This is the latest in a series of essays on AI Scaling. You can find the others on my site. Summary: RL-training for LLMs scales surprisingly poorly. Most of its gains are from allowing LLMs to productively use longer chains of thought, allowing them to think longer about a problem. There is some improvement for a fixed length of answer, but not enough to drive AI progress. Given the scaling up of pre-training compute also stalled, we'll see less AI progress via compute scaling than you might have thought, and more of it will come from inference scaling (which has different effects on the world). That lengthens timelines and affects strategies for AI governance and safety. The current era of improving AI capabilities using reinforcement learning (from verifiable rewards) involves two key types of scaling: Scaling the amount of compute used for RL during training Scaling [...] ---Outline:(09:12) How do these compare to pre-training scaling?(13:42) Conclusion ---
First published:
October 22nd, 2025
Source:
https://forum.effectivealtruism.org/posts/TysuCdgwDnQjH3LyY/how-well-does-rl-scale
---
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.
--------
15:39
--------
15:39
“Recommitting to Giving: A Personal Update” by frankieaw
TL;DR: I took the 🔸10% Pledge in 2016 and haven’t kept to it consistently. I’ve decided not to pay the backlog donations, and instead to recommit fresh from today, with simple systems to keep me on track. Sharing this for transparency and in the hope it may be helpful to others - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Why I’m posting In 2016, as a university student, I took the Giving What We Can 10% Pledge. I made my pledge publicly, and my social media profiles show the 🔸10% Pledge badge. For integrity's sake, I want to be equally public that I fell short—and [...] ---Outline:(00:29) Why I'm posting(00:52) What happened(01:53) Going forward ---
First published:
October 28th, 2025
Source:
https://forum.effectivealtruism.org/posts/3vcpERphsumgEzqeB/recommitting-to-giving-a-personal-update
---
Narrated by TYPE III AUDIO.
--------
3:59
--------
3:59
“Why Many EAs May Have More Impact Outside of Nonprofits in Animal Welfare” by lauren_mee 🔸, Animal Advocacy Careers
Many thanks to @Felix_Werdermann 🔸 @Engin Arıkan and @Ana Barreiro for your feedback and comments on this, and for the encouragement from many people to finally write this up into an EA forum post. For years, much of the career advice in the Effective Altruism community has implicitly (or explicitly) suggested that impact = working at an EA nonprofit. That narrative made sense when the community and its talent pool were smaller. But as EA grows, it's worth reassessing whether we’re overconcentrating on nonprofit careers, a trend that may be limiting our community's impact and leaving higher-leverage opportunities on the table. Why Now? As the EA movement has grown, it has attracted far more talent than the nonprofit sector can realistically absorb. This creates an urgent need to develop alternative pathways for talented, mission-aligned people. Under the current status quo, many end up feeling frustrated after going through multiple [...] ---Outline:(00:51) Why Now?(02:06) Important Caveats(03:20) The argument for roles outside of non-profits(03:25) Institutions Dwarf Nonprofit Capacity(05:19) Salaries Are Covered Outside the Movement(06:12) Counterfactual Impact Is Often Greater(06:57) A Healthier Distribution of Talent(07:54) Why This Might Be Wrong Advice(08:24) The Challenges of External Roles(10:13) Why These Risks Still Seem Worth Taking(10:57) Why Steering Everyone Toward Nonprofits Might Hurt the EA Community(11:03) Nonprofit Roles Are Saturated(11:27) Nonprofits Have Low Absorbency(12:24) Too Many Advising Channels, One Bottlenecked Funnel(12:49) Nonprofits are not a good fit for everyone, and they may be a much better fit for roles in other sectors.(13:31) Important Final Caveats ---
First published:
October 16th, 2025
Source:
https://forum.effectivealtruism.org/posts/FAmCmCavZ5vTzbRcM/why-many-eas-may-have-more-impact-outside-of-nonprofits-in
---
Narrated by TYPE III AUDIO.
--------
15:19
--------
15:19
“Framing EA: ‘Doing Good Better’ Did Worse” by Rethink Priorities, David_Moss
Summary As part of our ongoing work to study how to best frame EA, we experimentally tested different phrases and sentences that CEA were considering using on effectivealtruism.org. Doing Good Better taglines We observed a consistent pattern where taglines that included the phrase ‘do[ing] good better’ received less support from respondents and inspired less interest in learning about EA. We replicated these results in a second experiment, where we confirmed that taglines referring to “do[ing] good better” performed less well than those referring to “do[ing] the most good”. Nouns and sentences Nouns: The effect of using different nouns to refer to EA was small, but referring to EA as a ‘philosophy’ or ‘movement’ inspired the most curiosity compared to options including ‘project’ and ‘research field’. Sentences: “Find the most effective ways to do good with your time, money, and career” and “Effective altruism asks the question of how we [...] ---Outline:(00:12) Summary(01:23) Method(02:18) Taglines (Study 1)(03:40) Doing Good Better replication (Study 2)(05:23) Sentences (Study 1)(06:45) Nouns (Study 1)(07:41) Effectiveness focus(07:55) Conclusion(08:56) Acknowledgments ---
First published:
October 27th, 2025
Source:
https://forum.effectivealtruism.org/posts/Y6zMpdwkkAQ8rF56w/framing-ea-doing-good-better-did-worse
---
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.
--------
9:35
--------
9:35
[Linkpost] “The Charity Trap: Brain Misallocation” by DavidNash
This is a link post. In Ugandan villages where non-governmental organisations (NGOs) hired away the existing government health worker, infant mortality went up. This happened in 39%[1] of villages that already had a government worker. The NGO arrived with funding and good intentions, but the likelihood that villagers received care from any health worker declined by ~23%. Brain Misallocation “Brain drain”, - the movement of people from poorer countries to wealthier ones, has been extensively discussed for decades[2]. But there's a different dynamic that gets far less attention: “brain misallocation”. In many low- and middle-income countries (LMICs), the brightest talents are being incentivised towards organisations that don’t utilise their potential for national development. They’re learning how to get grants from multilateral alphabet organisations rather than build businesses or make good policy. This isn’t about talent leaving the country. It's about talent being misdirected and mistrained within it. Examples Nick Laing [...] ---Outline:(00:36) Brain Misallocation(01:16) Examples(05:37) The Incentive Trap(07:48) When Help Becomes Harm(08:48) Conclusion ---
First published:
October 23rd, 2025
Source:
https://forum.effectivealtruism.org/posts/6rmdyddEateJFWb4L/the-charity-trap-brain-misallocation
Linkpost URL:https://gdea.substack.com/p/the-charity-trap-brain-misallocation
---
Narrated by TYPE III AUDIO.
Audio narrations from the Effective Altruism Forum, including curated posts and posts with 125 karma.
If you'd like more episodes, subscribe to the "EA Forum (All audio)" podcast instead.
South Africa