Epistemic status: These are my observations as someone who recently completed the CEA Career Bootcamp (June 2026), after ~3 months of engagement with EA and AI safety for the first time. I'm an outsider to EA / AI safety space by conventional measures with my enterprise SaaS background, Bangalore location, no research credentials and no prior community involvement. Some of what follows is my direct experience while the rest reflects conversations with others in my bootcamp cohort, a few of of whom are from the Global South. I'm sharing this partly because the bootcamp encouraged it, and partly because a few people in the cohort echoed these thoughts independently. But mostly because I wanted to hear of my blindspots from those more embedded in the ecosystem.

Some points need further work (see the section on “Points which need further work”), but I did not want to hold off sharing till then. So, the most useful thing you could do after reading this is correct me. If you know where the generalist/ops roles actually are, or have seen the geography problem addressed well somewhere, I'd rather know than have this post just sit there by itself.

________________________________________________________

What surprised me (positively)

The EA/AI safety space feels more alive than almost any other professional or impact-oriented space I've encountered. I say this having spent some time adjacent to social impact circles in India.

A few specific things that stood out:

• The community infrastructure is quite remarkable. Job boards, detailed career profiles, free advisor tools, curated databases, ecosystem maps, video lectures, bootcamps; And  a lot of it high-signal, not just corporate stuff or fluff pieces. The amount of content freely available to anyone who wants to engage is almost daunting. In a good way.
• There are identifiable town-squares: The EA Forum, CEA bootcamps, LessWrong, AstralCodex, Alignment Forums -  and a lot of them follow common posting and discussion norms, which makes it easier to orient. Probably also because a lot of them use the same infra as LEssWrong. This isn;t something I can say about most fields.
• People in visible roles make themselves accessible: This still surprises me. It lowers the barrier to engaging meaningfully, and it contributes to the community feel in a way that I don't think traditional corporate or even social impact spaces manage.

I likely wouldn't have uncovered most of this - at least not as quickly - without the structured, curated nature of the bootcamp. The bootcamp functions partly as a momentum-builder for people who might otherwise take much longer to get up to speed. 

But frustratingly, that momentum gets difficult to sustain. More on that below.

What struck me as fractured

The liveliness and accessibility are real. But they're unevenly distributed.

• The Global South is largely absent from conversations: This showed up both in the community infrastructure I encountered and in explicit conversations with bootcamp cohort members from non-Western contexts. It's not just representation in the abstract; it's about who's present when opportunities are actually being created and shared.
   
• Pedigree, credentialism, and who-knows-who remain significant gatekeepers. The AI safety space is full of credentialled PhDs from areas ranging from STEM to philosophy, which makes sense given the nature of the work. But beyond researchers, it also attracts accomplished generalists from journalism, policy, and adjacent domains to support corresponding functions at top organisations. 
  
  What's thinly spread, and often either unadvertised or not actively hired into, are support-staff and generalist roles. There're posts and conversations about a genuine demand for these roles, but the path to them isn't well-mapped. And the scale of available opportunities does not match the scale of need expressed in some conversations.
   
• A lot of what appears visible is still only visible to some: The bootcamp and the wider ecosystem do a fairly good job of making opportunities visible; with the caveat that you don't know what you don't know. 
  
  Many opportunities seem to be shared and created only in conversations happening in rooms rather than town-squares. I have strong feelings that this is the case, based on both my own experience applying to roles and events and the observations of others in the cohort.
  
  Even the opportunities that are surfaced aren't yet "normalised." Paths to them are not reliably accessible to people who lack the right network entry point.

The overall vibe of the space felt to me like that of a hip part of town with a thriving cafe scene with a lot happening. But a good chunk of it is inaccessible to most people in practice, beyond being able to talk about it online.

A note on diversity efforts: From what I've seen, the community acknowledges the homogeneity problem and has made genuine efforts to address it. But those efforts have so far focused primarily on race and gender rather than on geography, socioeconomic background and educational pedigree. This feels like an incomplete approach. The educational and organisational pipeline currently funnelling people into EA is producing relative homogeneity along dimensions that don't show up in the standard diversity frame.

An open question

Through a project called Orbit, I’m testing the hypothesis that structured, curated introductions can help diffuse network capital and opportunity access more broadly which should led to better cross-strata mobility. The bootcamp itself feels like partial evidence that this is possible: it makes opportunities visible to people who wouldn't have found them otherwise.

But if many opportunities are still being shared in closed conversations rather than public town-squares, the question becomes: how do we get those conversations accessible more widely?

I don't have a confident answer. One approach might be starting with a curated introduction layer, using the existing community infrastructure as a foundation. Something like what the CEA bootcamp already partially offers. I'm curious whether others have thought about this, tried it, or have evidence about what has or hasn't worked.

Points which need further work

• …Global South is largely absent from conversations…
  • Evidence? What kind of conversations specifically? 
• … either unadvertised or not actively hired into, are support-staff and generalist roles….
  • More concrete evidence here
• The diversity note at end:
  • Potentially inflammatory; Review if accurate and can be expressed better
• Open question: A bit too abstract. Needs to be tightened.
• “Cross-strata mobility” in my Orbit thesis

_________________________________________________________

If you're reading this and thinking "that's not quite right - here's where the generalist roles actually are" or "there are resources for non-Western entrants that you missed" - please comment. That kind of correction is one of the primary reasons I'm posting this rather than keeping it in my notes.

Thanks to @Carson for feedback on an earlier draft.