A few times I've tried not deleting comments, but banning the commentators for a token amount of time with an explanation. Theoretically that seems good? But in practice they post my ban note on twitter and get their friends to send me hate mail...

Does anyone have a theory of comment moderation? I struggle people seem to be engaging in good faith, but are voicing opinions that are ~crazy. Removing these people feels dictatorial, but if you don't then you get gigantic useless threads of people trying/failing to change their mind.

me: i just can't understand why everyone takes themselves so seriously psychologists: that's because you have high neuroticism me: aha

What's with all the slide decks? dynomight.net/slides/

we're approaching superhuman coding and math but somehow LLMs are still mid at "writing"? this is undertheorized

ambiguous

hacker news has once again detected my secret plan to profit by writing about aspartame on my non-monetized pseudonymous blog

I quit drinking for a year dynomight.net/drinking

sans mots www.brasstacks.blog/interview-wh...

Theorem: You should make a better pingback Proof: 1. It would be better if discussions happened in 500-3000 word chunks 2. Partly that doesn't happen is that blogs don't encourage discussion 3. The "pingback" was one idea for solving that 4. But the pingback is ripe for abuse □

I feel that AI safety folks should stress the possibility of long timelines, because: 1. Long timelines are in fact possible (hot take). 2. Long-timeline scenarios are the ones where AI safety work has a higher (nonzero?) chance of positive impact.

I endorse this shape.

I think you can get any three of {healthy, cheap, low-effort, good-tasting}, but not all four.

Folks on HN are NOT buying Google's new process for side-loading apps on Android. As far as I can tell it's just a one-time 24-hour waiting period. Which... really isn't that bad? news.ycombinator.com/item?id=4744...

Can you create a "subreddit" on top of bluesky? Maybe some kind of custom feed or something? I want something intentionally fragmented, a sub-community where visibility of posts is entirely based on opting into the sub-community, not poster ID.

> once there were forums > then reddit ate all the forums > then reddit became crap > now there are no more forums > oooookkk?

LLMs predict my coffee dynomight.net/coffee/

good results

WHY DOESN'T THIS PLACE HAVE POLLS ARE POLLS BAD THEY SEEM OK

Bit ashamed to say that I'd never heard of Wilberforce, who appears occupy a special position on the (best human) × (best name) historical Pareto curve. en.wikipedia.org/wiki/William...

> change background to cycle over ~100 of my favorite paintings > it's nice, art is good, who know > it's so nice that start leaving second monitor empty just for the vibes > something is wrong here

"Software enables the enforcement of arbitrary rules that no human being would have the heart or foolishness to enforce."

The modern formatting addiction in writing dynomight.net/formatting/

Or maybe this should be S tier? It's hard not to admire the hustle.

Pop quiz: Infer the axis comparison

This is not as crazy as it sounds! In the US Civil War, 2/3 of soldiers died from disease rather than in battle. I could easily see myself predicting that better weapons would reduce deaths. (It's even conceivable that they did?)

Maybe there's a pattern here? dynomight.net/pattern/

i'm not being fed into a wood chipper you're being fed into a wood chipper

this tweet is such a wild ride how can you write something like this and not conclude that you should never ever ever ever work with the DoW? seriously, if this is your position, how is it remotely feasible?

Anyone remember when NASA announced in 2018 that they would be sending humans back to the moon by 2024? www.nasa.gov/specials/apo... I guess... everyone found this so implausible that it wasn't even worth making fun of them?

All: Please consider politely asking niplav to do this.

@niplav.site can you please do a self-experiment where you do double-blinded randomized saline infusions (or placebo) and check for cognitive effects?

An under-theorized bit of convergent evolution: >50% of modern people write with way too much "formatting" with constant disorienting bold text and recursive bullet points. I think this reflects a lack of confidence in the power of paragraphs. Experts don't do this. But LLMs have the same addiction?

Humans are unusual in that instead of laying around until hungry and then finding something to eat and then resuming lying around, they are obsessed with "doing stuff".

Personally, I think the reason that (secretly) using AI for writing is bad is that it disrespects readers. 1. They want to know. 2. So you should tell them. Openly using AI for writing is fine. But currently almost no one does that, because no one would read it. (Which is the point...)

Apparently this is true

Part of me wants to embrace a sort of Marxist worse-is-better attitude. Why not let grade inflation spiral forever until we can eventually get rid of grades? (Only part.)

I tried to explain what's happening in that recent paper about the heritability of human lifespan: dynomight.net/lifespan/

no KYC for chocolate bars, for now

This isn't getting enough attention. The CBP proposes to exclusively accept US visa-waiver applications exclusively through a mobile app, shutting down the website. Meaning you cannot visit the US unless you use a phone controlled by one of two US companies? What?

Say you wanted to create a trash blog with low-effort few-paragraphs thoughts posted several times a week. (Like they used to do in the long-long ago.) I think this is not: - mastodon - bluesky - a "newsletter" Is this format just lost in the mist of time?

AI is sycophantic is it

Asked some AIs who wrote latest post: Kimi K2: Dynomight GLM 4.7: Nate Soares Claude 4.5 Opus: Dynomight DeepSeek Chat V3.2: Scott Alexander Qwen 3: Dan Luu GPT 5.2: Scott Alexander Gemini 3: Dwarkesh Patel Llama 4 Maverick: Scott Alexander Grok 4: Scott Alexander Mistral Large 3: Scott Alexander

Why read novels? dynomight.net/novels/

Is there a label for the common thing where people say someone was wrong because of a technical interpretation of their words while ignoring the ~universally understood cultural meaning of those words?

what in the world doi.org/10.32996/jpb...

the true werewolf meta is everyone covering their faces and sitting in silence then voting at random

Good if make prior after data instead of before dynomight.net/prior/

Why the chicken crossed the road, according to various entities dynomight.net/chicken/

Underrated reasons to be thankful V dynomight.net/thanks-5/