Context

It’s been about nine months since I originally wrote my "Sam's hot Takes on Community Building" Forum Post, and it got a lot more engagement than I expected! Several EA uni group organisers reached out to let me know they’d changed their approach this year after reading it, which was really exciting. More broadly, it seemed to resonate with a lot of community builders who felt quite lonely in their discomfort with habits that have become normalised in EA community building.

I’ve now spent nine months actually trying to enact a lot of what I advocate for. I wanted to share an honest account of what happened: what worked, what didn’t, and what I’d tell someone trying this now.

What held up

I still fully stand by the altruism-first framing. When talking with other organisers and EA affiliated people about my post and approach, this is the number one thing that people have been bullish about! One organisation that’s taken this direction recently is Impact First, directed by @ Alicia Pollard. I’m really excited by what they’re doing and I think them existing helps open the Overton window toward the kind of community building I advocate for and others want more of. The more organisations that centre genuine motivation for impact over EA jargon fluency, the easier it gets for individual groups to do the same without feeling like they’ve gone rogue.

The D&D Dungeon Master analogy for facilitation has also been one of the most practically useful parts of the original post. I’ve used it training facilitators at Leaf and more ad hoc at Kairos and various EA uni groups, and it just clicks for people in a way that abstract “facilitation best practices” don’t. The way it captures power dynamics, the balance between freedom and structure, knowing when to impose and when to let people explore, all of it translates really naturally. If you’re trying to improve at facilitation or training other facilitators I’d recommend framing it more through this lens and maybe even trying your hand at DMing as a toy scenario for facilitation!

I’ve also softened a bit on fellowships. The original post was probably too broadly negative. The problems I flagged of power dynamic exploitation/ unawareness, deference, high attrition, treating people as career profiles, etc. are still real, but they’re problems with how fellowships are typically run, not necessarily with the format itself. I do think underpinning fellowships and structured learning with a solid community focus is more sustainable and healthier for the group. However, if someone is genuinely thoughtful about those dynamics, has good facilitation instincts, and treats a fellowship as one tool among many rather than the unquestioned default, they can work. The issue is when groups run them naively without thinking about what they’re creating socially.

EA Bristol

EA Bristol has been my biggest source of reflection on this post, because it contains both a proof of concept and a failure mode for my model.

Our first event was a low-effort pub quiz with vaguely “EAy” questions. We did one Instagram post to advertise it and got over 10 people to show up, which was pretty remarkable given that EA Bristol’s previous traction was maybe 4 people in a room discussing things awkwardly. Afterwards, we offered going to the pub in a completely no-pressure way and everyone came. We stayed chatting for hours with the organisers leaving before the participants. It was genuinely fantastic and an awesome feeling given the conviction I had after writing the initial post!

Notably, the group had more diversity (gender, ethnicity and academic focus). We just missed a lot of the problems EA groups typically run into. I don’t think this was a coincidence. Me being non-binary and another committee member not being white likely helped signal that this wasn’t a group exclusively for a narrow demographic. Who is visibly running the group affects who feels welcome. Moreover, lowering the bar for entry can make the space feel less pressured and competitive, avoiding compounding intersectional effects that can push marginalised groups away from such spaces.

Most importantly, several of the attendees who were second years or above specifically told us they’d been interested in EA Bristol the previous year but something had put them off. The demographics of the people running it. The weird structure of the fellowship. A sense of competitiveness. The lack of any social focus. These were people who already had the altruistic motivation, but found something about the traditional EA group packaging alienating enough to not show up. That’s not me arguing in the abstract that these problems exist; that’s real people telling me they experienced them. The takes I advocate for in the original post counterfactually surface exciting people more traditional groups miss out on altogether!

We followed this up with more casual socials all with similarly positive turnout and good vibes. For a few weeks, it really felt like proof of concept.

And then it stalled.

I got majorly sidetracked with Leaf, Non-Trivial, ERA, and other commitments. Events became less frequent, more ad hoc, and the momentum just evaporated. My co-organiser @Chris Clay, kept trying to run things but had very little success without that initial momentum. There was also a very successful offshoot AI Safety club (BAISC) that came out of Pathfinder (Kairos) by @anirdesh_shankar, which is still running with really consistent great turnout and engagement, but did split some attention away from EA Bristol.

What I learned from this

The model I advocate for is more vulnerable to capacity loss than the traditional EA one. When your week-to-week draw is social stickiness and the energy of specific people rather than the structure of a reading group or fellowship curriculum, things fall apart faster when those people aren’t around. A fellowship has a schedule that runs itself to some extent. A community built on relationships needs the relationship-builders to actually show up.

This isn’t unique to my approach, every student society faces the succession problem. But my model does amplify it. The spontaneity and responsiveness is what people experience and enjoy, but without enough planning behind the scenes, it collapses when one person gets busy.

If I were doing this again, I’d want a slightly larger committee with clear capacity commitments and a plan for the whole term mapped out in advance. Things were too ad hoc, and while that works when you have momentum, you need something to fall back on when capacity is constrained. That’s probably the biggest structural lesson: the social momentum and in-the-moment presentness should be what people see at the front, but behind the scenes you need enough structured scaffolding that it doesn’t all rest on one or two people’s availability.

Any uni society is fragile. Fellowships can be a band-aid that gives you structure for 8 weeks without people necessarily being there for the right reasons, being committed to staying, or actually enjoying being part of the group. But I do think my model needs more intentional infrastructure around it than I originally gave it credit for, and I’d rather be honest about that than pretend the early wins tell the whole story.

I do still think the response to my original argument as well as EA Bristol's first stint as a proof-of-concept do lend enough credence to the takes that it feels important to return to them and continue to move the discussion around community building in a positive direction rather than falling back on the same playbooks.