I was mostly attracted to LW for epistemic rationality, and I think epistemic rationality (rather than life advice or creativity) is LW's main advantage compared to all other online communities. It was also the subject of many posts by Eliezer, including the sequence Mysterious Answers which the wiki calls "probably the most important core sequence". But your list doesn't mention it, and indeed it doesn't get much discussion on LW lately, though people like Abram and Scott Garrabrant (and me) are trying. I wonder how things ended up like this.

To me, the best content in the old Less Wrong does not fit on any of you islands. It was developing theoretical rationality.

It would probably go on on near the AI safety island, but I feel like that is not fairly representing its generality.

Those islands you describe seem real (although "no smiles allowed" doesn't resonate). But I think the sort of islands I'm looking for are arranged on a different axis that this.

I want to talk about AI, and Instrumental Rationality, and Questionable Practices That Might Help Your Epistemics Or Might Destroy Them, and High Adventure, and Blood for the Art God, and a bunch of other things – they all seem fairly connected to me (since I'm a person, and everything I do is connected in some way).

The Islands I'm interested in are "the island where everyone is required to be good at X, but not required to be good at Y" and "the island where everyone is required to accept norm/paradigm Z because life is short and we can't argue about things forever", etc.

My underlying island that I'd like to cultivate, regardless of which object level topic we're discussing, is something like:

Accomplishing a Goal. We're not just talking for no reason, we're talking because we're brainstorming ideas or figuring out how to learn the current best practices or building an understanding of something. The goal might vary, which might change some of the norms that are ideal for the job.

Leveling Up. In addition to solving an object level goal, we're trying to grow in the process (at epistemics, at learning, at doing)

Disagreements are there to be resolved, and knowledge built upon. 10 years ago it was still a thing that I'd run into people arguing about creationism, I'm deeply glad I don't have to worry about that in my current filter bubble, and I'd like my bubble to spend the next 10 years getting to the point where things that are current considered contentious are considered resolved. (And, to be clear, I'd like to get there by actually resolving things and people changing their minds, rather than people filtering themselves out)

I currently believe a keystone of this is that people should be striving to have belief structures that are "cruxy" – i.e. have reasons for believing the things you do, and when you encounter disagreement, attempt to resolve that disagreement so that we don't have to hash it out over and over. (sort of a "hacker ethos" for beliefs. The hacker mentality is that problems are solvable, solutions are shareable, and no problem should have to be solved more that once).

Perhaps ironically, discussion that doesn't try to seriously resolve disagreement feels like either politics or recreation, both of which are nice and all but not what I'm interested in LessWrong for.

(This doesn't mean I expect disagreement to always get resolved anytime soon, just that long as it isn't, you should have a nagging sense that your job isn't done)

Introspection, Extrospection and Social Skills necessary for collaboration - This has some worldview baked into it that not everyone shares, and that I'm not currently sure about the framing of. I'm open to running threads that don't assume that worldview (and certainly am up for participating in other people's islands that don't). But, threads that don't assume this involve me running at a psychological deficit and will probably burn me out over time.

I expect the people I'm actively working with to be able to notice their own needs, notice the reasons for their beliefs, and be able to communicate about them in a collaborative (rather than confrontational) fashion. If you're at a point where if something threatening to you happens, and you don't feel like you have the social tools to deal with it non-confrontationally, I'd prefer you talk to me (or someone else you trust) about it one-on-one until you've gotten a handle on it.

An object-level norm that I want to push for: being allowed to revive old threads. LW right now feels to me like a "travelling band/circus/horde" that flits about from one party to the next. Part of the problem is that the frontpage cycles so quickly, and part of it is that almost nobody feels comfortable commenting on something more than a week old.

I might even make the following feature request: allow people to "refresh" old threads at the cost of karma (say proportional to how old it is) and push it to the top of frontpage. This is something I would definitely like to use to periodically provoke discussion and elaborate on certain posts in the Sequences, for example.

Fluff and Fiction: Blood for the Art God! Fun over fact.

I think I object to this description of the island for several reasons, assuming it's the same island I have in mind (probably you're being tongue in cheek here but even so). The goal for me is not to have fun, although fun is a nice side effect and explicitly stifling the fun would make the thing work worse; the goal for me is to change people's behavior (in ways they would endorse) by dealing with emotional blocks.

The target audience I have in mind is roughly people like me 2-4 years ago: interested in things like AI safety in some abstract sense, but with substantial motivational / emotional problems getting in the way of doing anything about it. It's hard to meaningfully contribute to AI safety research if, for example, you have a fear of stating opinions that haven't been vetted by an authority (this isn't the problem I was running into, I made it up, but I think things like this are reasonably common), ultimately stemming from a sense of feeling generally unsafe socially. Fears like that can go really deep and be really resistant to even being noticed, let alone debugged (by which I mean, it took me over a year to notice that I had something like this and several years of trying out all sorts of things to make noticeable progress on it, and I don't expect this to be particularly unusual).

I think it's important for us to be open, individually and collectively, to trying pretty weird shit in the name of tackling bugs like this, and accordingly the norms I want on this island are some combination of 1) being pretty open to people writing things in the emotional / poetic / illegible direction, or in general being experimental and trying things out in writing, and 2) being pretty intolerant of people who try to shut down 1).