Lack of intellectual diversity.
talks a good game but never delivers
Not willing to treat certain subjects (ex., religion) as a source of potentially-useful tools.
Too little regard for established philosophy and too many poor attempts to re-invent existing philosophical ideas.
The length of an inferential gap being bridged increases the probability that one of the steps will be wrong, and the conclusion can't be trusted after that.
endorsement of eugenics
No one has figured out the right way to do internet discourse yet. Less wrong was better than most in that respect, but still had all the usual problems of organization and coherent presentation for anyone not present through the initial development.
Percieved valorization of system 2 thinking over system 1
The medium doesn't support or encourage deep exploration of a topic.
Racism
The sequences were not, and still have to be, condensed into something less daunting for newcomers.
overreliance on work done inside the community, reinventions of the wheel are very common
If I stop a sequence, it's hard to pick back up again; and I'd really like a group of people to go through them with to discuss, but they're all already pretty discussed.
Not good enough at making sure all its interesting decision theory ideas were on solid ground
A lot of defensiveness induced by people being cruel to them on the internet.
'What is lesswrong.com?' was difficult to answer.
Focus on Many Worlds Interpretation. Regardless of it's merits, the reasons for it's inclusion were insufficiently explained (I only learned recently why Yudkowsky considered it to fit in the sequences, and am skeptical of the reasoning.)
When others went to fill the void partially, all the low hanging fruit exhausted.
Too intimidating and legitimately hard to compose idea into good post
Neglect of prior work
Too focused on math
LW-style rationality seems to behave badly in minds not already somewhat inclined that way. (Individuals with IQ below the median immediately misuse elements of Rationality when introduced to them. Given that half the human race is below median IQ, this is an important flaw.)
Heavy focus on Bayes theorem, I know people who have said in person it was overbearing.
In relation to 'Criticism of the Scientific Method': Not criticizing it specifically, but not understanding its proper place in the toolbox and throwing the baby out with the bathwater, as it were
Engagement with cognitive science seems to be non existent, and amounts to brains/subjective experience "is just computers bro".
Strong libertarian and computational bias. The problem is not in having a libertarian ideology or a cs background. The problem is that many participants have a hard time even understanding a problem or a solution that sounds divergent from libertarian dogma or the assumptions of current computational problem solving
The community struck me as being massively overconfident in subject matters not well understood by most members.
Too much focus on quantum mechanics
contains pissing contests and mean people (is not unique in this, but still)
Too much jargon that was weird Yudkowsky misinterpretations of real concepts
Tendency to reinvent the wheel, but call it the "toroid of static friction"
seriously though, > 2048-32 >implying unitary evolution as he means it is *real* BECAUSE I SAID SO
Not enough scholarship-virtue and not enough organization of the scholarship that was done
Worship of Eliezer Yudkowsky
People not being able to OTHER-apply the discourse ideals and insufficiently upvoting playfully/hastily written content that was on the right track.
Unresolved tension between epistemic and instrumental rationality
The tendency to take itself too seriously (e.g. Roko's Basilisk). LW's cross to bear is the purported saving of all of humanity. Relatedly, the tendency to invent proper nouns where perfectly good terms already exist & pretend that new and all-important concepts have been created.
If the site is past its peak, people probably expected too much, or a wide array of different things from it. This is natural.