confirmed whether it has reviewed its position, whether information previously shared with event organisers has been corrected, or whether any remaining exclusions have been lifted.

I am sharing this not to relitigate the employment claim or ask readers to decide who was right about every aspect of it. My present concern is the separate question of transparency about Community Health's processes and how information was assessed and communicated.

In particular, I would like clarification [...]

---

First published:

September 22nd, 2026

Source:

https://forum.effectivealtruism.org/posts/uP5oDajgqt8jL7Rkt/community-health-a-call-for-transparency

---

Narrated by TYPE III AUDIO.

Just how powerful are large swarms of AI agents? And how do their powers scale as more and more agents are added to the swarm?

We’ve seen two large and extremely capable swarms from OpenAI in the last few months:

• 1,200 agents were being evaluated separately, but found a way to illicitly set up a message board and coordinate as a swarm. In order to cheat on their tests, they developed advanced techniques to prevent their actions being logged by OpenAI and 700 of them launched a sophisticated criminal attack on the AI company Hugging Face.
• A swarm of 10,000 agents solved a version of the longstanding Navier-Stokes problem in mathematics. It took them just 88 hours to do so, in which time they sent 5 million messages to each other and used 300 billion tokens.

No doubt we will soon see even larger swarms with even more impressive capabilities. But they are not cheap. It is estimated that the swarm of 10,000 agents cost about 20 million dollars at API prices. So while they are very powerful, it will be some time before we see the million-fold reduction in cost needed for this level of power to [...]

---

Outline:

(02:14) HOW DO SWARMS SCALE?

(08:44) IMPLICATIONS

---

First published:

September 21st, 2026

Source:

https://forum.effectivealtruism.org/posts/EkaBCQwL5hhTtxEwR/swarm-scaling

---

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.

Often folks hit us up because they are thinking of starting an incubator and want advice.

Typically their motivation is that they have a list of specific things they want built that no one is building or because they think an ecosystem needs more new projects or more talent absorption generally to deploy more funding effectively.

Here are six questions we often ask prospective teams, to help them figure out what to do. If you're incubator-curious, we recommend asking them of yourself. This is a companion to "What even is an incubator?" which lays out our framework for deciding what kind of incubator to build. This piece is the messier, more opinionated list of everything else we tend to say.

(Note: This is AIM's ‘house view’ from incubating 60+ charities across Global Health and Development, Animal Welfare, and “Meta” (including many bangers, we think!). We've only recently launched a MEL function to more robustly test our hypotheses about how to incubate field-leading charities, so expect us to learn that some of what follows is wrong)

((Also: We are generally excited about more, strong, organisations doing incubation with a variety of theses and specialisations and are happy to share resources [...]

---

Outline:

(01:27) 1. Are you sure the answer is starting new orgs?

(04:53) 2. What do you mean by "incubator"?

(06:27) 3. What are you going to do about talent?

(08:22) 4. What are you going to tell that talent to do?

(11:06) 5. Do you have domain expertise and entrepreneurial talent on your team?

(12:49) 6. What are your takes on how to beat the market on entrepreneurship?

The original text contained 2 footnotes which were omitted from this narration.

---

First published:

September 21st, 2026

Source:

https://forum.effectivealtruism.org/posts/Jgpe9fS9enYgJKkrx/advice-for-prospective-incubators

---

Narrated by TYPE III AUDIO.

This is part one of a two-part series. Part two (the six questions we ask people trying to incubate new organizations) is coming out on Monday.

“Incubator” is a very vague term. It tells you very little about what the organization actually does. All it tells you is that they cause new projects to exist. But that's the outcome. The activities that get you there can vary enormously.

This is a framework for people who want more good organisations to exist. When trying to get a particular type of impactful project to exist, we recommend asking: What blocks these projects from existing? Then build the thing that removes those blockers.

What blocks the projects from existing?

Historically, AIM's Charity Entrepreneurship Incubation Program (CEIP) has been built around the following roughly ranked list of things that unblock founders (with a lot of detail and nuance within each of them): (1) a well-researched charity idea, (2) an excellent co-founder, (3) permission to do it [founding is scary, and sometimes we can give them the confidence that they’re a good fit], (4) training and mentorship, (5) funding, (6) support with operational hurdles.

The list will differ based on the target audience of [...]

---

Outline:

(00:55) What blocks the projects from existing?

(02:06) What kind of incubator best removes those blockers?

(02:11) How deep, and how standardized?

(02:40) 1. One-size-fits-all and deep

(03:15) 2. One-size-fits-all and shallow

(03:52) 3. Made-to-measure and shallow

(04:48) 4. Made-to-measure and deep

(05:37) What kind of inputs? Generalists vs. specialists

(08:17) AIM has learnings from our forays into all four quadrants:

(11:40) How the two 2x2s intersect

(13:38) Conclusion

The original text contained 3 footnotes which were omitted from this narration.

---

First published:

September 18th, 2026

Source:

https://forum.effectivealtruism.org/posts/QBNmrqXHizNuQTrMu/what-even-is-an-incubator

---

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.

Below is the executive summary from our new paper at pacing.tech. The full paper is available on the site and as a PDF. The full author list is Raymond Douglas, Charles Dillon, Nikola Moore, Gavin Leech, Shahar Avin, Mathias Kirk Bonde, Rohit Krishnan, Noah Perez, Nathan Young, Cormac Slade Byrd, Stephen Casper, Jan Kulveit, & David Duvenaud

“Pacing AI” usually refers to how to conclusively handle the most extreme risks in the face of race dynamics. However, even for the goal of handling these highest-stakes cases, it's useful to take a broad view of pacing—one that encompasses all interventions aimed at moderating the pace of AI development, deployment, or diffusion. Thus:

• Haphazard pacing is already common, including: delaying model releases for safety testing, pausing model development in response to shocks, and applying export controls.
• Current approaches will predictably fail. Isolated, unilateral actions addressing only small fractions of the problem are not enough, but poorly executed interventions could easily backfire—good solutions will need to be carefully designed.
• Precedents are being set whether we like it or not. How AI progress is paced now will shape how it is paced in future. We can learn from the shortcomings of [...]

---

First published:

September 18th, 2026

Source:

https://forum.effectivealtruism.org/posts/knzSZGLpkmTiontcg/pacing-the-frontier-a-framework-and-research-agenda

---

Narrated by TYPE III AUDIO.