You can use https://web.archive.org/ for deleted web pages, e.g. https://web.archive.org/web/20250426145325/https://forum.effectivealtruism.org/posts/Pc3CFbYxPXgyjoDpB/seven-ways-to-become-unstoppably-agentic 

The author also deleted their EA Forum and LessWrong accounts, so you'd need to reach out to them directly to ask why

I don't think "EA Funding" is that useful of a term here. My sense is that forecasting is not funded by a large number of small retail donors thinking about forecasting as a category, but by few large institutions funding specific projects for specific reasons (which are sometimes not just effectiveness-related, and usually not public so hard to evaluate)

[The internet] was then not very useful until the 1990's.

I don't think this is true. Emails and FTP were established in 1971 and used a lot by academics, scientists, and the military^[1]

1. ^{^}

   From Gemini:
   
   The utility of email, FTP, and remote login (Telnet) during the 1970s and 1980s repaid the original government grants in three primary ways:

   1. Elimination of Duplicate Hardware Costs
   In the 1960s and 1970s, computers were multi-million-dollar mainframes. Prior to ARPANET, ARPA frequently had to purchase separate, identical computers for different research institutions. The network allowed a researcher at UCLA to log into and utilize a specialized mainframe at MIT. The cost of developing and laying the network infrastructure was significantly lower than the cost of buying duplicate hardware for every university the Department of Defense funded.

   2. Accelerated Scientific and Defense R&D
   Email and FTP collapsed the time required for complex collaboration. Instead of mailing magnetic tapes or waiting months for academic papers to be published and circulated, researchers shared datasets, software code, and peer reviews instantly. This rapid iteration sped up advancements in computer science, aerospace engineering, and defense logistics, delivering immense strategic value to the military and government.

And to add some obligatory nitpicking, "Individual starfish typically consume around 0.5 mussel per day although maximum feeding rates of 0.8 mussels per hour have been recorded for larger individuals"

I agree that the value of many interventions is sensitive to specific moral weights, but I disagree with "therefore the increase in subjective wellbeing from life-saving work is nowhere near as high as it could be for e.g. mental health types of work".

The increase in subjective wellbeing from GiveWell-funded work seems really high, and it could be competitive with mental health types of work. (or not, as different kinds of wellbeing can be reasonably valued in very different ways)

E.g. HLI "higher risk, higher reward" "Promising Charities" at https://www.happierlivesinstitute.org/charities/ are both also funded/recommended by GiveWell.

Worth noting that besides HLI focusing on happiness, AIM/Charity Entrepreneurship just incubated https://www.betterfuturesguide.org/ which seems to focus entirely on poverty reduction, and GiveWell is expanding their work on "Livelihoods Programs", which weigh income gains 2x higher than they normally would.

(I'm sure you know all the above, just writing it out for people with less context)

That's not clear to me: all GiveWell interventions have lots of life-improving benefits besides life-saving.

E.g. for the AMF, 33% of the estimated value comes from long-term income increases, and for each life saved there's ~200 malaria cases averted, which likely significantly increases subjective wellbeing

Thank you! Here's a link from web.archive.org of the EA Forum citation https://web.archive.org/web/20230715000000*/https://cdn.shopify.com/s/files/1/0255/4986/5032/files/DC-CASE-STUDY_NEW-BRAND_WEB2.pdf?v=1653689936 

But yeah if we can't trust that there was a real significant population reduction it doesn't mean much

I haven't read the whole post, but "519 g of fertility bait prevents one rodent birth" seemed implausibly high. I asked Gemini to review it, and it came out with this:

This 519g figure assumes wild rats will drink 10% of their body weight in bait every day as their exclusive hydration source. But real-world data shows intermittent grazing is enough to cause cumulative infertility.

For example, in the Washington D.C. ContraPest pilot trial (Nov 2019–Oct 2020):

• Site A had a starting colony of 391 rats.
• Over 12 months, the population crashed by 88% (the juvenile count specifically dropped from 121 to just 2).
• The entire colony consumed only 1.8L of bait all year.

If it truly took 519g to prevent one birth, 1,800g would have only prevented ~3.5 births for the whole colony.

Was it correct? I'm mostly curious about whether current LLMs can already help improving these estimates, or their reviews have too much noise

orgs like GiveWell are still getting a lot of funding

It's not just that these orgs are still getting a lot of funding:

1. their funding is significantly increasing
2. there's many more of them
3. many of them are making more and more varied grants themselves, e.g. GiveWell making 2 <$100k grants in 2026 which they didn't use to do 5 years ago, Founders Pledge brand new Catalytic Impact Fund

there were more fellowship and grant and award opportunities than I could possibly apply to. It does not feel like that today.

I'm surprised by this, I think there's a ton today. I'm not following this space actively but, besides the >100 job openings and >3 AIM programs mentioned above, here's some off the top of my head:

1. High Impact Professionals Impact Accelerator Program
2. CEA bootcamp (which as far as I know is not mainly about AI)
3. School for Moral Ambition fellowships and circles
4. Magnify Mentoring mentee applications (I think it now accepts more people than WANBAM did five years ago, but can't quickly find numbers. I see it got $371k from Coefficient Giving in August 2025, and their revenue seems to be increasing)
5. Animal Advocacy Careers course and career advising
   1. Their Job Board has 21 job openings from last week

You can also have a look at the most recent posts tagged "opportunities to take action" and the EA opportunities board, there's lots of non-AI stuff, enough to overwhelm newcomers as much as EA in 2021, and likely way more than EA in 2017.

Also in general if Coefficient Giving and others are making more grants to more things, it likely means that there are more opportunities.

funding for non-AI projects has dried up

What are you basing this on? I think the opposite is going on. Some datapoints that come to mind:

1. Coefficient Giving more than doubled their funding for GiveWell for 2026, adding $175M on top of the existing $100M. They also started two new funds
2. GiveWell's funding from non-Coefficient Giving donors is also increasing
3. Founders Pledge went from $25M money moved in 2022 → $80M in 2023 → $140M in 2024, and other major funders are emerging
4. Giving Green influences >$17M/year in climate donations, and recently started research into biodiversity projects
5. The EA Animal Welfare fund raised >$10M/y last year and is now targeting $20M/y
6. https://jobs.probablygood.org/ has 148 roles published in the last 4 days, only 10 of which are explicitly categorized as AI safety (although a few more involve AI)
7. Charity Entrepreneurship is launching more and more charities per year, and AIM as a whole has more programs