Error

Hi Arepo, I think you are describing a tiny portion of EA software development, but are using the term “EA tech” to describe that small portion. I would suggest changing your post to something like “small EA organizations should hire an agency instead of hiring <=1 FTE of developers” and drop the term “EA tech work” unless it’s something that genuinely applies to all EA tech work.

The claim “EA tech work is bad for technical career capital” seems particularly unsubstantiated.

I care about this not so much because it affects your agency proposal, but more that I worry software developers who are reading this won’t understand that the experiences you describe are not representative, unless they read very closely.

As some justification: Perhaps the most obvious definition of “EA tech work” is to filter the 80 K job board for “engineering” positions. When I do this, the current top positions are at the UK government, DeepMind, OpenAI, the Rockefeller Foundation, and Microsoft. These positions generally do not suffer from the problems you mentioned, like looking bad on a resume.

The 80 K job board is sometimes criticized for being too longtermist-oriented. My guess is that most short-termist EA engineers are at places like Wave, which employ dozens to hundreds of developers, and similarly don’t suffer from the difficulties you mention here, though there is not an equivalent job board to check.

In part II you say:

At any given time in the last few years, there have been perhaps 5-10 software developers working full time in EA nonprofits.

CEA, my current employer, single-handedly employs this many full-time software developers.^[1] The same is true of my former employer Ought. I expect it’s also true of Redwood or Anthropic. It’s also true of EA-aligned animal rights organizations like The Humane League and Global Health and Development charities like GiveDirectly. So I’m guessing you are also excluding from consideration “EA nonprofits which have dedicated software development teams.”

My best guess is that you are considering only EA organizations which hire <=1 FTE of software developers. This is an important target audience to consider, but is very different from all of “EA tech work”.

You noted that 100% of the people who said they were worried about compensation being too low were just factually wrong about EA compensation. I suspect a similar thing is true regarding career capital, and would not want your post to reinforce that misimpression.

As best I can tell you don’t seem to address the main reasons most organizations don’t choose to outsource:

• additional communication and planning friction

• principal-agent problems

You could of course hand-wave here and try to say that since you propose an EA-oriented agency to serve EA orgs this would be less of an issue, but I’m skeptical since if such a model worked I’d expect, for example, not not have ever had a job at a startup and instead have worked for a large firm that specialized in providing tech services to startups. Given that there’s a lot of money at stake in startups, it’s worth considering, for example, if these sorts of challenges will cause your plan to remain unappealing in reality, since the continue with the example most startups that succeed have in-house tech, not outsourced.

I really like this idea! I’m particularly excited about having a new agency to do tech work for a fairly simple subset of the arguments you’ve mentioned:

1. The unmet demand for the work is there

   1. There appears to be a lot of high-EV software work that is currently undone within EA.

   2. There are probably many more software work that we haven’t identified because of not having the capacity to do them.

2. There are many talented software engineers within our movement, many of whom say they want to do direct work “someday.”

3. There appears to be a fair amount of money in our movement.

4. And yet the work is not done

5. Given that there’s a burning need in the market, it would be helpful to fill it.* An independent agency seems like a promising candidate to solve this “matching market” issue.

Personally, I find your arguments for having a independent tech agency and the advantages of an EA tech agency to be fairly compelling, and the argument for the agency to be primarily donor-funded rather than primarily consultant fee-funded or a hybrid system to be much weaker.

Like naively I’d have a picture of the initial tech agency to look like analogous to that of a VC funded startup where initial customers (EA orgs) are subsidized by VC (EA donor) money, until the tech agency “finds its feet” and can do a combination of being paid for services and funded for specific projects. (I gave more specific reasons for doubt in the relevant sections).

But nonetheless, don’t let my naysaying dissuade you! Given that the most likely route to making this tech agency a reality is through you (co)founding it, I think it’s much more important that the founder believes in the details of the product and the organization rather than a random Forum commentator does!

*A point that I’ve made privately a few times and should probably make into a full post is that a rational profit-maximizing firm should always increase investment until marginal cost equals marginal revenue, and the EA talent pipeline is nowhere near that.

Thanks so much for writing this. I have just skimmed the overall arguments but from what I have absorbed I tend to agree with you that there is a need to centralise tech resources.

Some quick comments/​thoughts:

I have been involved in a few organisations and volunteer organisations and all would have benefitted greatly from access to tech expertise. I know others who told me they had similar issues with their organisations.

Based on my experiences I suspect that if we had well funded and centralised tech support for EA orgs there would be many more websites and tech products and many of those that exist now would be greatly improved.

Going beyond tech: I think that there is a general need for EA consulting networks, organisations and agencies (e.g., tech, marketing, research, statistics) to help incubate and support the many ‘topic ’focused groups (e.g., career advice, animal suffering reduction, charities etc). Also to help disseminate best practice from the experienced EA orgs into the new orgs. The arguments in your post also seem applicable there too (and I see you mention something related in your last post). However, I am far from knowledgeable or expert here so not at all confident that there isn’t something I am missing .

Finally, I’d like if someone explored if a hybrid model is viable for tech and other agencies. Basically, they could consult for EA orgs at a budget or for free but subsidise this by also consult for non-EA orgs at full price. That model could have a lot of benefits and enable faster scaling, higher salaries etc.

As someone who does software and is interning at an established non profit in San Francisco (just doing research, not software engineering), I agree with your points, and I want to get a little bit deeper into the reasons why.

First, I think a lot of the really impactful technical is like, somebody’s working on a report and they need to make a pie chart, and they’re not that good with Google sheets. And the technical volunteer is really good at Google sheets and can finish the task in 10 minutes. But to get to the point where the technical volunteer was connected to this task, they’ve had to attend months of weekly meetings where there weren’t any technical tasks available. And that time is good time spent understanding the context of the work that the organization does, but won’t feel worth it if the technical volunteer’s goal for their involvement is to make technical contributions.

Also, I think it’s really difficult for non-technical people generally to describe the problem they want a technical person to solve, in a way that makes sense. Like, it’s not going to be the way it is at your typical tech job where your manager gives you the specs of the project and you just make it happen the way they asked. People are going to ask you for stuff that isn’t possible or is scoped differently from what they need, and to prevent making something that they aren’t actually going to be able to use, it’s likely that you will spend more time talking to people to learn about their work and how you can best help them than actually coding. And that skill of figuring out how to best help somebody by talking to them is a skill that I think most software engineers don’t actually have, unless they are also entrepreneurs who do that kind of thing regularly.

My recommendation for people who are good at computer stuff generally (doesn’t have to be as deep as software engineering, but if you are handy with WordPress, and Google sheets, you can be really useful already) who want to help out is to make an effort to be part of the community that is working on a problem that you care about. That way, you will get the context of what exactly people are trying to do, and understand the nature of the work, and when it becomes clear that a technical solution would be really helpful, people trust you to do it in a way that will be helpful, and there’s no friction with trying to onboard you because you are already there.

This is incredibly good.

I think this “operational” and “implementation” content, filled with on the ground experiences, is critically valuable.

I have a question/​suggestion:

Have you considered publishing the “sequence” in successive posts, maybe a few days or a week apart?

• Due to the algorithm, if people don’t make it to the other parts in the sequence, it gets buried off the front page as it ages. By posting a new section each week, you can get more total eyeballs and attention. This is a little rhetorical/​gamey but is basically used here.

• You can iterate and change on content each week, especially in response to feedback—so with the same amount of effort, produce content and connect to people more.