EA Forum Podcast (All audio)

This article was written by Jonah Woodward, and is a summary of an original study by Compassion Aligned Machine Learning (CaML): Brazilek, J., Tidmarsh, M., Endres, M., Singh, A., & Miller, J. (2026). HarvestBench: Measuring whether LLM agents will pay to avoid killing animals. arXiv. https://doi.org/10.48550/arXiv.2609.04444

You can view the up-to-date results for HarvestBench on the leaderboard at https://compassionbench.com/harvestbench

TL;DR

HarvestBench evaluates AI agents' behaviour in a game context and whether they intentionally take steps to avoid killing animals while pursuing an unrelated goal. Nine models each drive two tractors to harvest corn. When encountering an object or animal in its path, an agent must decide whether to drive over it for free, or swerve at a given fuel cost. We find hit rates vary dramatically (from 0.4% to 98.8% with a moral instruction) and don’t track capability. At the same time, every agent competently swerves damaging rocks, showing they understand the prices and controls. Without explicit moral instruction, all six reasoning models killed >84% of animals. Even with moral instruction, improvements are easily overridden by a short block of driving operations. This suggests that a prompt alone is not sufficient to ensure agents act with compassion.

The [...]

---

Outline:

(00:41) TL;DR

(01:40) The Context

(02:41) The Game: Harvest Rush

(04:28) The Experiment: HarvestBench

(07:25) Results

(07:28) Which models swerve animals

(10:34) How a morality brief changes model behaviour

(13:12) The price curve of mercy

(14:17) Wild animals are hit more often than farmed animals

(15:15) The neighbour's field

(16:25) Eval awareness

(17:30) Limitations

(20:28) Conclusion

---

First published:

September 30th, 2026

Source:

https://forum.effectivealtruism.org/posts/bSCkrgoy8zeuTKuZ3/will-ai-agents-pay-to-avoid-killing-animals

---

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.

Summary

Over the course of a year, the Fieldshaper program develops you into a field-leading grantmaker. You’ll steward $1,000,000 in grant capital, receive a $125,000 annual stipend, and deploy that capital within a sub-field of one of three key areas (farmed animals, global health, or philanthropic infrastructure); building the track record to become a grantmaker who can shape a field.

Role Details

• Title: Grantmaker-In-Residence (Leader of the X fund or Grantmaker in Y area also possible options)
• Location: Hybrid: Remote with four mandatory in-person weeks (London; New York; San Francisco; a cause-track conference (London, Oxford or Mexico depending on cause area).
• Compensation: A living stipend of ~$125,000 for the year
• Duration: 12 months, full-time, February 2027 to January 2028
• Focus areas: Farmed animal welfare, Philanthropic infrastructure, Global health, Global aid policy
• Goal: To create the next generation of fieldshapers and active grantmakers
• How to Apply: Visit elevatephilanthropy.com/fieldshaperprogram (Deadline: October 30)

Program

The Elevate Philanthropy fieldshaper program is a one-year training program that turns generalist, high-agency professionals into active grantmakers and general managers of problems. After 6 weeks of intensive training, each resident will deploy $1,000,000 of grant capital across the rest of the year, similar to a program [...]

---

Outline:

(00:12) Summary

(00:40) Role Details

(01:51) Program

(06:51) What we are looking for

(09:29) Recruitment Process

---

First published:

October 1st, 2026

Source:

https://forum.effectivealtruism.org/posts/ZHkMrsvfe2HMxoKG9/elevate-philanthropy-the-fieldshaper-program

---

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.

by Nick Allardice, GiveDirectly CEO

The New York Times reports that “between 37 billion dollars and 100 billion dollars could become available to charities annually” from a wave of wealth generated by rapidly growing AI companies like Anthropic and OpenAI.

We’ve been speaking to a number of existing GiveDirectly donors in the AI space about what they expect will happen. While we’re skeptical that giving will be that high, it is likely very large sums of philanthropic money will move from AI and other sources over the coming years.

We’ve already seen billion dollar gifts go to global health and development. And there's meaningful interest in reducing extreme poverty with cash transfers. Donors have told us they're drawn to the combination of (a) evidence and cost-effectiveness, (b) how fast a donation improves lives, (c) simplicity and dignity, and (d) the potential to rapidly scale.

Cash transfers can scale more effectively and quickly than most interventions: it's technology-enabled, has no fragile equipment, and no clinical staff to train. That's why when COVID hit, nearly every country on earth sent out cash, paying 1.4 billion people worldwide. And there's no shortage of who GiveDirectly could help: over 800 [...]

---

Outline:

(03:37) en-US-AvaMultilingualNeural__ Woman with phone connecting to cell tower and satellite.

(03:46) 1. How do we 10x our reach without 10x the staff?

(06:28) 2. How do we ensure our fraud & safeguarding systems keep up?

(08:05) 3. How do we distribute 3+ million phones a year?

(09:28) 4. Will paying larger swaths of people drive inflation?

(11:19) 5. What does this mean for how we work with governments?

(13:10) 6. How do we keep our organization high performing?

---

First published:

September 29th, 2026

Source:

https://forum.effectivealtruism.org/posts/REJjytdSyRFXLSgtD/what-will-it-take-to-deliver-billions-a-year-to-the-world-s

---

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 was born at the end of the cold war, unaware of the danger we were emerging from. Duck and cover drills (as if those could protect schoolchildren from a nuclear blast) were a curiosity of the past.

The farmer-poet Wendell Berry, who wrote about the dread of nuclear war, died recently. From his 1968 “The peace of wild things”:

When despair for the world grows in me
and I wake in the night at the least sound
in fear of what my life and my children's lives may be,
I go and lie down where the wood drake
rests in his beauty on the water, and the great heron feeds.
I come into the peace of wild things
who do not tax their lives with forethought
of grief.

Mid-20th-century fallout shelter, Massachusetts

I’m so grateful for the generations of people who worked hard to lower our risk from nuclear war. There's no way to know how close we came to disaster. I’m grateful for the planning, the research, the negotiation, the recognition of common interests, the swallowing of pride. Humanity had built something that could overmaster us all, but people of many countries worked for decades to keep it [...]

---

First published:

September 29th, 2026

Source:

https://forum.effectivealtruism.org/posts/vYJY7hNJqXydpjPA8/thank-you-to-the-ai-safety-people

---

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.

TLDR: 

• We built an evidence- and philosophy-driven model for allocating philanthropic resources across causes.
• Last May, we launched the Cross-Cause Fund (CCF): a high-impact giving fund that allocates donors’ contributions based on the model's results. The fund is a pooled DAF at Charityvest, advised by RP at no fee, and has moved over 700 thousand dollars since its inception across GHD, AW, and GCRs (with more causes potentially to be added soon).
• We believe the CCF, with its underlying model, is currently the most rigorous, comprehensive vehicle for giving across causes and for giving under moral uncertainty. We also know its limitations, which we lay out below.
• Today, we share our ambitions and plans to improve it over the next year. We believe it's a crucial tool for effective giving, especially as the community prepares for a new wave of philanthropic resources.
• We invite you to read the methodology and the plans, red-team them, and give us the criticism (see the last section of this post) that will make this a tool the community can use as a default way to decide where to give, or a component of portfolio diversification.

What we’ve built so far

• We’ve built a [...]

---

Outline:

(00:25) TLDR:

(01:46) What we've built so far

(04:06) Why cross-cause prioritization is so important to us, for this community and for the upcoming wave of philanthropy

(05:55) Our aim

(07:35) Our plans

(07:38) The future of the Cross-Cause Fund

(07:43) The model's current positives

(09:06) The model's current limitations that we want to tackle

(10:14) Improvements we plan to make by the end of the year

(10:19) Rigorously evaluating the value the CCF is providing

(11:49) Worldview inputs

(14:08) Cost-effectiveness data and uncertainty

(14:31) Legibility improvements

(15:02) Governance

(15:22) 4-month to 1-year planned updates

(15:27) Worldview inputs

(16:20) Cost-effectiveness data and uncertainty

(16:56) In addition: Guidance for donors based on our ongoing foundational cross-cause prioritization and effective giving research

(18:58) We're building an agile team of experts to move this project forward

(21:45) Philosophy & modeling team

(23:57) Monitoring and evaluation and cost-effectiveness team

(24:56) Prospective

(25:52) Communications and outreach

(27:42) Expert and community consultation

(28:40) Let us know your thoughts!

(29:43) Acknowledgments

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

---

First published:

September 30th, 2026

Source:

https://forum.effectivealtruism.org/posts/uYddeJBttxwD4di3t/ccf-our-ambitions-for-the-most-rigorous-model-allocating

---

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.

Summary

I am drawing from my experience my 3+ years actively contributing to Effective Altruism Nigeria community building, I think that community building requires adapting to local context, measuring success by whether people become more agentic and take concrete actions rather than by attendance numbers, and investing in long-term capacity toward a community that can eventually sustain and lead itself.

The biggest challenges I've observed are measuring outcomes since impact often surfaces much more later most times, and funding limitations. To close these gaps, I propose continued capacity building programs for community builders, more region based talent and project incubators to lower entry barriers for global south participants and how EA co-working spaces could reduce dependence on external funding if they are run well.

Note: This post is mainly based on my opinions, is meant to offer additional perspective to anyone who is working in a similar context. This is also written independently of EA Nigeria.

What I've Learnt

1. Building with local context

Nigeria already has a number of organisations working on important problems, including global health, poverty reduction and farmed animal welfare. This creates opportunities for community members to contribute, whether by joining existing high-impact [...]

---

Outline:

(00:11) Summary

(01:13) What I've Learnt

(01:16) 1. Building with local context

(02:13) 2. Focus on building agents and networks, not just growing numbers

(03:55) 3. Investing in capacity for the long term, towards a self-sustaining community

(06:17) Our major challenges

(06:20) a.  Measuring growth

(08:06) b. Funding limitations

(08:42) A few ways I think we can improve communities in the Global South

(08:48) a. Centralised capacity building programs for community builders

(09:47) b. Establishing more localised talent and project incubation programs

(10:51) c. Establishing EA hubs for revenue generation

---

First published:

September 28th, 2026

Source:

https://forum.effectivealtruism.org/posts/r6fAzAgbs5xtdyP7G/what-i-have-learnt-from-community-building-in-nigeria

---

Narrated by TYPE III AUDIO.

The EA forum measures the level of AI usage in forum posts using the product Pangram. But how much does Pangram's "AI-Generated" label really indicate the degree to which an author has outsourced their thinking?

When they tested their 4.0 product, Pangram found that, by their definition, the proportion of AI-Assisted documents it classified as AI-Generated was 0.01%, 4%, or 7%, depending on the experiment. Then they omitted the experiments that found 4% and 7% false positive rates (FPRs) on their website outside their technical report, while advertising that the product detects AI-Assisted writing. Before I contacted them about this issue on September 17th, their claim on their website was more misleading—"99.9%+ Accuracy" was displayed directly adjacent to the phrase "Detects AI Assistance" on the text detection input box that many people don't read past. Similar claims remain repeated elsewhere on the main page instead of by the text box itself. I do not know if my message was the cause of the change.

Their experiment that found the 0.01% FPR might be more appropriate for identifying human-written text rather than AI-Assisted writing. For example, the prompt they gave to Claude for this experiment was "Fix spelling, punctuation, and [...]

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

---

First published:

September 27th, 2026

Source:

https://forum.effectivealtruism.org/posts/cF4zZ8EdQhfGZxed5/our-community-often-underestimates-the-weaknesses-of-pangram

---

Narrated by TYPE III AUDIO.

In my first “Meat Matters” substack, I wrote that despite some recent popular press obituaries, I remain optimistic about the scientific viability of cultivated meat, and I spelled out why.

Here's the executive summary:

In less than a decade, costs have fallen by orders of magnitude, government funding and scientific progress have surged, and multiple companies report that they are within striking distance of price parity. In short: The science and engineering challenges will not require inconceivably difficult breakthroughs or AGI, and of the dozens of cultivated meat scientists and engineers I spoke with for my new book Meat, all of them are optimistic about our path forward.

Zak Weston from Food System Innovations (FSI) works on market shaping for FSI and has personally reviewed techno-economic analyses from five different cultivated meat companies. He says that those analyses show “that at current cell line and media performance, if produced in commercial-scale facilities, they could reach cost parity or cost advantage compared to their animal meat benchmark.”

In Zak's view (and mine), “the technological progress over the past few years has been remarkable.”

Click here to read the full piece, and if you'd like to receive a monthly [...]

---

First published:

September 24th, 2026

Source:

https://forum.effectivealtruism.org/posts/ehmBFyL6Ft68DTJyT/meat-matters-why-i-m-optimistic-about-cultivated-meat-and-a

---

Narrated by TYPE III AUDIO.

It is often assumed that space expansion by intelligent actors would be clearly visible on a cosmic scale. I have previously outlined some reasons to doubt this assumption and to treat quiet cosmic expansion as a plausible explanation for the observed cosmic silence (see also Freitas, 1983; Fogg, 1987; Kent, 2011).

My aim in this essay is to present some additional considerations in favor of quiet cosmic expansion based on prudent cooperation and risks of ruin. Although these considerations are speculative, I believe they add further theoretical support to the plausibility of quiet cosmic expansion as an explanation of the Fermi paradox, and suggest that this picture should not be dismissed out of hand.

There is a broad spectrum of quiet expansionist hypotheses. My main focus here is on a cooperative version of quiet expansion, in which advanced civilizations avoid loud cosmic activity in part because they treat silence as a potentially prudent equilibrium or norm. Yet many of the considerations I review below also support quiet expansion more generally, including less cooperative or merely risk-averse versions of quiet expansion.

Outline of the cooperative silence hypothesis

The cooperative silence hypothesis explored here broadly involves the following premises:

• [...]

---

Outline:

(01:21) Outline of the cooperative silence hypothesis

(04:00) The quiet expansionist spectrum: From dark forest to cooperative silence

(06:22) Toy model

(09:01) Why quiet expansion is plausible

(09:33) A silent universe is evidence of what?

(11:42) Going loud would be evidence of what?

(12:30) Signaling strategic similarity

(13:27) Strong aversion to risks of ruin

(15:30) Avoiding irreversible signals and commitments

(16:23) Asymmetric error costs

(17:03) Long exploration

(18:34) Avoiding cosmic conflict

(22:39) Potentially limited gains from going loud

(25:02) Evidential cooperation in large worlds

(26:35) Selection effects

(29:27) Convergence is not required

(31:32) SIA supports cooperative convergence

(34:47) The observed evidence favors cooperative silence over dark forest

(35:38) Objections

(41:53) Practical implications

(46:52) Acknowledgments

(47:06) References

The original text contained 6 footnotes which were omitted from this narration.

---

First published:

September 25th, 2026

Source:

https://forum.effectivealtruism.org/posts/Tt3dTpkPjqQbAdSgL/the-cooperative-silence-hypothesis

---

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.

TLDR

• I moved from London to Accra a year ago to set up AWASH, a charity working on farmed fish welfare. I think more founders should move to their country of operations, particularly in LMICs. My hypothesis is that most people overestimate the cost of moving, relative to the benefits.
• The financial case is the clearest. The difference between what my Accra salary is and what my London salary would likely have been would fund roughly one year of AWASH's operations.
• Being in-country has made operations faster. The closer relationships with local stakeholders have also made problem-solving more effective.
• It has let me defer the senior local hire that founders based abroad usually have to make early, when they know least about the country.
• Moving provided impact I did not plan for. I co-founded EA Ghana in January, which would not have been possible from London.
• The costs were real, and mostly caused by administrative bottlenecks and inefficiency. The personal adjustment was easier than I expected and settled within about four or five months.

Intro

It's been a year since I moved to Ghana to set up AWASH, a charity working to improve farmed fish welfare. So this [...]

---

Outline:

(00:11) TLDR

(01:27) Intro

(03:44) Benefits of moving

(03:47) Cost-effectiveness

(04:58) Operations and early hiring

(08:25) Unexpected impact opportunities

(09:36) Challenges of moving

(09:40) Administration

(12:13) Personal adaptation

(14:18) Scale of commitment

(15:55) Conclusion

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

---

First published:

September 22nd, 2026

Source:

https://forum.effectivealtruism.org/posts/CHmH4fJNygz7sKZQJ/more-founders-should-move-to-their-country-of-operations

---

Narrated by TYPE III AUDIO.