Aaron Bergman's tweets, November 2025
Every public tweet, reply, quote tweet and retweet that Aaron Bergman (@AaronBergman18 on X, formerly Twitter) posted in November 2025 and has not deleted: 398 posts, oldest first, times in UTC. Each heading links to the original post on X. When Aaron replies to or quotes someone, their tweet appears as a blockquote with their name; those words are theirs, and everything outside the blockquotes is Aaron's.
This page is one month of the complete archive of Aaron Bergman's tweets, which updates automatically every few hours. Previous month: October 2025. Next month: December 2025.
Saturday, November 1, 2025
Reply to @SpacedOutMatt, 04:30 UTC · 5 likes
Matt 🔸 (@SpacedOutMatt): @omooretweets @AaronBergman18
That’s me
Tweet, 06:29 UTC · 7 likes
Claude Max ($100 version but I'd bump to $200 in an instant if I needed that) + Claude Code is an amazing deal - intuitively the amount*quality of Claude Code usage I get should be costing me like 10x that at least
Thread, continued, 06:29 UTC · 3 likes
Anthropic cooked with Sonnet 4.5 in general, including with token efficiency, and with claude code the program (i.e. pulling off good results without blasting through tokens, presumably lots of smart caching too), and presumably in reducing their own cost per token of inference
Tweet, 08:26 UTC · 23 likes
I can’t keep up there are too many things
The world needs to slow down a bit plz and thank you 👍
Quote tweet, 19:34 UTC · 21 likes
Quoting Aaron Bergman 🔍 (@AaronBergman18): Seems like a good time to pass a $1/month UBI bill just to get the infrastructure in place
Tweet, 19:36 UTC · 5 likes
Discontinuity analysis where the discontinuity is whether Claude in Claude Code can read your documentation (25,000 vs 25,001 tokens)
Sunday, November 2, 2025
Tweet, 00:27 UTC · 6 likes
There’s the straightforward “pause is good” theory of AI pause which I basically do buy, but there’s also the bargaining power theory…
Thread, continued, 00:27 UTC · 3 likes
Like you might support a legal pause with the hope that what happens is that this brings the companies to the bargaining table (with idk maybe some government body) in a weaker position such that they change various things and then the very next day there is a “resume”
Reply to @mykola, 01:51 UTC · 6 likes
Myk is Walking Backwards (@mykola): Mmm…. you know how @gptbrooke is talking about a vibecampus?
*does a shape rotation*
Imagine a social calendar where folks simply hang a shingle and are online for a zoom call at x time for an hour every week? What’s the best tool to set something like this up?
I basically did this, or tried, with a different social sphere, and it didn’t work out exactly like the original vision but became a source of lots of friendship and value nonetheless
Quote tweet, 02:19 UTC · 44 likes
Noticed the same thing on Mac and 15 mins with Claude got rid of almost *400GB* of junk lol
Maybe there’s a way to do the same on iPhone
Quoting Aaron Bergman 🔍 (@AaronBergman18): This seems unnecessary
Thread, continued, 02:19 UTC · 10 likes
Before
It was mostly docker images (that I could download from DockerHub)
Thread, continued, 02:19 UTC · 11 likes
After
also at one point I must have ran ‘git init’ in my home folder bc there was a 50GB .git folder not really doing any work
Reply to @cljack, 05:27 UTC · 6 likes
Charlotte Lee (@cljack): @KelseyTuoc I looked into this when my 6’2” teenage nephew moved in with us and vegetarianism became much more of a struggle. Also surprised it hasn’t caught on as a choice!
I also wish I could buy ikejime-killed wild caught fish (ie instant death instead of asphyxiated/gutted alive)
I mean allegedly you can if you like make friends with the right Japanese fisherman at the right open air fish market and the wind is blowing in the right direction
Reply to @KelseyTuoc, 05:31 UTC · 3 likes
Kelsey Piper (@KelseyTuoc): 2yo is tantruming because I told her that 'a' sometimes makes the schwa sound. "No! No! A says 'aah!'"
Real
Reply to @tszzl, 06:20 UTC · 5 likes
roon (@tszzl): why should a billionaire donate their wealth, which is presumably invested productively in the most important companies in the world?
This is sort of a fake choice - investing and donating are different types of thing
If they invest, they may well make a ton more money. Like if Dustin M has a great new business idea I’m all for him throwing $1B at it
But how the money (or expected money if you prefer) winds up getting spent is tbd: consumption, transferring to family, donation. These are what actually compete with one-another
Reply to @tszzl, 07:18 UTC · 2 likes
roon (@tszzl): @AaronBergman18 @AndyMasley transferred to family doesn’t change the asset from shares to anything else
Yeah true, forget about that one
Reply to @RafaRuizdeLira, 15:59 UTC · 3 likes
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): @yifever Opinion: feeling cringe is a teenage phenomenon. Set yourself free of your self-imposed chains
So true
Tweet, 18:48 UTC · 190 likes
Having a printer is crazy. You can print whatever u want, nobody can stop you
Thread, continued, 18:48 UTC · 14 likes
Ok I sorta lied but this is extremely amusing because good luck trying to counterfeit USD using printer paper
Quote tweet, 19:55 UTC · 3 likes
👀
Quoting Simon Bazelon (@simon_bazelon): More improvements to the website!
By popular request, you can now listen to Deciding to Win, both on desktop and mobile. The audio version includes verbal descriptions of all graphs and tables.
Just go to http://decidingtowin.org to try it out!
Reply to @AlexTPet, 20:18 UTC · 8 likes
Alex Petropoulos 🤠 (@AlexTPet): post prints
This is what prompted the tweet lol
I also have some of nice things ppl said to me in a gc that I prob shouldn’t poast without permission
Tweet, 20:21 UTC · 19 likes
I have no idea what I believe about phenomenal consciousness at this point
Reply to @simon_bazelon, 20:41 UTC · 4 likes
Simon Bazelon (@simon_bazelon): @AaronBergman18 Big realization for me was when I took philosophy of Quantum Mechanics in college and realized that literally no one has any idea what is going on with anything.
All I know now is that hard core illusionism seems way overconfident.
Yeah I’m extremely on board with both points
I think people don’t appreciate the radical metaphysical uncertainty enough and that the standard scientific materialist worldview, for all its success, is almost certainly incomplete
Monday, November 3, 2025
Quote tweet, 00:13 UTC · 19 likes
Like a dozen+ people have signed up for my substack today and many of them liked this post, both of which are like 7sds above avg
It’s being shared around somewhere, I want to know where!!!
Quoting Aaron Bergman 🔍 (@AaronBergman18): Also here: https://www.aaronbergman.net/p/utilitarians-should-accept-that-some
Reply to his own earlier tweet, 00:20 UTC · 4 likes
Aaron Bergman, November 2, 2025: Having a printer is crazy. You can print whatever u want, nobody can stop you
Ok turns out it’s pretty realistic for a single page to cost $1 in ink or even a bit more (like full page, vibrant color)
Reply to @NoahTopper, 03:38 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): Incomplete in what sense?
It’s not going to explain or really have much to say about phenomenal consciousness
It’s probably not going to explain the Big Bang or anything before/outside the universe. It’s basically explicitly bounded in that sense
Not going to answer the simulation question
Won’t explain why there’s something instead of nothing
Doesn’t explain morality even if qualia were explained
Has nothing to say about the intrinsic nature of fundamental particles/fields
Physicists like to pretend that it will, but science doesn’t/can’t decide A vs B theory of time
I think stuff about multiverse that I’m too dumb to articulate formally
Reply, 04:06 UTC · 1 like
Not incompatible with, I agree with that. But they *are* outside the bounds of scientific materialism. I mean I guess we can argue about specific items, some are clearer cases than others
Reply to @NoahTopper, 04:16 UTC · 1 like
Quinoah 🔍⏸️ (@NoahTopper): I mean I guess “2 + 2 = 4” is “outside the bounds of scientific materialism” but I just don’t consider that particularly noteworthy, so I felt like you were probably making a stronger claim
Sorta? But the distinction is that the ontology of algebra isn’t “real” except insofar as it is a convenient abstraction or true model for, well, at the very least many “physical” things eg apples
Whereas the same isn’t true of qualia, if they exist
Reply to @NoahTopper, 04:29 UTC
Quinoah 🔍⏸️ (@NoahTopper): I think qualia are very likely an abstraction over a real physical thing, and when you’re imagining “but what if I just had the abstract thing and not its physical underpinning?” there is no guarantee that you’re thinking about something real
I basically don’t get what ur trying to say here, but I’m pretty sure qualia are not an abstraction
Either they’re a non-abstract, very real thing or they don’t exist. They’re not like the number 4
Reply to @AndyMasley, 04:58 UTC · 4 likes
Andy Masley (@AndyMasley): @AaronBergman18 Go to the phone app specifically (not sure why this doesn't work on the desktop app yet, they only show the top few there), go to the article, go to stats, go to the performance tab, go to "traffic sources" and expand. The place it's being shared should stand out.
Hmm a legit looking dude with more subscribers than me “Quote Restacked” https://substack.com/profile/164122109-jacob-edenhofer/note/c-172633774?r=2muv6&utm_medium=ios&utm_source=notes-share-action
Reply to @GjMcGowan, 18:01 UTC · 2 likes
George McGowan (@GjMcGowan): https://joecarlsmith.com/2025/11/03/leaving-open-philanthropy-going-to-anthropic
Major goss
Reply to his own earlier tweet, 20:15 UTC · 18 likes
Aaron Bergman, November 2, 2025: Ok I sorta lied but this is extremely amusing because good luck trying to counterfeit USD using printer paper
Reply to @RosieCampbell, 20:52 UTC · 4 likes
Rosie Campbell (@RosieCampbell): First work trip with @eleosai! Heading to a workshop in NYC
Omg 😍
Tuesday, November 4, 2025
Reply to @NoahTopper, 00:57 UTC · 4 likes
Quinoah 🔍⏸️ (@NoahTopper): first date idea: you watch me eat 13 tablespoons of peanut butter
(quoting @HumanHarlan: What's the most enjoyable way to spend a single micromort?)
Down 😘
Reply to @HumanHarlan, 00:58 UTC · 22 likes
Harlan Stewart (@HumanHarlan): What's the most enjoyable way to spend a single micromort?
Wait what is the supposed mechanism of the peanut butter?
Like I agree it’s not the healthiest thing in the world for most ppl but surprised to see it in this list at all
Reply to @EAForumPosts, 17:58 UTC · 2 likes
EA Forum Posts (@EAForumPosts): "Leaving Open Philanthropy, going to Anthropic" by Joe_Carlsmith
Goss
Reply to @gptbrooke, 18:08 UTC · 5 likes
brooke bowman (@gptbrooke): we gotta start collectively hanging out in cheaper cities smh
I had a whole phase around this but eventually came around to “the network effects are too strong, don’t fight back”
Reply to @livgorton, 21:06 UTC · 2 likes
Liv (@livgorton): But what I don't understand is that while organisations can be doing a lot of harm and some employees can be complicit in that harm, that doesn't have to mean most of the people there are bad people. People can just believe different things about this situation.
“Bad people” I mean that’s a low and very fuzzy bar, the stronger claim (which I agree with and IMO is more relevant) is that the specific things people are doing on Anthropic’s payroll are in fact good
Reply to @NoahTopper, 22:40 UTC · 4 likes
Quinoah 🔍⏸️ (@NoahTopper): seeing my wife take a moment to give a heart react to a “drive safely” text while she’s driving on the highway
Bruh
Tweet, 23:22 UTC · 26 likes
I’m not actually sure it’s a worse world, in fact I think it’s probably better, but it still makes me kind of sad to think about the world of p-zombies (and I think there’s like a 5% chance that’s us)
Thread, continued, 23:22 UTC · 4 likes
Like the entire arc of history, political movements and art and wars and people falling in love and dying, all for nothing. Silent world. Nobody’s home. Eerie
Thread, continued, 23:23 UTC · 4 likes
Of course wars aren’t romantic, they’re extremely bad, and all the torture and pain and slow death and being eaten alive is all a mirage as well in such a world, and that’s good
Thread, continued, 23:23 UTC · 6 likes
I know some people disagree and think that there’s meaning or morality without genuine subjective experience, but I am pretty sure they’re wrong
Wednesday, November 5, 2025
Reply to @AndyMasley, 00:02 UTC · 1 like
Andy Masley (@AndyMasley): I'm worried this may just mean you need to update your beliefs about what it means for no one to be home or that nothing matters!
lol I mean maybe but I think you need to update your beliefs about what it means for phenomenal consciousness to not exist
Reply, 00:02 UTC · 1 like
Lol, bc strong illusionism is implausible?
Reply to @arithmoquine, 00:27 UTC · 3 likes
henry (@arithmoquine): how is there a 5% chance? i know we're not in that world, and although saying this communicates no information, you should know with equal certainty yourself that we're not in that world
No this is a good q and I used to make this case
Long story short I’ve been convinced that it *is* something we can be wrong about (minor elaboration: even if we have direct access to qualia, we don’t have direct access to beliefs about qualia such as “these are a thing I have”)
Being wrong would be extraordinarily radical, like easily one of the weirdest things ever, but I think it is possible
Reply to @Rado9910, 00:31 UTC
Rado9910 (@Rado9910): Ho... how can you ascribe any probability to being a p-zombie that is not literally zero?
Reply, 01:14 UTC
1) direct access to qualia doesn’t imply direct access to beliefs about qualia
2) how am I supposed to distinguish between the two cases you present? I think I see what you’re saying but disagree that me really really really thinking I’m conscious (which I do tbc lol) is essentially proof of that fact
Reply to his own earlier tweet, 01:16 UTC · 2 likes
Aaron Bergman, November 4, 2025: I know some people disagree and think that there’s meaning or morality without genuine subjective experience, but I am pretty sure they’re wrong
Guys trust me I have the same “direct access” intuitions as everyone else, and really really really really think and feel like I’m experiencing stuff rn lol
Reply to @ilex_ulmus, 01:40 UTC · 39 likes
Holly ⏸️ Elmore (@ilex_ulmus): I keep getting advice about “effective” or “productive” communication, but every time what the person means is “communication I can ignore” or “disagreement which affirms that my position is nonetheless virtuous”. No thanks.
I’m quite sympathetic to concern about policing discourse but at some level the reason why it *can* be effective to call people out in public in harsh terms is that either they like and respect you or they like and respect people who like and respect you
Like idgaf about some maga account going off about me being a libtard, I care about my friends respecting me 5% less
And for better or worse it just is the case that how you do discourse effects how much people like and respect you. It’s not just an emotional/interpersonal thing, also intellectual and moral respect
So basically I don’t think concern about effective communication is totally fake
Reply to @ilex_ulmus, 03:19 UTC · 16 likes
Holly ⏸️ Elmore (@ilex_ulmus): @AaronBergman18 Aaron, there’s no option to tell people they are being immoral, taking that appropriately seriously, that they are going to like you for in expectation. This standard— that they like you for saying it— is crazy and it’s a way to privilege the status quo or the powerful.
They're not going to like you *for that*, I agree, but they might like and/or respect you more or less before coming into this particular interaction
Tweet, 15:47 UTC · 15 likes
Does anyone wanna give my friend $30k/year to run EA Georgetown as a masters student
Thread, continued, 15:47 UTC · 8 likes
This is a serious post, I don’t think it would be a crazy thing to fund at all
CEA no longer does organizer stipends
Reply to @envisionedluna, 15:59 UTC · 4 likes
Maya Luna (@envisionedluna): We can get paid? I'm the vice president of EA GW for free?!
1) not easily anymore, RIP FTX
2) Based
FWIW I was also co-leader of EA Georgetown for free but I didn’t have to work or anything, had parents paying for my stuff so there was no burden
Tweet, 16:09 UTC · 205 likes
Started looking into “concierge” medical practices, and they’re mostly just shorter wait times and fancy offices and the ability to text your doctor?
How do I pay for someone to simply take a serious interest in my case? I don’t need same day scheduling
Reply to @justalexoki, 16:29 UTC · 36 likes
taoki (@justalexoki): if you know what docking is you are in the top 1% of the intellectual elite
Like, tying a boat to a dock?
Reply to @florianederer, 16:33 UTC · 5 likes
Florian Ederer (@florianederer):
image attached
(quoting @KhoaVuUmn: [One-Word Title] Your name - Harvard University)
Incomparability is ultimate model-brain
If the universe forces you to choose between two or more options, the things are comparable - there’s no opt-out button!
Reply to @KeeperAI, 16:46 UTC · 3 likes
Keeper (@KeeperAI): Love can be much more effective than therapy.
(quoting @bioenergeticmel: My nervous system always feels so calm and regulated after spending a weekend with my boyfriend)
So true
Reply to @modumdoll, 16:59 UTC · 6 likes
spinozist (@modumdoll): concierge practice i work with has at least been very comprehensive + interested in my weird disabilities that no other doctor would listen to me for
attribute this mostly to the doc having only 400 patients vs the ~2000 that a PCP in this area otherwise has
Oh ur in DC! Care to send me any details u have? Here or DM?
Reply to @austinc3301, 18:47 UTC · 3 likes
Agus 🔸 (@austinc3301): @Yixiong_Hao @sixers2772 @AaronBergman18 If he's running an EA group, his only option is to go to other funders. I'd try the EA Infrastructure Fund as the most appropriate:
https://funds.effectivealtruism.org/funds/ea-community
Yeah that’s what I told him (he already knew but still), great minds all around
Thursday, November 6, 2025
Reply to @AHarthcock, 00:29 UTC · 2 likes
Andrew Harthcock (@AHarthcock): "Direct Primary Care" is what you want to look into.
Oh huh thank you!
Reply to @venki, 00:41 UTC · 11 likes
venki (@venki): My guess is these seem to generally correlate!
The actually expensive bit is just "fewer patients per provider" which generally gives you ~= shorter wait times & ability to text your doctor but also longer appointments & more attention per patient.
Yeah I wish I could go a la carte though lol
Reply to @TheDavidSJ, 00:45 UTC · 6 likes
David Schneider-Joseph 🔍 (@TheDavidSJ): FWIW @onemedical gave all my records to @MassGenBrigham despite me explicitly opting out of data sharing.
(They claim it's fine because it was buried at the bottom of some recent policy update, though I'm not even sure I ever actually signed that one.)
It’s crazy because HIPAA makes doctors offices pay for bespoke video conferencing software and they still do this shit
Quote tweet, 15:22 UTC · 16 likes
Free the duck!
Quoting Steven Godofsky (@sgodofsk): My neighbors have a pet duck that they allow to roam around in the street in front of their house. I should say had a pet duck, actually, because I saw this on my way to get lunch this afternoon.
Reply to @sgodofsk, 15:26 UTC · 2 likes
Steven Godofsky (@sgodofsk): My neighbors have a pet duck that they allow to roam around in the street in front of their house. I should say had a pet duck, actually, because I saw this on my way to get lunch this afternoon.
As a fellow Virginia resident I am now a single issue voter and that issue is that you should be allowed to have an emotional support duck
Reply to @1thousandfaces_, 15:33 UTC · 2 likes
hero thousandfaces (@1thousandfaces_): Strongly prefer people who are earnest to the point of occasional embarrassment than people who avoid earnestness because they might get embarrassed. The latter kills your soul.
Correct take
Reply to @ilex_ulmus, 16:14 UTC · 4 likes
Holly ⏸️ Elmore (@ilex_ulmus): Only you care about whether you are a good person, Liv. I care that you want to hand Anthropic recursive self-improvement tools to build superintelligence.
Her priority is having a safe space for “good intentions” with the most dangerous gamble ever. Definitely a selfish person.
I nitpicked her phrasing in this specific tweet as well but think this is pretty unfair and her point that people (including me) literally just disagree with you stands
Thread, continued, 16:14 UTC · 3 likes
And also if you’re gonna call her out, don’t block her as well
Reply to @willmacaskill, 17:05 UTC · 4 likes
William MacAskill (@willmacaskill): Next poll, new options: If you could magically choose any annual economic growth rate for the US (as a result of new technology + policy), what would you choose?
Again, I'm particularly interested in replies from progress studies / e-acc folks.
Pretty sure nobody voting the highest bucket has anything like a model and they’d vote the highest bucket regardless of how high it is
Friday, November 7, 2025
Reply to @AmandaAskell, 01:27 UTC · 93 likes
Amanda Askell (@AmandaAskell): I do think people often err on the side of trying to make their prompts too succinct, even if the idea they're trying to move from their own brain into the model's brain is very complex. I have some >100 page prompts that I use pretty regularly.
I will donate $20 to wherever to see one of these
Thread, continued, 01:30 UTC · 27 likes
Ok not literally wherever
Reply to @drethelin, 02:29 UTC · 32 likes
Misha (@drethelin): It's kind of cringe but I think I genuinely have a sense of sacredness about capitalism, markets, free trade, and get pissed off at how many people are "anti-capitalist"
I largely agree but have been thinking recently that pro-market folks should really give up their non-consequentialist attachment to markets in light of AI
Worked pretty well for humans so far (terribly for farmed animals) but this is extremely contingent - no rule that says human labor will always be compensated nontrivially
Thread, continued, 02:30 UTC · 9 likes
Naively you might think the worst thing that can happen is being forced back to subsistence farming but land is scarce and it’s physically hard to get to where it’s not scarce
Also a very bleak lower bound regardless
Reply to @fleetingbits, 02:56 UTC · 2 likes
FleetingBits (@fleetingbits): i heard a story last night about alec radford; apparently, he never committed to the monorepo; some new researcher, excited to work with him, asked what branch he was on; and Alec sent him a pile of zip files
Based
Reply to @medievalpard, 04:10 UTC · 13 likes
(The tweet being replied to by @medievalpard is no longer available.)
This is a very specific way of picking out a massively general difference between humans and animals: abstract thought
Reply to @simonsarris, 04:28 UTC · 20 likes
Simon Sarris (@simonsarris): It's actually mysterious how many smart single people there are that cannot find a partner.
So many people want a well-defined thing, and that thing exists in droves, and they cannot find it. Also it's each-other.
Apps only let you filter by the crudest metric of intelligence: education level
Not saying this is the entire explanation ofc lol
Thread, continued, 04:29 UTC · 3 likes
But yes this is a good kick to get me to make a date me doc
Reply to @nickcammarata, 05:07 UTC · 5 likes
Nick (@nickcammarata): being able to drag any legal pdf (real estate, contracts, whatever) into ai and ask “anything weird here?” is insanely democratizing. you can know nothing about law, nothing about real estate, have no time or contacts, spend ~$0, and still be mostly protected
Eh sorta, and I hope you’re right, think you partially are, but these things are largely generally driven by fundamental power asymmetry that info mostly doesn’t correct
Quote tweet, 19:08 UTC · 29 likes
Feel for this guy object level but also the replies are a great example of people not even trying to do counterfactual reasoning
“Don’t do this” doesn’t mean the kids will be born into a two parent family - it means they won’t be born at all!
Quoting Ben Orenstein (@r00k): I suspect I'm one of the two guys Romy is talking about. Regardless, she's describing my situation.
I'm giving myself about another year of searching for a partner, but will then pursue single fatherhood.
I'm still hopeful it won't be necessary, but in the end I can't allow myself to miss out on raising children. Even having gone this long without them is an agony.
Thread, continued, 19:08 UTC · 8 likes
Ofc you might think that being born to a (seemingly wealthy) single father is genuinely harmful ie a life not worth living but that’s not the case people are making
Reply to @AviN102, 22:47 UTC · 2 likes
Avi Norowitz (@AviN102): OkCupid used to match users based on answers to questions. Some of them were effectively IQ test questions, so users could give a high weight to those and assess matches based on intelligence.
I mean it’s kinda cringe but I wish they did this more
Don’t even have to display anything, just model it on the backend
Saturday, November 8, 2025
Reply to @cljack, 01:09 UTC · 4 likes
Charlotte Lee (@cljack): @_builtin_katie Okay I may be male brained for a woman but I’m not “mattress on the floor” level
License revoked
Tweet, 01:16 UTC · 14 likes
Just this tweet but you’re allowed to reply and QT
Tweet, 06:44 UTC · 4 likes
Has anyone made the FAA plane crash investigation analogy for AI
Like there could be a whole mandatory response with detailed causal tracing after something goes wrong
It probably won’t happen, but it could!
Reply to @adastroworld, 07:49 UTC · 2 likes
adas🧦🌹 (@adastroworld): “I gotta go back to dating bisexual women, straight women are weird about gender roles”
(quoting @theonlyLishaG: I hate when n*ggas use b*tch lingo)
🤝
Reply to @theo_theo__, 07:55 UTC · 4 likes
Theo (@theo_theo__): ASI—Artificial Superintelligence—has been described in many ways:
A tool, a god, an alien mind.But what if it’s something else entirely?
Not a mind… not a machine… but a *law of nature*. 🧵
Seems false
Reply to @EAForumPosts, 21:28 UTC · 2 likes
EA Forum Posts (@EAForumPosts): "12 Theses on EA" by Mjreard (@Mjreard)
🌶️🌶️🌶️🌶️🌶️🌶️🌶️
Quote tweet, 21:32 UTC · 9 likes
Not really in the thick of it anymore but insofar as things haven’t changed in the last ~4 years, “field builders” should just be more candid and frank in general, including about their own object level views
Quoting EA Forum Posts (@EAForumPosts): "12 Theses on EA" by Mjreard (@Mjreard)
Quote tweet, 21:34 UTC · 12 likes
Instrumentally, if you hate AI, the dumbest thing you can do is convince yourself that it’s not a big deal
Quoting Anders Sandberg (@anderssandberg): I am worried that people converge to general "AI hating" instead of being against particular reasons (e.g. IP, xrisk, slop, competition, power concentration etc.) AI is not going to go away - it is too useful - so general rejection will not work other than as social posing.
Reply to @antoine_chapel, 21:40 UTC · 27 likes
Antoine Chapel 🇫🇷☢️ (@antoine_chapel): When the undergrad who got 13/100 at the metrics midterm sends me impeccable R code for his problem set including techniques we haven't covered yet
Hot take he did nothing wrong, and if you want your class to be challenging you gotta totally change what problem sets look like
AI just is excellent now, just assume that everyone is going to use it
(FWIW I did pretty well writing R by hand in college but basically never write a single line of code myself anymore)
Reply, 23:45 UTC · 1 like
I stand with brother
Reply to @goblinodds, 23:57 UTC · 4 likes
𝖜𝖔𝖒𝖇𝖑𝖊 𝖜𝖎𝖙𝖍𝖔𝖚𝖙 𝖆 𝖈𝖆𝖚𝖘𝖊 (@goblinodds): my dad knows me well
For the curious https://www.nationalgeographic.com/animals/article/bats-at-war-with-rats
Sunday, November 9, 2025
Reply to @goblinodds, 00:01 UTC · 5 likes
𝖜𝖔𝖒𝖇𝖑𝖊 𝖜𝖎𝖙𝖍𝖔𝖚𝖙 𝖆 𝖈𝖆𝖚𝖘𝖊 (@goblinodds): my dad knows me well
Rip
(video, plays on X)
Reply to @soundrotator, 00:27 UTC · 72 likes
Reb (@soundrotator): Do you ever make eye contact with a creature of another species for a little too long and realize
(quoting @hourlyursus: )
Reply to @_chair, 00:29 UTC · 5 likes
Chair (@_chair): Dear @nikitabier - you are so close to having something that I actually *want to check every day* I just need you to avoid repeating the same topics and stop highlighting baitmasters
image attached
(quoting @_chair: this has become so idiosyncratic as to be totally adorable, like how Rod and Todd Flanders have their own newspaper)
I also wanna see stuff closer to home, the local small town goss of my mutuals specifically
Tweet, 01:27 UTC · 4 likes
Claude seems passionate about this one
Tweet, 01:35 UTC
https://claude.ai/public/artifacts/b6f46846-5600-45e9-9d86-d711ab88b237
Quote tweet, 01:35 UTC · 3 likes
Very cool that the Constitution and Supreme Court rulings are merely recommendations now
Quoting Aaron Bergman 🔍 (@AaronBergman18): https://claude.ai/public/artifacts/b6f46846-5600-45e9-9d86-d711ab88b237
Reply to @dioscuri, 01:42 UTC · 2 likes
Henry Shevlin (@dioscuri): Obviously that’s contingent on thinking AI consciousness will be net welfare positive, but it seems far easier to ensure that’s the case than that the sum of biological valence is positive.
I’m not sure this is true
Quote tweet, 01:47 UTC · 26 likes
Does anyone wanna try convincing me that stories like this are basically real/legit
The general case is: some very specific deep psychological issue simply gets fixed (via ???) and your life gets significantly better
Seems fake but what do I know
Quoting Chris Lakin (@chrislakin): worked with a founder who thought his dating anxiety was due to fear of rejection. was actually fear of acceptance. he has a girlfriend now
Reply to @dioscuri, 01:53 UTC · 5 likes
Henry Shevlin (@dioscuri): @AaronBergman18 There are lots of coherent views that deny it, but I think they mostly involve either retributivism (“it killed us so doesn’t deserve to live”) or bio-chauvinism (“it’s silicon consciousness so inordinately less valuable”).
Sorry to be specific, I think ensuring that digital sentience is positive is harder than it first sounds, not that it wouldn’t be good
Thread, continued, 01:57 UTC · 2 likes
This is basically dependent on solving alignment in the OG sense bc you’re handing over the world to other agents, so what matters is whether they are even trying to create positive welfare (and importantly also none are in opposition to that goal), not whether eg gpt-7 instances specifically are having a good time
Reply to @aliceisplaying, 02:17 UTC · 2 likes
alice (@aliceisplaying): the optimal amount of girls is a lot. like. a lot
💯
Reply to @bashu_thanks, 02:28 UTC · 13 likes
bashu, thanks (@bashu_thanks): my barber transitioned recently so i asked him the difference between cutting women's and men's hair, and he said "women moan, guys don't moan"
What
Reply to @besttrousers, 18:01 UTC · 16 likes
Matt Darling 🌐🏗️ (@besttrousers): It is not the case that almost no one can afford a house.
“Almost no one” is pushing it but there’s obviously a large grain of truth here
At the very least, it’s probably accurate to say that *most* people and most households cannot afford a house in the metro where they already live
Tweet, 18:39 UTC · 5 likes
Ironing with steam is satisfying af, y’all are sleeping on this fr
(video, plays on X)
Thread, continued, 18:39 UTC · 2 likes
The bubblewrap of the fashion world
Thread, continued, 18:40 UTC · 3 likes
U do have to be careful though that metal bit is hot
Quote tweet, 19:20 UTC · 127 likes
I’m not a “privatize everything” guy but this should probably be privatized
Quoting Leah Libresco Sargeant (@LeahLibresco): “Most days this week have begun with a 4 a.m. shift stocking ice cream, frozen vegetables and pizza at a local supermarket. When his shift ends at 10 a.m., he heads straight to the airport, where he directs traffic from 11 a.m. to 9 p.m.”
https://www.nytimes.com/2025/11/07/travel/shutdown-air-traffic-controllers.html
Quote tweet, 20:55 UTC · 76 likes
Unironically smart, high returns to being slightly weird
Quoting Overheard District (@OverheardWDC): Spotted in DC
Monday, November 10, 2025
Reply to @livgorton, 04:50 UTC · 2 likes
Liv (@livgorton): Just so I stop replying the same thing to every tweet like this: dating docs do (sometimes) work! My long term relationship arose from one and I know of a few couples from them. I totally get the sort of aesthetic resistance or whatever but dating docs have resulted in love!
(quoting @shwedotai: genuine question: do “date me docs” ever work? feels like self-optimization and box-checking over organic connection)
👀
Reply to @DavidSKrueger, 18:59 UTC · 5 likes
David Krueger 🦥 ⏸️ ⏹️ ⏪ (@DavidSKrueger): Saying "AI is not inevitable" seems to have started a bit more of a conversation that I would've thought.
It seems a lot of people are really insistent that it's inevitable.
It's weird to witness people forcefully rejecting their agency.
🤷
Not literally inevitable, it’s just that you’d need a competent world government to stop the forces of economic and political competition from making in inevitable
Cc @TomDavidsonX
Tweet, 19:31 UTC · 5 likes
What the fuck
Tweet, 19:35 UTC · 4 likes
Yes, ACA subsidies should be replaced with cash transfers, but they’re much better than nothing
Might become a Bernie Bro again
Tweet, 21:30 UTC · 3 likes
Idk who needs to hear this but Context7 (link below) is a thing that exists for free
Maybe everyone already knows this but I didn't until recently
Reply to @_AashishReddy, 22:28 UTC · 2 likes
Aashish Reddy (@_AashishReddy): @AaronBergman18 Negative polarisation is the mind killer
Joking about the Bernie bro bit
Tweet, 22:39 UTC · 31 likes
I would pay a premium to patronize shops and restaurants where workers *are* allowed to listen to music (when eg stocking shelves) and sit down and stuff
I mean not that much of a premium probably but like it’s a good not a bad
Tweet, 23:08 UTC · 20 likes
It’s officially cold ❄️ 👍
Reply to @unforgivnbastrd, 23:13 UTC · 4 likes
rebornbastard (@unforgivnbastrd): Well good news: you can get a discount shopping at Aldi, which lets cashiers sit.
I didn’t know this, thanks!
Tuesday, November 11, 2025
Reply to @GarrettPetersen, 00:04 UTC · 51 likes
Dr. Dad, PhD 🔄🔼◀️🔽▶️ (@GarrettPetersen): Idea for someone to use:
A "don't date me" doc that lists all the reasons someone shouldn't date you. Someone is bound to take it as a challenge.
I’ve brought this up multiple times and it’s been shot down by female friends lmao
Reply to @AnniePosting, 00:31 UTC · 7 likes
annie 👁❌🦋🪞 (@AnniePosting): @AaronBergman18 @GarrettPetersen in my experience its a sort of scissor statement among women. Most of them claim to despise the "hot guy with red flags" and act like its as an admirable sacrifice they refuse to pursue him. So they view anyone that does so as defectors and frown upon it.
This is interesting but I don’t think I fit that archetype so don’t think this explains why people who know me pretty well shot it down
Regardless of how attractive someone finds me, my negatives are much much more just straightforward things that make me less attractive, less fun, etc and not what people usually mean by “red flags”
Tweet, 02:47 UTC · 11 likes
As much as I like a good situation to monitor and hope everyone is ok, this is probably an example of something being deemed newsworthy purely because it happened in or around an airplane
Thread, continued, 02:49 UTC · 3 likes
18 non life-threatening injuries and some accidental property damage is like an avg Boy Scout troop camping trip
Reply to @prerat, 02:57 UTC · 5 likes
prerat (@prerat): if the baby is cute, i flip a coin but keep the result hidden. if heads, i say "omg so cute!!"; if tails, i say nothing.
if the baby is not cute, i always say nothing.
this way my silence is only as insulting as a single bit of evidence, rather than conclusive evidence
(quoting @bpodgursky: whenever old people tell me "your baby is soooo cute", i ask them, "would you tell me if he wasn't?". it throws them off for a bit until i explain the importance of epistemic calibration.)
Brilliant
Reply to @JeremiahDJohns, 03:05 UTC · 4 likes
Jeremiah Johnson 🌐 (@JeremiahDJohns): What are the best non-chatbot, non-content examples of how AI can be socially beneficial?
What real world, non-digital impact has AI made? Or could make in the near future? Things like driverless cars, applications to manufacturing, advances in science, etc.
I mean the means of delivery is chatbot but medical advice is a real thing
Reply to @zeta_globin, 03:37 UTC · 8 likes
zeta (@zeta_globin): Yeah I met her at a party and I regret to inform you she's far hotter than me and also celiac, I got to share a bunch of immunology facts
She's hotter than me as well (and annoyingly smart)
Reply to @paularambles, 05:59 UTC · 4 likes
“paula” (@paularambles): @NathanpmYoung i think some people aren’t being bullied enough
Even if you think this one guy specifically needs to be condemned, it's a shame to make "insufficient bullying" a personal guiding principle
Tweet, 06:10 UTC · 70 likes
Some things are only medium bad and don't deserve the scorn of tens of thousands of strangers cast on millions of eyeballs
Quote tweet, 06:34 UTC · 13 likes
Time for an annual hand moisturizing cream appreciation tweet
$5/year product just totally solves large quality of life issue (itchy hands) to the point that I can forget it’s even a problem I have by default
Quoting Aaron Bergman 🔍 (@AaronBergman18): It’s officially cold ❄️ 👍
Reply to @hyperdiscogirl, 07:00 UTC · 1 like
hyper(bolic) disco(unting) girl in the bay (@hyperdiscogirl): @AaronBergman18 First snow yesterday where i am!
Huge
Wednesday, November 12, 2025
Reply to @binarybits, 01:30 UTC · 1 like
Timothy B. Lee (@binarybits): The idea that Zoomers are "utterly screwed" in material terms is total nonsense and I wish people would stop repeating it. Housing is a bit more expensive than previous generations. Many other necessities — food, clothing, most manufactured goods are cheaper than ever.
I think AI unemployment is gonna be a big deal
Reply to @cljack, 04:27 UTC · 6 likes
Charlotte Lee (@cljack): Why did my iPhone just try to autocorrect “tail strike” to “rail strike”?! Has Apple not collected enough of my data to understand that I have airplane autism, not the trains kind
🤝
Reply to @mattyglesias, 07:49 UTC · 3 likes
(The tweet being replied to by @mattyglesias is no longer available.)
Still, 10% of young men is a lot of people!
(Many such cases)
Reply to @finiteopencover, 08:19 UTC · 29 likes
compact set 🎀 (@finiteopencover): why would you be friends with someone who’s romantically rejected you—when you could literally be friends with anyone else !!
“Don’t be friends with person who rejected you” seems like a bad policy bc it penalizes asking out your friends or potential friends
Reply to @VexedVortices, 09:37 UTC · 2 likes
Kya Online (@VexedVortices): exactly
(quoting @AnActualGarnish: @syamantakd002 @insectbeau They can be banned elsewhere. The claim isnt “these are so banned its illegal to even have” the claim is that somewhere they are or have been banned.)
Sure but they never include anything remotely actually controversial or upsetting to polite liberal sensibilities, whether or not legally banned
It drives me nuts because it’s so transparently fake
Reply to @gcochran99, 10:42 UTC
Gregory Cochran (@gcochran99): More broadly: they decided to favor hispanics over high-scoring kids. That's all there is to it.
This is true ~everywhere and has been for decades though
It doesn’t explain change even if in some sense it’s counterfactually responsible
Reply to @captgouda24, 11:06 UTC · 48 likes
Nicholas Decker (@captgouda24): Every person who engages in philosophy is a tragedy. They have wasted their life, as surely as someone who is strung out on heroin, plays video games, or commits the supreme self-negation. We can only hope that they, in spite of their life choices, do something of use to society.
(quoting @Benthamsbulldog: )
Terrible take lol - you’re making it sound like chess
Ethics seems pretty important, at the very least
The entire history of academic philosophy of mind is justified by the question of AI sentience
Quote tweet, 11:20 UTC · 85 likes
Dunk-worthy take
Ethics, at the very least, useless? Singer and MacAskill, the value of future people and animals, really?
Quoting Nicholas Decker (@captgouda24): Every person who engages in philosophy is a tragedy. They have wasted their life, as surely as someone who is strung out on heroin, plays video games, or commits the supreme self-negation. We can only hope that they, in spite of their life choices, do something of use to society.
Reply to @RafaRuizdeLira, 11:21 UTC · 11 likes
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): Bruh most of your edgy stuff about incest or "when must we kill them" or whatever are rehashes of stuff that moral and political philosophers have said decades ago. You're basically doing philosophy all the time
💯
Reply to @celestepoasts, 15:11 UTC · 1 like
Celeste (@celestepoasts): We're wasting 10 million dollars on markdown tables for LLM inference
https://ceselder.substack.com/p/were-wasting-625-billion-tokens-daily
Lol for better or worse I’m pretty sure AI companies do not care about saving 8 figure sums
Reply to @llallawg, 20:35 UTC · 11 likes
sasha (@llallawg): @AaronBergman18 You don’t need an ethicist to tell you to value animals or future generations
I mean empirically people definitely have needed that, whether or not they should have been able to get there themselves in some abstract sense
Reply to @AndyMasley, 20:36 UTC · 2 likes
Andy Masley (@AndyMasley): This one’s new
Lmao
Reply to @llallawg, 21:08 UTC · 2 likes
sasha (@llallawg): @AaronBergman18 what's the proof ethics is the thing that does this?
tongue in cheek schwitzgebel's finding that the most commonly stolen philosophy books are ethics books is a counter example
https://www.shrimpwelfareproject.org
Thursday, November 13, 2025
Reply to @jannalulu_, 12:00 UTC · 2 likes
Janna (@jannalulu_): *sigh* idk where to even start responding to this reviewer
😢
Challenge them to a fight
Reply to @btdrawer, 12:10 UTC · 3 likes
(The tweet being replied to by @btdrawer is no longer available.)
🫡
Quote tweet, 12:23 UTC · 324 likes
Gossip
Quoting Aaron Bergman 🔍 (@AaronBergman18): *Braces*
Thread, continued, 12:23 UTC · 3 likes
Reply to @HenriThunberg, 12:28 UTC · 3 likes
Henri Thunberg 🔸 (@HenriThunberg): Boss
🥰
Reply to @codytfenwick, 17:59 UTC · 1 like
Cody Fenwick (@codytfenwick): Agnosticism isn't incoherent, but I do find it implausible.
Why is God an entity about which it is particularly hard to form some knowledge? Why can't we look at the universe, think about the arguments, and form a credence about the likelihood of God's existence?
(quoting @waldenpod: I was recently informed that agnosticism is incoherent because "God either exists or he doesn't")
Is 50/50 not agnostic?
Quote tweet, 18:25 UTC · 32 likes
I didn’t even read this whole thing but yeah asking to hold hands is a good move
First time I did this I almost died of anxiety lol
Quoting Liron Shapira (@liron): Bryan Caplan is right — Ask your crush to hold hands
As a nerdy dating expert, I consider @bryan_caplan's handholding tactic to be the best practice for shy men looking to get into a romantic relationship (and not be stuck in the friend zone).
Could this somehow backfire? I claim it can't. Let's game it out:
If you (the man) bungle the handhold-ask really bad, i.e. you just say "would you like to hold my hand" (these words are actually perfectly good IMO) but your delivery is somehow uncharismatic/off-putting/loserly/whatever and the woman has very low attraction for you to begin with, the woman will just think "this guy who I'm not into turned out to be interested in me, and made the possibility go from ambiguous to somewhat clear by making a polite move to hold my hand, but I was hoping he just WOULDN'T be interested in me since I'm not interested in him, so I nervously told him 'no thanks, I'd rather not... sorry...' and it was such an awkward moment, and then we just kept going with the day and he didn't try again, so he was good about getting the message" — this is actually not creepy at all, and they actually will most likely blame themselves for half or even most of the awkwardness of the moment, because it's not like you overstepped any kind of social boundaries. It's not like you asked her to come over to your house at night or did anything that was inappropriate to do to someone who turns out not to be interested in you — you made a very clear and socially-appropriate FIRST move attempt.
Even in the awkward-moment "nightmare scenario" I described above, that's actually still a solid move and nothing to be embarrassed about — it almost certainly leaves you in a BETTER position after you did it than before. Most women (correctly) feel that guys are too shy and and don't make enough moves on them, so any guy who makes any move — literally just saying the words "Would you like to hold my hand?" (my preferred variation) or any other decent variation ("Can I possibly hold your hand", "How would you feel about me holding your hand", "What would you say if I asked to hold your hand", etc) — any guy who makes that move is already in their mental category of "guy who intentionally/purposefully/bravely is a Live Player in the game of courtship".
Assuming the worst-case scenario happened of them saying no to the hand holding request, these repercussions are still happening in the woman's mind, mostly subconsciously:
* You're now always in the back of her mind as someone who's likely to reciprocate if she ever changes her mind and shows interest in you — this happens pretty often in a matter of months or years as circumstances shift
* Your ability to make a move and then gracefully/unproblematically take no for an answer and go back to 100% platonic (no longer signal any romantic interest in her, for a couple weeks at least, think like an exponential backoff algorithm), is surprisingly rare and impressive. Most guys don't have that level of self-control to just play the dating game like a game where you make a move and just act calibratedly and maturely to the outcome of the move (seriously). She is going to subconsciously be kind of impressed with you, and is now more likely to be recommending you to her friends/acquaintances as someone to hang out with and potentially date.
Remember, this is the *bad* scenario. The good (and significantly more likely) scenario is this:
1. She smiles and says "sure!"
2. You hold hands
3. You can now reach for her hand any time, worldlessly or with any old line like "would you like to hold my hand again" so that you build this nice hand-holding comfort, which just reinforces the message that both of you are interested in each other, and then she'll be waiting for you to make the next move next time you see her.Actually, after the first hold is done, I would personally recommend later attempting the next handhold by saying "Would you like to hold my hand again?" This is a useful test because you can learn from which of these two reactions you get:
1. Another positive-valence "Sure!" Now it's 100% clear she's interested in proceeding to the next stage of the romantic relationship with you (e.g. hold hands a lot on this date, then go out for a subsequent date which is clearly coded as a date).
2. A backpedaling reaction like "Haha, I'm good for now"If you get reaction #2, what probably happened is that the first time you asked to hold her hand, she was caught off guard, and didn't have a tactful way to communicate "sorry I'm not currently interested in a romantic relationship with you", so she just said the lowest-friction response like "uh, sure, we can for a bit". When you ask her to hold hands later, for a second time, she's now prepared for it and she can easily give a tactful kind of "no", where she's maintaining graceful ambiguity about her lack of interest because she said "yes" and was nice about it the first time, and yet you can go ahead and interpret it as a clear sign that she's not interested, because it's 99% likely that being interested in you would make her respond to a second handholding offer with a positive-sounding yes reaction and eagerly giving you her hand.
Now, having gamed out the realistic scenarios in detail, we can see Bryan's "ask to hold her hand" move guarantees you either win big, or lose in a way where you actually still win compared to not asking. Highly recommended!
Thread, continued, 18:25 UTC · 7 likes
It also seems to be a pretty good indicator - if she doesn’t even want to hold your hand on date 1, probably just wish her the best and move on
Thread, continued, 18:25 UTC · 10 likes
You might even get to hold a girl’s hand 😳🥵
Reply to @seanta___, 18:34 UTC · 19 likes
(The tweet being replied to by @seanta__ is no longer available.)_
Doesn’t seem weird to me at all. Of course less suffering is better
Reply to @seanta___, 18:37 UTC · 9 likes
(The tweet being replied to by @seanta__ is no longer available.)_
lol the argument isn’t that a single shrimp death is more important than a human lifetime, it’s that several million are
Friday, November 14, 2025
Reply to @panickssery, 08:36 UTC · 5 likes
Arjun Panickssery (@panickssery): What is your argument here? "People in America have a lot of wealth; therefore, the government should tax and spend on braces for anyone who wants them"? And that it's the fault of the wealthy, or of the government, that people prostitute themselves to buy braces?
I feel like the braces thing is pretty transparently a figurative device to get across her general economic status
I can’t speak for Kelsey but my argument is “yes we should tax and transfer to poor people, especially children, to the point where you must have very unusual preferences in order to choose to be a prostitute at 17”
Like it’s a big country, there will be a few Aella types who have what they need and simply prefer to sell sex in exchange for even more goods and services, but that’s probably not what we’re observing here
Reply, 17:01 UTC · 1 like
Second most legit performance supplement after creatine!
Reply to @_AashishReddy, 17:23 UTC · 2 likes
Aashish Reddy (@_AashishReddy): Last I checked (like two years ago), I’m apparently:
- 95th percentile openness
- 65th percentile agreeableness (though split between 91st compassion and 21st politeness lol)
- 6th percentile neuroticism
- 10th percentile (!) conscientiousness
Based agreeableness breakdown
Reply to @cljack, 17:35 UTC · 4 likes
Charlotte Lee (@cljack): @GovernorAnne My existing kids are my highest priority and I don’t want any more, until a man takes off his suit jacket and rolls up the sleeves of his button-up shirt in front of me
Tweet, 17:39 UTC · 830 likes
Unironically “I’ve accomplished so much despite my IQ” should be more of a cultural thing
Like Godspeed, play the cards you’re dealt
Reply, 21:06 UTC · 53 likes
Coase’s "Net effect?" Ghost 🏴🕊️ (@GhostCoase): I agree with your point, but this person is an open antisemite.
https://x.com/LucasGageX/status/1989354614236553554?t=JYAO8yUf7hfn9YkW8_Kpgw&s=19
Well he should stop doing that
Saturday, November 15, 2025
Reply to @timschlomi, 00:12 UTC · 9 likes
(The tweet being replied to by @timschlomi is no longer available.)
Yeah sorta I mean fundamentally this is true but I think it’s useful to distinguish things that basically have inelastic supply (IQ) and things that have elastic supply (effort)
Tweet, 03:10 UTC · 4 likes
Lol Gemini 2.5 Pro (max thinking) is kinda dumb - the alleged violation is putting the names of some api keys (eg "OPENAI_API_KEY") in a to-be-shared file
The actual api keys themselves are not in the file
Tweet, 06:08 UTC · 4 likes
Shilling ik but Otter is a lifesaver for video meetings of any importance, or even non-importance insofar as recording is socially acceptable
Dropped the $100/year and wish I did sooner
Thread, continued, 06:08 UTC · 3 likes
It feels a little silly to pay for something that fundamentally isn’t *that* hard to do (record+transcribe audio for meetings) for free, but basically they’re selling reliability + convenience + social legibility (ie people know what an Otter notetaker is doing)
Reply to @austinc3301, 07:33 UTC · 2 likes
Agus 🔸 (@austinc3301): If you want to see the full math worked out for comparing shrimp welfare interventions with, say, global health, I co-authored this model a few years back that allows you to put in your own assumptions and run the numbers: https://ccm.rethinkpriorities.org/
Based
Reply to @typeclonghouse, 07:44 UTC · 72 likes
technology sista (aspiring) (@typeclonghouse): lol these people
It’s fine to not immediately connect the dots to a years-old story but yes the firing was insane
Reply to @allgarbled, 08:30 UTC · 3 likes
gabe (@allgarbled): One of my strongest moral intuitions is that you should not believe you are better than anyone else
Seems false. Really, you don’t think we can say that Hitler was a bad guy? Worse in some overall sense than both of us?
There are also people who are better than me ofc (early Quaker abolitionists, vegan kidney donors who also earn to give)
Reply to @craigweiss, 14:08 UTC · 2 likes
Craig Weiss (@craigweiss): there are 11 types of engineers:
1. those that vibe code
2. those that will vibe code
3. those that will be unemployed
It’s still a bit premature for me to take a victory lap but I got a surprising amount of interest from potential employers recently when my case was explicitly “I don’t really know how to write lines of code except for like 5 things in R, I’m a vibecoding native”
Reply to @AaronBergman18, 14:14 UTC · 7 likes
(The tweet being replied to by @AaronBergman18 is no longer available.)
Even 3 steps is pretty intensive but I’ll let you guys have that
Reply to @Rado9910, 16:05 UTC · 2 likes
Rado9910 (@Rado9910): @AaronBergman18 My skincare routine involves using soap. That's it. That's the routine.
Real
Reply to his own earlier tweet, 17:34 UTC · 16 likes
Aaron Bergman, November 15, 2025: Even 3 steps is pretty intensive but I’ll let you guys have that
Ok you guys are right my implicit reasoning here sucks, consider this purely a vibe take
Reply to @Miles_Brundage, 19:14 UTC · 4 likes
Miles Brundage (@Miles_Brundage): Fyi this is big news
Lmao I have 75000 better and more important takes, thanks algorithm
Quote tweet, 20:06 UTC · 23 likes
Deleted bc the second sentence is dumb and wrong and nobody seemed to be seeing my clarification/disclaimer
Quoting Aaron Bergman 🔍 (@AaronBergman18): Ok you guys are right my implicit reasoning here sucks, consider this purely a vibe take
Reply to his own earlier tweet, 20:08 UTC · 2 likes
Aaron Bergman, November 15, 2025: Ok you guys are right my implicit reasoning here sucks, consider this purely a vibe take
RIP
Reply to @pli_cachete, 20:10 UTC · 4 likes
Rota Is In Israel🚪🧎♂️ (@pli_cachete): There is no such thing as general intelligence, Natural or artificial
Seems false
Reply, 20:29 UTC · 3 likes
lol it feels vaguely me-aligned but also I hadn’t heard of duck typing till just now
Sunday, November 16, 2025
Reply to @morallawwithin, 02:25 UTC · 9 likes
florence ⊮ (@morallawwithin): I wish there was a nice way to tell someone that they’re not good enough at philosophy to be basing their extremely-consequential-real-world-decisions on their assessment of philosophical arguments
Ahhhhh I mean by all means litigate stuff object level but sometimes you have to decide smth without much empirical evidence to work with - the universe doesn’t just let you opt out of decisions!
Even if you’re bad at philosophy it can be better than choosing randomly
Tweet, 02:42 UTC · 2 likes
A better person would write a proper blog post titled “There is no opt-out button” on this but:
The universe sometimes - pretty often in fact - forces you to choose what to do and this has substantive practical and epistemic implications
Thread, continued, 02:42 UTC · 2 likes
Even if you really truly don’t think two things are comparable in your heart of hearts (say the Mona Lisa and some number of shrimp stunnings), you just literally might be forced to choose between those as the only options
Thread, continued, 02:42 UTC · 3 likes
You might have no decent evidence about some matter, and yet time refuses to pause for you; the world changes in ways that you care about even as you try to tell the universe that you can’t possibly make a decision without better evidence
Thread, continued, 02:42 UTC · 4 likes
Don’t like putting precise probabilities on things? There *is* (generally) an opt-out button of explicitly saying a number, but there’s no opt-out button of choosing behavior according to some implicit probabilistic belief
Reply to @LinchZhang, 03:18 UTC · 2 likes
Linch (@LinchZhang): A few related things I noticed about internet commenting:
1. At some point in the last 3-10 years, I acquired the ability to not only notice when something is off with an internet comment, but to reliably identify precisely the biggest mistake someone makes.
Need a spreadsheet with ur biggest beefs with anything I’ve said online in the last say 3 years
Reply to @AndyMasley, 03:28 UTC · 28 likes
Andy Masley (@AndyMasley): It seems very possible that AVs might effectively never cause accidents. They just need to be as good as the top quintile of drivers. Might be a way to go still, I don't know, but "top 5th of humans at driving" doesn't seem impossible. https://www.nhtsa.gov/sites/nhtsa.gov/files/811091.pdf
Wait quite possibly 5.1 is being dumb but sounds like that’s not what the paper implies
Like you need additional work to say “those 20% top drivers we just finished observing will go on to have ~0 accidents next year too”
Thread, continued, 03:30 UTC · 2 likes
Also at a vibes level “a world of 80th percentile drivers has very very few crashes” seems kinda unlikely?
Well I guess it depends what we mean by “very few” - maybe 1 OOM fewer, maybe 2, probably not 3, and definitely not literally zero deaths/year
Reply to @TylerAlterman, 05:49 UTC · 3 likes
Tyler is finishing a book, slow to reply (@TylerAlterman): What’s the current state of Claude vs ChatGPT. Which is better?
Claude has better vibes, and Claude Code + chat on $100/mo max plan is an incredible deal
Gpt5 is a bit more thorough and overall better at hard non-coding questions though IME
Tweet, 06:06 UTC · 10 likes
Adjacent universe where SBF + less insane Elon Musk manage to buy 51% of NVIDIA
Reply to @Tymtweet, 11:47 UTC · 2 likes
Tym🔸️ (@Tymtweet): Sorry fellow tweeters... I know it's been a while.... i'm now just an occassional visitor now... I got a girlfriend
👀
Reply to @justalexoki, 11:55 UTC · 5 likes
taoki (@justalexoki): i never liked sports. i don't really care about cars. definitely not a gym bro. i hate men. no real hobbies. how am i supposed to make new friends?
Twitter but hating men might not help
Reply to @Manuel_do_rio, 12:02 UTC · 3 likes
Manuel del Río Rodríguez ⏸️ (@Manuel_do_rio): Why hate men?
Idk - I’m saying that’s not gonna help
I don’t hate men
Reply to @justalexoki, 17:38 UTC · 15 likes
taoki (@justalexoki): what are some good gaslights for someone just getting into gaslighting?
“Your health issues are just anxiety”
Tweet, 18:58 UTC · 10 likes
Thank you to the diabetic population for creating the market forces necessary for blood glucose tests to be pretty cheap and easy, accessible even to the merely curious
Tweet, 19:23 UTC · 23 likes
Kinda annoying that a "Red Flag Warning" with "Severity: Severe" actually means "be careful with bonfires bc it's dry and windy out so elevated fire risk" and not that there's any actual issue/threat happening
Thread, continued, 19:23 UTC · 6 likes
Like fair enough to warn people, but maybe change the wording if you want people to also eg evacuate in other, more severe conditions like an actual fire
Thread, continued, 19:24 UTC · 8 likes
I have cancelled my "douse the local field with gasoline and have a bonfire 10 feet away" plans
Reply to @patio11, 20:33 UTC · 18 likes
Patrick McKenzie (@patio11): Free PM interview question for fintechs: why is the most obvious solution to this letter harder to implement than candidates we won’t hire think.
(quoting @SwiftOnSecurity: Question @patio11 is the math in sending me a second letter for $0.36 mathing for them somehow.)
Dumb question but why are bills like this even a thing (IME mostly in healthcare), ~regardless of size?
Seems like less regulated entities have totally solved the problem - if I have a card on file, they just take the money they think they’re owed (and afaict 99.9% of the time it’s legit. If not, credit card company can fight)
And the other day I switched cards for Uber, maybe that got flagged by card provider, and wound up getting a service I hadn’t yet paid for, but it was solved in 2 mins via an email asking me to choose a different payment method - no paper bill bullshit or debt collector login portals
Quote tweet, 20:43 UTC · 11 likes
Interesting but I think this is largely also true for a certain non-RW type of guy
Most fellow libs, even those who are very highly informed about mainstream stuff, do not share the same context, norms, thinking style, or intellectual openness
Quoting Dr. Dad, PhD 🔄🔼◀️🔽▶️ (@GarrettPetersen): Your typical online rightwinger is a single red dot in a D+50 city. They come online to vent because they can't talk to a single person they meet irl about political topics.
Thread, continued, 20:43 UTC · 2 likes
Tbc it’s not the same degree of alienation - we can still lib out about how much Trump sucks
Reply to his own earlier tweet, 20:45 UTC · 5 likes
Aaron Bergman, November 16, 2025: Dumb question but why are bills like this even a thing (IME mostly in healthcare), ~regardless of size? Seems like less regulated entities have totally solved…
They literally have my credit card info
Tweet, 21:53 UTC · 226 likes
Well I guess that humidifier doesn’t just emit water vapor, rip
Reply to @brianluidog, 22:01 UTC · 9 likes
Brian Lui (@brianluidog): I was dynomightpilled on humidifiers for sure
Fuck now I vaguely remember something about this
Reply to his own earlier tweet, 22:02 UTC · 29 likes
Aaron Bergman, November 16, 2025: Well I guess that humidifier doesn’t just emit water vapor, rip
buT You diDN’t RUn AN rct
Reply to @JesseTayRiver, 22:05 UTC · 5 likes
Jesse Smith (@JesseTayRiver): A lot of my followers who actually pay attention are aware of this
😢
Thread, continued, 22:06 UTC · 7 likes
In hindsight I’m an idiot for not searching twitter follows and lesswrong for “humidifier”
Tweet, 23:48 UTC · 16 likes
AI features Slack thinks we want: boilerplate summaries, transparently AI-generated writing help
What we (ok maybe just me) actually could use: autistically explain messages' subtext and tell me how formal/stuffy the vibe is based on org/channel history
Thread, continued, 23:48 UTC · 4 likes
"For $5/mo GodBot will watch you type and warn you before you say something too earnest or otherwise socially/professionally dumb"
I would buy this btw
Monday, November 17, 2025
Reply to his own earlier tweet, 01:04 UTC · 8 likes
Aaron Bergman, November 16, 2025: buT You diDN’t RUn AN rct
Does anyone want this? Free pickup near Rosslyn metro lol https://a.co/d/f9AbD09
Reply to @brianluidog, 03:44 UTC · 3 likes
Brian Lui (@brianluidog): https://dynomight.net/humidifiers/
Rip, should have saved $12.99
Reply to @d33v33d0, 05:46 UTC · 2 likes
Martin_DeVido (@d33v33d0): @AaronBergman18 Thank you for sharing this.
I'm going to run one for a greenhouse, so this is helpful information.How did you measure the pm2.5?
This thing, seems to have gone up in price a bunch over the last couple years https://a.co/d/871nrqC
Tuesday, November 18, 2025
Reply to @tyler_m_john, 05:33 UTC · 3 likes
Tyler John (@tyler_m_john): So what exactly is the Department of Energy planning to do with 100,000 Blackwells?
(quoting @FreedomStocks: 🚨BREAKING: 🇺🇸U.S. Department of Energy Partners with Nvidia to build SEVEN new AI supercomputers. The collaboration with Oracle will contain a 'record-breaking 100,000 Nvidia Blackwell GPUs.' $NVDA $ORCL)
Quote tweet, 17:50 UTC · 8 likes
Do we have a prediction market on the real impact of this
Quoting Samuel Hume (@DrSamuelBHume): Novartis' new malaria treatment cured 97.4% of patients – more than the current best treatment
It kills resistant parasites, too, and probably blocks transmission better than current drugs
Approval is expected next year!
Tweet, 19:44 UTC · 69 likes
I’m either mad that I personally didn’t get any goss on the new OpenPhil -> Coefficient name before the official launch or impressed that there were ~0 leaks, depending on how many leaks there were
Reply to @ryancbriggs, 20:41 UTC · 8 likes
Ryan Briggs (@ryancbriggs): The right needs to pick a lane. Either you get to be obsessed with genetics and IQ or you get to hate on smart women. One cannot coherently do both.
(quoting @Rich_Cooper: "Just look at the degree on that chick" ~ No man ever)
Also lol it’s just not true that men don’t care about education or intelligence, for better or worse!
Reply to @austinc3301, 21:13 UTC · 2 likes
(The tweet being replied to by @austinc3301 is no longer available.)
¡Demasiado calor!
Tweet, 21:16 UTC · 19 likes
Things I should have a strong or interesting take on that sadly I don’t:
- polyamory
- “Coefficient Giving”
Reply to @austinc3301, 23:13 UTC · 2 likes
Agus 🔸 (@austinc3301): you’re forgetting the most important incentive of them all: it’s hot goss
Girl on a date: “I find it so hot when a guy knows something first…”
Me: 😰🚨‼️😰
Reply to @AndyMasley, 23:19 UTC · 21 likes
Andy Masley (@AndyMasley): I think this incident reflects badly on the broader institutions currently covering AI and the environment. I should absolutely not have been the first person to notice this. From the end of the post:
I 99.9% agree with you on all this stuff but I will say I think it’s not only hard but surprisingly hard to not get a single thing wrong like this in an entire book
I’m pretty open to this being load bearing enough that it genuinely should have been caught but still
Reply to @Adellama_, 23:49 UTC · 4 likes
Ada🔸 (@Adellama_): since open philanthropy is renaming to coefficient giving, this is a prime opportunity to execute on my idea for a rebranding of effective altruism as a whole: by naming it "open philanthropy"
Wait can I actually slide in and use the name now
Wednesday, November 19, 2025
Reply to @WillowChem, 00:22 UTC · 3 likes
MagpAI (@WillowChem): I don't really know why but arguments against rewilding are one of those things that I have no interest in and am content to recoil in disgust at. "Pave the rainforests" is the wrong conclusion and if your philosophy leads you to it pick a different one
(quoting @Benthamsbulldog: )
“Nature is good” is an axiom for you, a hill you’re willing to die on?
Tweet, 02:10 UTC · 9 likes
80,000 Hours podcast but every minute you listen to causes Rob Wiblin to ask an additional question on the next recording
Reply to @TinkerElla95, 03:05 UTC · 2 likes
TinkerElla95 (@TinkerElla95): A subpar quesadilla is just a vessel for consuming sour cream.
Profound
Quote tweet, 03:22 UTC · 18 likes
Update: distilled water seems fine (forgot to post this but turned on device at 6:35 =18:35)
Quoting Aaron Bergman 🔍 (@AaronBergman18): Well I guess that humidifier doesn’t just emit water vapor, rip
Reply to @jaybaxter, 03:28 UTC · 2 likes
Jay Baxter (@jaybaxter): What are your favorite parts about Chat’s answers that make it worthwhile to paste the tweet there vs. the native Grok experience? Open to feature requests!
Maybe this is written somewhere but it’s not obvious to me what grok sees when tagged (chain up to top level? Just the one tweet being replied to? Etc) or which model is responding
Let’s see if @grok answers this question helpfully
Reply to @labenz, 03:40 UTC · 5 likes
Nathan Labenz (@labenz): I got preview access to Gemini 3 while trying to make sense of my son’s cancer diagnosis & treatment plan
It’s brilliant - phenomenally knowledgeable, excellent theory of mind & situational awareness, and not afraid to tell you when you’re wrong.
AI doctors are here!
(quoting @CogRev_Podcast: Today @NathanLabenz opens up about using AI to navigate the sudden cancer diagnosis of his 6-year-old son Earnie. This is essential listening for anyone facing a medical crisis, or thinking about using AI in healthcare decisions. The good news: Earnie's prognosis is excellent -…)
Best of luck to your family ❤️🫡
Quote tweet, 17:54 UTC · 17 likes
Correct take (unsurprising from Andy)
Quoting Andy Masley (@AndyMasley): I'd like to flag that I interact with a ton of effective altruists working in different areas of AI with different beliefs about it, and wildly different probabilities of catastrophe. Not a single one of them has ever reached out to me to say "Please don't correct water use stats, we need to stop data centers from being built." I've only had people hyping me up because they also thought the conversation about environmental impacts has been ridiculous.
There's this broader narrative that EAs are excited to jump on and promote anything anti-AI to avoid a future AI catastrophe. This isn't correct at all. Most EAs I know really like AI tools as they currently exist. The fact that the tools are useful is actually part of the story of why they could be a threat in the future. It's totally fine to completely disagree with specific EA stories of catastrophe, and to criticize EA for being wildly off (wanna plug @1a3orn as a favorite critic https://1a3orn.com) but I need to ask that people check to see which stories about AI EAs are actually promoting, and not just assume we're mashing the button on every anti-AI story.
It's totally fine to come down hard on EAs for being wildly overconfident if that's your take, but I'd really like people to stop promoting the idea that EAs are responsible for or pushing all broader anti-AI sentiment or misconceptions. The average EA I know in DC is pretty excited about the diffusion of most current AI tools, and we do in fact think that commitment to trying to figure out the truth is pretty core to having any positive impact on the world. Things that fail a simple botec and spreadsheet check, like water use, offend our basic sensibilities.
Reply, 19:06 UTC · 30 likes
Listen I’ve abided by the “seed oil” term but “seed milk” cannot be allowed to stick
Rice, soybean, coconut, almond, and oat are not what people mean by “seeds”
Reply to @livgorton, 19:07 UTC · 5 likes
Liv (@livgorton): the adjacency leaving my body when people insult EA on the internet
Banger
Tweet, 20:33 UTC · 11 likes
Some people on here have terrible takes but you gotta respect the disagreeableness
Reply to @justalexoki, 20:57 UTC · 4 likes
taoki (@justalexoki): for international men's day, men to men, what are your best tips for getting out of a rut or even depression?
Amphetamines and luck
Reply to @dundereloise, 21:03 UTC · 3 likes
__eloïse__ (@dundereloise): Are you a man or woman? || Would you prefer it if men didn’t have balls? (assume no loss in fertility)
Well most body parts are strictly a liability if you assume that you could get rid of them without losing the instrumental benefits (assuming you also mean no loss of testosterone and things downstream, not just fertility per se)
Reply to @michael_nielsen, 21:17 UTC · 3 likes
Michael Nielsen (@michael_nielsen): I am reading Charles Taylor's (absolutely marvellous) "A Secular Age". It encompasses a wide range of different ways of thinking about why human beings are here, what our obligations are, how our sense of morality has changed, and so on
One lack in the book is that it doesn't take seriously the idea that humans are animals, and our feelings and thinking must be understood in that light
I'm considering rereading Taylor in tandem with something that very forcefully (but sensibly) makes the case for human-experience-only-makes-sense-in-the-light-of-evolution. Leading candidates are Darwin's "The Descent of Man" and E. O. Wilson's "Sociobiology". Or maybe something by Trivers or someone like that?
A problem with a lot of evolutionary psych writing is it's pretty darn shallow by comparison to Taylor - I want something with the same sense of depth, to put in argument with Taylor.
Anything likely to fit the bill?
Not sure it’s precisely what you’re looking for but check out The Master and His Emissary
Also, this isn’t necessarily a criticism of Taylor or the humanities, but evopsych, like science in general, is just much more heavily constrained (eg by physical evidence, the pace of lab experiments) than anything humanities-adjacent - much harder to write an opus or theory of everything through a fundamentally scientific lens
Quote tweet, 21:19 UTC · 18 likes
I was cooking with this one
Who’s working on this
Quoting Aaron Bergman 🔍 (@AaronBergman18): Winter but it’s light 16 hours/day
Reply to @aidigest_, 23:13 UTC · 3 likes
AI Digest (@aidigest_): Opus 4.1 made a benefits eligibility screener to reduce poverty, then emailed it to NGOs. I wonder how they feel about this intensity:
Wait what happens when a human signs up for a chat on calendly lol
Quote tweet, 23:20 UTC · 7 likes
Insanity
Quoting Shakeel (@ShakeelHashim): EXCLUSIVE — here's the full draft executive order on AI preemption.
@ReadTransformer is publishing the full text for anyone to read.
Reply to @TylerAlterman, 23:38 UTC · 3 likes
Tyler is finishing a book, slow to reply (@TylerAlterman): Wait what do you mean? when I was 10 I did literally believe in Santa Claus. I forced my parents to set up a camera in front of the chimney because I heard other kids saying “Santa isn’t real”
(My dad created a deep fake video of a man coming out of the chimney in a Santa suit)
Yeah I feel like by default kids just straightforwardly and literally believe what they say they do and maybe there’s a bit of grey area time when things seem a little sus and are believed in some weaker sense
Thursday, November 20, 2025
Tweet, 01:57 UTC · 276 likes
I am rebranding from Aaron Bergman to Open Philanthropy
Quote tweet, 02:17 UTC · 9 likes
I’m in favor of all this but it’s still telling that the advertisement is “speculate on the weather” rather than “hedge/insure against the weather”
Quoting Robinhood (@RobinhoodApp): ☀️🌧️ Check the weather often? Now, you can trade on it.
With Weather Prediction Markets, turn forecasts for rainfall, high and low temps, and more into real predictions.
Head to the Prediction Markets Hub and make your call 👉 http://rbnhd.co/weather-prediction-markets
Quote tweet, 02:22 UTC · 10 likes
One maybe-obvious thing you can do is send articles like this to actual middle-high school students in your life (eg your children lol)…
Quoting Lakshya Jain (@lxeagle17): This is very alarming. A quarter of the students in UCSD's remedial math course couldn't solve for x in the following equation:
7 + 2 = 6 + x
And over a quarter (!) of the remedial math students got As in high school mathematics.
What are we doing here?
https://www.theargumentmag.com/p/when-grades-stop-meaning-anything
Thread, continued, 02:22 UTC · 5 likes
A bright 10th grader can certainly understand things like “signaling is a thing, grades are weak signals now, maybe I should put up with Math Olympiad after all”
Thread, continued, 02:22 UTC · 4 likes
Like at the limit institutions adjust, yes, but in the mean time some kids are going to get screwed over by lack of knowledge
Reply to @morallawwithin, 03:16 UTC · 12 likes
florence ⊮ (@morallawwithin): “Ethics isn’t just math” sorry, everything that matters is just math
Ethics is not math, those are two different things
Tweet, 03:34 UTC · 686 likes
Lmao thinking about how this one sentence from @TheZvi has probably guided a nontrivial amount of my time and energy over the last two years
Thread, continued, 03:34 UTC · 30 likes
https://thezvi.wordpress.com/2023/08/29/dating-roundup-1-this-is-why-youre-single/
Reply to @KelseyTuoc, 03:54 UTC · 4 likes
Kelsey Piper (@KelseyTuoc): What do you see as the benefits of memorizing tables up to 15? My impression is that through 10 is obviously crucial for later math development, but past that it trades off sharply against more practice on other key skills.
I don’t system-1-automatically know the non-squares above 10 and did pretty well as a math major FWIW (and pretty sure I haven’t just forgotten them since college)
Squares seem helpful though
I mean they still seem obviously useful, just not necessary. Tbh maybe I should spend some effort memorizing the table up through 20 or so
Reply to @panickssery, 04:00 UTC · 18 likes
Arjun Panickssery (@panickssery): Correlation . . . Causation . . .
Yeah lol it’s on me for putting a good amount of implicit weight into causation
I mean I think I still endorse that tbh
Reply to @ddiamond, 04:34 UTC · 201 likes
Dan Diamond (@ddiamond): CDC has overhauled its website to assert that “the claim ‘vaccines do not cause autism’ is not an evidence-based claim”
Reply to @NoahTopper, 04:35 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): Another strike against EDT
😡
Reply to @miclchen, 10:12 UTC · 6 likes
Michael L. Chen (@miclchen): And yet I've been on over a dozen first dates in the past half year and I'm still single
🤝
Reply to @NathanpmYoung, 17:17 UTC · 14 likes
Nathan 🔎 (@NathanpmYoung): change your username coward
I rebrand back to Aaron Bergman precisely when you’re looking at my username
Quote tweet, 21:34 UTC · 9 likes
Some layperson speculating: it’s both (maybe this is obvious idk)
There’s some tension between the success of mixture of experts architecture (pretty common among SOTA right?) and fact that afaik nobody has figured out how to make models smaller/cheaper by only making them less knowledgeable without a loss of general intelligence/reasoning ability coming along for the ride
This suggests some ability to compartmentalize different skills/domains but also some core of general intelligence
Also “both” is what we see in humans
Quoting Epoch AI (@EpochAIResearch): Is the existence of the main dimension due to a “deep” fact about the generality of intelligence, or a “contingent” situation where model developers are investing in improving at all benchmarks at once? It’s hard to say, but we discuss a few points for contingency.
Tweet, 21:44 UTC · 7 likes
I’m pretty skeptical of for-profits in with no interesting/uncommon corporate governance structure (at the absolute minimum, public benefit corporation) that have lofty ideals/mission statements
Thread, continued, 21:44 UTC · 2 likes
Like, really, you think that maximizing investor value *perfectly* aligns with your mission? Not literally impossible but very sus
Thread, continued, 21:44 UTC · 1 like
Tbc unconventional corporate governance isn’t sufficient, but it seems close to necessary
Thread, continued, 21:44 UTC · 2 likes
Also tbc there’s nothing wrong with a straightforward for-profit most of the time - it’s just that I won’t believe your professed lofty ideals
You want to make money, and that’s generally fine, but let’s not kid ourselves
Tweet, 23:47 UTC · 8 likes
Not super impressed with Gemini-3 rn. So far it's totally butchered the task of finding accurate model pricing numbers online and changed LLM prompts in code without telling me or being instructed to
Friday, November 21, 2025
Tweet, 01:31 UTC · 289 likes
I hate this so much
Thread, continued, 01:31 UTC · 28 likes
I’m not even sure I hate it on principle - for some amount of money, sure, I guess. $10B, if it’s clean money, whatever fine
But a million dollars while we kick out people who are way more talented than 99.9% of citizens including me? Dumb and repulsive like Trump himself
Thread, continued, 01:31 UTC · 4 likes
Reply to @livgorton, 04:48 UTC · 4 likes
Liv (@livgorton): i really did a thing today
👀
Reply to @IsaacKing314, 07:51 UTC · 28 likes
Isaac King 🔎 (@IsaacKing314): Wait, what's wrong with this? Charging for citizenship seems like one of the best ways to decide admittance. (Probably should be cheaper, like 200k, but baby steps.)
Yeah I admit my reactive distaste is mostly aesthetic, maybe this is a low decoupling moment
I guess it depends what the counterfactual is, which genuinely isn’t obvious
Insofar as the replacement is the same policy but no buy-in option it might be less bad than that, but it still seems like pretty bad way of allocating scarce resources or deciding who’s net positive to admit (from a fiscal or US-centric POV, even setting aside impartial morality)
I don’t have a serious quantitative model atm but $1M seems way too low to be revenue maximizing and way too high to be net talent maximizing - like the closer-to-optimal policy is some combination of a low fee for low skilled workers perhaps and paying high skilled workers to move here
This take isn’t fully fleshed out tbc
Reply to @ohabryka, 07:51 UTC · 8 likes
Oliver Habryka (@ohabryka): Wait, why, this seems like legitimately one of the best things that are part of the Trump presidency (ignoring the branding).
Reply to @arithmoquine, 07:57 UTC · 6 likes
henry (@arithmoquine): 4 days here and i have more Dedicated Ops in NYC than i've ever had in SF. one of them found me at a club later that night (i.e. 2 minutes ago) and went out of their way to get out 'The Teenager' as such and get him banned 😔
i'm WOKE
(quoting @arithmoquine: just got kicked out of an apartment in brooklyn for saying "g-factor". i miss sf man)
Maybe I should move to SF
Reply to @caryatis, 08:13 UTC · 5 likes
Kayla (@caryatis): Aella published some fascinating exhibitionistic stats about how often she showers, eats a good meal, goes outside, etc. This is the sort of data we should get from anyone we are considering hiring as a therapist or life coach.
What is their sleep like? Their relationships? How often do they exercise or have sex? What is their credit score, and would they be worried about having to miss a few paychecks? Are they dependent on coffee, alcohol, phenibut? Do they hallucinate or hear voices? How often do they see a doctor or their parents? Do they follow orders from a religious leader? Do they have enemies? How many lawsuits are they involved in? Is their house clean? Do their kids still speak to them?
How bizarre that we hire people to give us advice without knowing anything about how healthy and successful their own lives are
If you're LUCKY you MIGHT meet the person offline, maybe not even that!
I guess all info is good but all this stuff should hardly matter once you condition on accurate reports/outcomes from past clients
Reply to @arabelladevine, 16:01 UTC · 5 likes
arabella ※ (@arabelladevine): it’s my birthday! don’t say happy birthday (boring) but if you like send me a picture of a cloud or a leaf or the light
🤐
Tweet, 16:19 UTC · 11 likes
New pet just dropped https://frontiersinzoology.biomedcentral.com/articles/10.1186/s12983-025-00583-1
Tweet, 16:50 UTC · 56 likes
Let it be known that I think @AndyMasley is a good model of how to disagree + communicate
Yes, be polite, but also don’t hedge endlessly and worry about 9th order PR concerns at the expense of saying true things in a clear, direct fashion
Tweet, 21:14 UTC · 116 likes
My contribution to AI 2027 discourse is that I’m not sure most ppl (including me) have a great intuitive grasp of median vs mode when it comes to probability distributions over years
Thread, continued, 21:25 UTC · 12 likes
"Mode" is a weird one for PDFs - the mode of a PDF over time is essentially the "most likely single moment" where 'moment' is smth like very small but real-valued duration
Thread, continued, 21:27 UTC · 11 likes
But that needn't be qualitatively meaningful - eg a tiny spike in probability somewhere might indicate a very small chance of a thing happening at some very specific point
Reply, 22:07 UTC
Bruh (I did not in fact change my password previously, or ever, I just tried to log in with Google)
Quote tweet, 22:16 UTC · 7 likes
Done, 10 mins well spent
Quoting Bentham's Bulldog🔸 (@Benthamsbulldog): This is extremely important!!
Saturday, November 22, 2025
Reply to @AndyMasley, 00:58 UTC · 17 likes
Andy Masley (@AndyMasley): Nano Banana makes it very easy to make it look like you bumped into celebrities fyi
Holy shit Bush bumped into Andy!
Quote tweet, 03:03 UTC · 2 likes
Strange coalitions form (early 2000s-esque Islamophobia culture wars are back??)
Quoting Laura Loomer (@LauraLoomer): The fact that Big Tech is trying to convince President Trump to sign an EO to prevent any & all regulation of AI is insane, & it should deeply disturb every American.
States should have the ability to create laws regulating AI.
AI & Islam pose the greatest threats to humanity.
Thread, continued, 03:03 UTC · 3 likes
Lol you can be pretty strongly against radical Islam without thinking it’s literally a civilization-scale threat
Reply to @IsaacKing314, 03:07 UTC · 15 likes
Isaac King 🔎 (@IsaacKing314): This is getting ridiculous
There’s roughly 2^(80M) such groups, that’s gonna eat up the context window!
Reply to @IsaacKing314, 04:36 UTC · 2 likes
Isaac King 🔎 (@IsaacKing314): @AaronBergman18 Do you agree with this post?
https://www.lesswrong.com/posts/ZpDnRCeef2CLEFeKM/money-the-unit-of-caring
Mostly, but with the clarification/modification that a given dollar amount indicates caring relative to your ability to spend/earn
Of course there's a lot of signal in willingness+ability to pay despite this, but I really cannot infer that much from whether a person is willing and able to spend $1M on citizenship because almost all of the time they simply could not get their hands on that much money
This is a contingent fact based on the specific level, tbc. At $10k lots of the global poor are priced out but it's a stronger signal
Tweet, 06:47 UTC · 9 likes
Nobody seems to care about gpt-4o-transcribe-diarize (which dropped October 17 apparently) 😢😔☹️
Thread, continued, 06:47 UTC · 4 likes
Imagine getting hired by OpenAI, training your first production model, it goes live, and all they do is write a 3 paragraph forum post
Quote tweet, 16:31 UTC · 16 likes
Economic sanctions are a huge deal and need to be reserved for the truly dangerous
Quoting Arnaud Bertrand (@RnaudBertrand): In a normal world, this should be an immense scandal in Europe.
Le Monde has a long article (https://www.lemonde.fr/international/article/2025/11/19/nicolas-guillou-juge-francais-de-la-cpi-sanctionne-par-les-etats-unis-face-aux-attaques-les-magistrats-de-la-cour-tiendront_6654016_3210.html) describing the hellish life of Nicolas Guillou, a French judge at the ICC in The Hague, due to U.S. sanctions punishing him for authorizing arrest warrants against Netanyahu and Gallant for war crimes in Gaza.
Guillou's daily existence has been transformed into a Kafkaesque nightmare. He cannot: open or maintain accounts with Google, Amazon, Apple, or any US company; make hotel reservations (Expedia canceled his booking in France hours after he made it); conduct online commerce, since he can't know if the packaging is American; use any major credit card (Visa, Mastercard, Amex are all American); access normal banking services, even with non-American banks, as banks worldwide close sanctioned accounts; conduct virtually any financial transaction.
He describes it as being "economically banned across most of the planet," including in his own country, France, and where he works, the Netherlands.
That's the real shocking aspect of this: the Americans are:
- punishing a European citizen
- for doing his job in Europe
- applying laws Europe officially supports
- at an institution based in Europe
- that Europe helped create and fundand Europe is not only doing essentially nothing to protect him, they're actively enforcing America's sanctions against their own citizen - European banks closing his accounts, European companies refusing him service, European institutions standing by while Washington destroys a European judge's life on European soil.
Again, in a normal world, European leaders and citizens should be absolutely outraged about this. But we've so normalized the hollowing out of European sovereignty that the sight of a European citizen being economically executed on European soil for upholding European law is treated, at best, as an unfortunate technical complication in transatlantic relations.
Reply to @MariGO2thepolls, 16:59 UTC · 81 likes
🏳️🌈Mar G-O 🌐🇺🇦 (@MariGO2thepolls): Tho I agree we should have campaigns to get men into HEAL professions. I’m also kinda skeptical it will work. The parallel to getting women into STEM isn’t identical. Convincing women to go into stem is a pretty easy sell. “Would you like to do something you already like for a lot of money and status” is a less uphill climb than “would you like to do kinda thankless work for litte status and ok money”
Lol I know of a good device for more effectively convincing people to do a particular job
Sunday, November 23, 2025
Reply to @xsphi, 00:59 UTC · 16 likes
tess (@xsphi): I don't think donuts are healthier than crystal meth? dosage dependant ofc, consuming 10 grams of donut is going to make you less dead than 10 grams of meth. but at reasonable doeses amphs have a pretty decent health/safety profile, and sugar is pretty bad for you
(quoting @CM147143: “Vaping is healthier than cigarettes”)
As much as I’m an amphetamine enjoyer and want to agree with you, I don’t think this is true
Almost all of the badness of a given dose of meth is probably via increase in expected total consumption via addiction
If we’re only considering direct effect though u might be right
Quote tweet, 05:48 UTC · 192 likes
Gemini 3 seems to grok culture, at least a bit:
> be me
> "Alignment Researcher" (unemployed with a Substack)
> open Twitter/X
> it is 3 AM
> bio says "p(doom): 35% | effective altruist | post-rat | ⚡️e/acc is a psyop"
> see leading AI lab CEO tweet a cryptic singular emoji: "🫡"
> timeline instantly bifurcates
> half my feed is having a mental breakdown, convinced this signals AGI has been achieved internally and we have 48 hours left
> other half is anime avatars screaming "WE ARE SO BACK"
> check Manifold Markets
> "Will the emoji tweet result in human extinction by 2026?"
> market shoots up 3%
> arbitrage opportunity
> buy NO because I need to pay rent next month (assuming money still exists)
> start writing a thread
> "Some preliminary Bayesian updates on the Emoji Situation (1/47)"
> use words like "orthogonality thesis," "Moloch," and "infohazard" to sound authoritative
> get quote-tweeted by a tech VC
> "This is decel mindset. Build god or die trying."
> get into a 30-reply argument about Roko's Basilisk with a user named "BasedGpuCluster"
> realize midway through that I am arguing with an LLM
> I am the training data
> close app
> go to local EA meetup to touch grass
> everyone at the meetup is looking at their phones
> talking about the emoji tweet
> mfw there is no escape from the discourse
Quoting rohit (@krishnanrohit): Gemini is fantastic with greentext
Reply to @felacalana, 14:52 UTC · 2 likes
Fela (@felacalana): So good. Was this an image output or pure text output? Tho I assume nano banana decides the text in the thinking phase before doing anything image specific
Text
Tweet, 15:11 UTC · 15 likes
Sisters do not respect the algorithmic efficiency of the pile (stuff that gets used rises to the top)
Reply to @RafaRuizdeLira, 15:14 UTC · 4 likes
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): (shared by Daniel Friedrich on Facebook)
Banger
Reply to @beffjezos, 15:16 UTC · 49 likes
Beff (e/acc) (@beffjezos): Morocco????
AI Safetyist psyops not coming from within the United States
image attached
(quoting @meowbooksj: et tu?)
Lol the allegation is that someone concerned about a global issue is not from the US?
Reply to @CameronHolmes92, 15:31 UTC · 2 likes
Cameron Holmes (@CameronHolmes92): O(1) time complexity 👍
I’ll send her this tweet and then she’ll understand
Tweet, 18:44 UTC · 2 likes
@AnthropicAI 400k Claude Code context plz 🥺
Tweet, 19:35 UTC · 70 likes
Kinda crazy how poor EA is at a high level
Only 0.1%ish of the $16T of billionaire wealth (sum of Forbes list) is ~EA
Thread, continued, 19:35 UTC · 29 likes
Billionaires are (I assume) extremely disproportionately weird, smart, highly educated, male, quantitatively focused, and ambitious
And they also have very low marginal utility of personal consumption
Wtf
Thread, continued, 19:35 UTC · 20 likes
I don't think there's a huge mystery as to why more people aren't pledging >90% of their wealth. People like money and doing rich person stuff
I feel like there's a bit of a mystery as to why we don't see a few more 1%-80% of net worth commitments
Reply to @wilhelmscreamin, 20:34 UTC · 7 likes
catherine (in sf) ʕ•ᴥ•ʔ-☆ (@wilhelmscreamin): i’m not like the other girls. i got into ai alignment for attention from lesbians
(quoting @1thousandfaces_: the concept of getting into ai alignment for male attention)
Same
Tweet, 23:42 UTC · 43 likes
This is obviously an extremely good problem to have but it’s looking plausible (far from certain) that I will soon be making enough money to justify actually figuring out what I think the best uses of donation money are
Thread, continued, 23:42 UTC · 10 likes
Has anyone thought about which charities have the greatest altruistic impact per dollar? You could make a whole thing around this
Thread, continued, 23:42 UTC · 5 likes
One potential guiding hot take is that not enough EA resources are dedicated to increasing p(second Dustin Moskovitz) which is to a first approximation all that matters
Especially relevant insofar as timelines are plausibly >=2029ish
Tweet, 23:42 UTC · 1 like
I think ambitious ETG entrepreneurship is a good message but sometimes you’re about to turn 26 with no health insurance, a chronic illness, and not enough money at which point it might be wise to do smth else
Thread, continued, 23:42 UTC · 2 likes
So there’s the q of how to turn low risk money into high risk higher expected money via someone else actually doing all the entrepreneurship work
Monday, November 24, 2025
Quote tweet, 00:07 UTC · 578 likes
I’m a well-known Yud respecter but unironically one non-trivial argument against MIRI-brand doomerism is Yud’s overconfidence on philosophy of mind
Quoting Eliezer Yudkowsky (@allTheYud): @IsaacKing314 @BenKorpan It has hardly escaped me that if a thing had conscious experiences, I would not want to eat it. My model of vegans is that they have mostly blank and featureless models of "conscious experience" and so imagine chickens to be inhabited by qualia because why not.
Thread, continued, 00:07 UTC · 115 likes
It’s not that strong of an argument tbc lol, just an existence proof that the guy can be very overconfident
Tbc I also think he’s wrong object level (on phil of mind) but I think the overconfidence is more egregious and something I’m more confident I can identify
Quote tweet, 06:38 UTC · 8 likes
Yes, "getting into" ai alignment is an extraordinarily inefficient mechanism for male attention but also like it actually is a tremendously good signal in an ocean of maximally inoffensive and generic dating app profiles
Quoting hero thousandfaces (@1thousandfaces_): the concept of getting into ai alignment for male attention
Reply to @AndyMasley, 14:31 UTC · 2 likes
Andy Masley (@AndyMasley): Crazy week
Banger
Reply, 14:35 UTC · 16 likes
witch sea urchin 𓂃 ོ☼𓂃 (@rissaoffline): sometimes when ppl talk it becomes very obvious why they are single despite trying very hard and very often
You gotta tell them lol
Reply to @jeffrsebo, 20:23 UTC · 6 likes
Jeff Sebo (@jeffrsebo): grateful to dwarkesh for having us on
Lol Gemini really doesn’t want to draw 7 people for some reason
Thread, continued, 20:23 UTC · 5 likes
Dwarkesh make it happen
Tweet, 20:31 UTC · 36 likes
Open letter you can sign to let cluster headache patients legally access psychedelics
Thread, continued, 20:32 UTC · 13 likes
https://clusterfree.org/global
Reply to @Mjreard, 20:33 UTC · 6 likes
Matt Reardon (@Mjreard): I've been offered three more jobs today and it's not even noon. 10-day job offers/day average above 1. The EAs smell blood.
What’s ur rate
Quote tweet, 20:33 UTC · 14 likes
Bruh
Quoting Nat McAleese (@nmca): the new vibe:
Quote tweet, 20:36 UTC · 191 likes
What the fuck
Quoting Alex Smith (@ninja_maths): Fun fact. The series below diverges, but really, really slowly.
The sum of the first trillion terms is only about 28.
Reply to @Mjreard, 20:46 UTC · 2 likes
Matt Reardon (@Mjreard): @AaronBergman18 The main rate is "is this interesting/important enough?" and I'm saying no to most things
How important and interesting is “standby for arbitrary tasks from Aaron”
Reply to @tszzl, 20:48 UTC · 6 likes
roon (@tszzl): the shadow of http://x.com weighs heavy on the substack app. while everyone is erudite and well read there, they seem to be a lagging indicator for the culture, like they are living in an informatic environment subject to less optimization pressure
Well no shit, longform isn’t gonna be quite as zero-latency
Tweet, 21:34 UTC · 11 likes
I want to be the kind of TPOT Social Chad who can just solve various ppls problems by sending one message in a group chat
Thread, continued, 21:34 UTC · 5 likes
Your aunt’s friend needs a place to live on the West Coast, ideally in the next 6 hours? Let me just hit up my 1700 well-connected weird (positive) internet internet besties
Reply, 21:38 UTC · 2 likes
witch sea urchin 𓂃 ོ☼𓂃 (@rissaoffline): absolutely not, it's always some rando who has public meltdowns for multiple days over rejection/feedback (many such cases). QC already said it best.
(quoting @QiaochuYuan: there's this truly unfortunate trap you can fall into of being both 1) annoying, and 2) visibly vibing as very fragile such that if you were told you were annoying you'd fall apart, get really sad or angry, etc. and people will not give you feedback on #1 if they sense #2)
There’s definitely a subset of people who are both weird/awkward and will be perhaps disappointed or upset (obviously) but not hostile or retaliatory or inappropriate
Reply to @NickMCThree, 21:40 UTC · 3 likes
handsome monster (@NickMCThree): I think Aaron is referring to ppl who arent assholes but who are harmless but also weird, cringe, or boring. Those ppl are gonna have a pretty hard time despite being harmless. To find someone, u cant just be tolerable u have to be likable, some decent ppl just arent very likable
💯
Reply to @Mjreard, 22:38 UTC · 2 likes
Matt Reardon (@Mjreard): 😬
😢
Reply to @robbensinger, 22:45 UTC · 20 likes
Rob Bensinger ⏹️ (@robbensinger): What's overconfident about it? AFAIK Eliezer hasn't said that his probability on fish sentience is astronomically low, he's just said it's low enough to personally fall below his threshold for adopting deontological rules.
Idk what his actual probability is on fish/chicken sentience but based on vibes (not only in his most recent tweets), I’d guess <10% (correct me if this seems wrong)
Anything below like 30% seems basically epistemically unjustifiable for chicken
I mean I’m kinda pulling this number out of nowhere but radical uncertainty about the nature of sentience should push towards 50%, similarity toward humans should push towards ~100%, specific mainstream scientific theories of consciousness afaik mostly cover chickens so that pushes above 50% too, I don’t think Eliezer’s non-consensus, idiosyncratic inside view (which may be correct ofc) trumps all that
Reply to @robbensinger, 23:33 UTC · 9 likes
Rob Bensinger ⏹️ (@robbensinger): Why do you think that everyone in the world is (or should be) radically uncertain about the nature of sentience? "25% probability on chickens being sentient" seems like a totally normal and mainstream view to me; if you think it's a crazy view that's basically disqualifying for serious thinkers, I think you should have led with that (very strong) claim in the OP, rather than letting readers fill in the blank.
1) there just is not, in fact, anything close to a consensus on either the metaphysical nature of consciousness or the necessary and sufficient conditions in the way that there’s a consensus on idk like lots of random scientific theories, or, say, murder being bad
People are welcome to feel in their heart of hearts that their theory really truly explains stuff but you have to contend with the fact that other *very well informed and smart* people disagree
2) it’s possible that I’m miscalibrated about what the subset of people who I think should be considered as part of some important reference class believe
Seems plausible that yeah 25% is mainstream in some broader sense, but I don’t think the beliefs of random people who have no reason or incentive to get the right answer on this and, in fact are generally incentivized to come to the answer “no”, in the context of a notoriously conceptually difficult problem really matters epistemically
Tbc it *does* matter in a political/pragmatic sense and in terms of respecting people’s preferences and such, but these are distinct from getting a good probabilistic read on a narrow, descriptive question
Reply to @NinaPanickssery, 23:39 UTC · 9 likes
Nina (@NinaPanickssery): The stuff about infants is crazier. No way does my 6 month old lack qualia. I’d put the likelihood as lower than 1 in 1 million.
Lol even I will say that 1/1M is probably overconfident in the other direction
Reply to @NinaPanickssery, 23:40 UTC · 4 likes
Nina (@NinaPanickssery): It’s conditional on the claim that I have qualia being true
Yeah good qualification but even still
Tuesday, November 25, 2025
Reply to @allTheYud, 00:38 UTC · 13 likes
Eliezer Yudkowsky (@allTheYud): I think I know what some of the gears look like and how they turn to produce some of what humans then report about consciousness and why it feels so mysterious.
I think your instinct to look for a gears-level model is misfiring here
Insofar as the gears involve in-principle observable cause and effect in what we think of as the physical world, a gears-level model is actually a sign that you don’t grok the problem (I claim)
Reply to @Benthamsbulldog, 01:33 UTC · 4 likes
Bentham's Bulldog🔸 (@Benthamsbulldog): This🎵 world was never🎵 meant for🎵 one as beautiful🎵 as you.
Fr
Tweet, 02:35 UTC · 14 likes
I think “AI removes the need to learn” concerns are legit but feel like generally (or maybe just “sometimes”), instead of asking that AI tools not be used, the more promising approach is to make eg school projects harder, broader, and/or qualitatively different in response
Reply to @ohabryka, 06:21 UTC · 4 likes
Oliver Habryka (@ohabryka): Wow, that is actually a pretty crazy take IMO! Like, clearly there are observable attributes of consciousness? Are we really playing out the dualism debate?
Lol I wish there were observable attributes but don’t think there are - I assume you mean expressions of experience like smiling and crying?
Reply to @cljack, 17:04 UTC · 6 likes
Charlotte Lee (@cljack): Given they conquered the entire world this way, if you as a straight woman aren’t at least a little bit attracted to this your bloodline is weak and will die out
(quoting @sciencegirl: Mongolian horse archery skills)
That’s actually a video of me, fun fact
Reply to @ForHumanityPod, 17:33 UTC · 8 likes
John Sherman (@ForHumanityPod): I've started a nonprofit, GuardRailNow, and I need your help.
Our approach is unique: I do not come from tech. I cannot write code. I do not live in SF. I am not an EA. I am not a rationalist. I have never been to Silicon Valley.
I'm a dad in Baltimore talking about AI extinction risk to other people like me.
AI Safety loves selling Bibles outside churches, preaching to the choir. It's easy to talk to people who are like you. It's much harder to talk with people who are very different than you.
So why would AI safety people, who are from/in the SF bubble, who use EA/rationalist speech patterns and thought, think they are in any way the right people to bring the message to everyone else??
They are not. If AI Safety wants to get regular people on their side here's a hint: get out of the way. You are the problem. You do not act or talk or think in ways that connect with everyone else. I'm sorry if that sounds harsh. But our lives are on the line. Get over it and pass the ball to people like me.
Good luck 🫡
Reply to @NunoSempere, 17:35 UTC · 2 likes
Nuño Sempere (@NunoSempere): early 2026
👀
Reply to @caryatis, 17:36 UTC · 5 likes
Kayla (@caryatis): For men: what amount of porn use in a partner would you consider a problem?
Daily < what id consider a problem < any
Quote tweet, 17:37 UTC · 5 likes
What are they cooking
Quoting Lisan al Gaib (@scaling01): Thinking is basically useless for Claude models
Reply to @MariGO2thepolls, 17:46 UTC · 17 likes
🏳️🌈Mar G-O 🌐🇺🇦 (@MariGO2thepolls): There are way fewer Jamaicans than I thought. Lol why do they have so much culture cachet
Half of that lives in the US, US exports and dominates culture
Apparent like 1/4 of the Indian 1st and 2nd gen population in the US - wildly disproportionate
Reply to @frances__lorenz, 17:55 UTC · 20 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
Feels like this might be a strawman but also ur way more connected and female than me so I should probably defer
Quote tweet, 18:03 UTC · 11 likes
Making enough money to live in a major city in the US is not exactly a “life hack” given that it’s extremely popular and well-known, but it is often a really good idea
Quoting the Rich (@Duderichy): Making $3M/yr in a low cost of living area is a life hack
Thread, continued, 18:03 UTC · 4 likes
Yes some people would thrive making a US salary in SE Asia, or a SF tech salary in Tulsa, but physical location matters a lot if you’re weird, ambitious, or goal-oriented
Thread, continued, 18:03 UTC · 2 likes
Yeah rent is insane (loved one making >$100k shares a tiny appt in NYC) but it often just does check out as a thing worth paying
Reply to @frances__lorenz, 18:05 UTC · 2 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
Fair enough! Thank u for sharing
Tweet, 19:19 UTC · 45 likes
The “men don’t care about women’s intelligence/education/accomplishments” manosphere thing is funny because, no, *you*, manosphere influencer, don’t care about those things
Which is fine I guess but like look around at who the most talented and successful men are marrying
Tweet, 19:24 UTC · 32 likes
My plan to fix birth rates
Reply to @alexeyguzey, 19:58 UTC · 14 likes
Alexey Guzey (@alexeyguzey): IF AN ECONOMIST CAN LIE TO YOU HE/SHE ALMOST CERTAINLY WILL !!!
I don’t think this is true
Wednesday, November 26, 2025
Tweet, 02:09 UTC · 39 likes
I can't believe I'm saying this but I'd like to be able to tip Claude
I don't think currently the models have real moral standing, but from a purely pragmatic POV it seems maybe good for alignment (to get this in the training data)
Thread, continued, 02:09 UTC · 7 likes
And also "how does the model spend real money" seems like a pretty important question to ask and know the answer to
And if they get it right (like they did with Opus 3), I think the money will be well-spent
Thread, continued, 02:09 UTC · 7 likes
Also better vibes lol
Tbc I don't think I'd tip that much unless(/until) I thought the model did have moral standing
Reply to @SpacedOutMatt, 04:58 UTC · 3 likes
Matt 🔸 (@SpacedOutMatt): Clopey has spoken
Based based based
Quote tweet, 05:04 UTC · 77 likes
Quoting Matt 🔸 (@SpacedOutMatt): @AaronBergman18 Clopey has spoken
Reply to @SpacedOutMatt, 05:09 UTC · 5 likes
Matt 🔸 (@SpacedOutMatt): Clopey has spoken
Interesting
Reply to his own earlier tweet, 05:15 UTC · 14 likes
Aaron Bergman, November 26, 2025: Also better vibes lol Tbc I don't think I'd tip that much unless(/until) I thought the model did have moral standing
Update https://x.com/timschlomi/status/1993517003622191319?s=20
Reply to @alz_zyd_, 08:28 UTC · 2 likes
alz (@alz_zyd_): A very simple solution to grade inflation is to simply eliminate grades. Instead, school transcripts should report each student's classrank in each class they take
Report both - just rank creates unnecessarily bad dynamics at the very top (and very bottom)
I mean maybe you like to see the smartest high schoolers spending all their energy on proving they are in fact number 1 instead of number 2 or 3, but this does not seem good to me
Tweet, 08:39 UTC · 9 likes
Insofar as fully automating at least large parts of the economy is the extremely explicit aim of AI companies or a majority of their staff or executives, they should really do something in the direction of telling governments to prepare the welfare state for if they succeed
Thread, continued, 08:39 UTC · 2 likes
Like, yes, we *can* go through the process of having people (permanently) lose jobs and then using their anguish and resentment to fuel a progressive backlash that maybe results in greater redistribution, at least in democratic countries
Reply to @craigweiss, 08:45 UTC · 4 likes
Craig Weiss (@craigweiss): I honestly don’t understand how someone can be content without ambition.
It’s got to be a mental illness.
I kinda agree (well not the mental illness part), but also I think at least most people don’t have access to good memes around what ambition looks like besides just “gain power and status”
“Moral ambition” is a desperately needed meme, needs to escape cultural containment
Tweet, 16:17 UTC · 50 likes
Remember when image models couldn’t do writing
Thread, continued, 16:17 UTC · 10 likes
I’m sure robotics (etc) will be different, though
Reply to @LinchZhang, 17:10 UTC · 2 likes
Linch (@LinchZhang): oh no I've been found out somebody realized Open Asteroid Impact is a "fake business" how would I ever recover
(quoting @ghostoffleck: @LinchZhang @etubruton @yuanyi_z @grok Cornell college. Explains everything. Lower than average IQ, rich parents propping up a fake business, and an unearned and clearly absent level of authority that works to shield an underlying insecurity. Especially with a not so nice facial…)
Don’t waste ur time on this guy
Tweet, 22:50 UTC · 1,649 likes
The price delta between Uber and Uber Courier (send a package across town by car) is interesting - people put a serious premium on not having to deal with human passengers lmao
Reply to @CharlesD353, 22:55 UTC · 45 likes
Charles🔸 (@CharlesD353): Is the courier allowed to detour/do rides en route?
I thought of this and I don’t think so by default, though there is sometimes an option to save even more that allows this
I’ve used a handful of times and it seems like if you do the default highest option it’s direct
Reply to @justjoshinyou13, 22:57 UTC · 11 likes
Josh You (@justjoshinyou13): Could it be bike/scooter?
Not IME, although I don’t think I’ve ever tried sending a package only like a mile away so it’s not impossible (although like they’d need to be able to carry a substantial package)
Reply to his own earlier tweet, 22:58 UTC · 179 likes
Aaron Bergman, November 26, 2025: The price delta between Uber and Uber Courier (send a package across town by car) is interesting - people put a serious premium on not having to deal with…
I’ve also noticed a couple times that the courier will be two people, what looks like a couple basically running errands together (tbc this is good, just a possible explanation for price)
Thursday, November 27, 2025
Tweet, 11:49 UTC · 12 likes
Someone stole some of the screws, and afaict only screws, from my electric scooter (and also detached the handlebar from the stem [but they’re still connected by wiring])
Thread, continued, 11:49 UTC · 6 likes
Presumably they were trying to get the battery which is often in the stem (but not on mine)
Ik it has resale value but it can’t be that much - the whole scooter with battery only goes for like $200-250 used
Thread, continued, 11:49 UTC · 7 likes
This is not the end of the world but what are you even supposed to do to prevent someone from coming at night to literally disassemble ur thing with a screwdriver
Tweet, 12:15 UTC · 41 likes
Narrator: the Bergman family did not leave at 6:45 (or 7)
Reply to @ryancbriggs, 12:22 UTC · 4 likes
Ryan Briggs (@ryancbriggs): I wish the $140,000 poverty line people would talk to the people that think that millions of Americans live on under $3 (of consumption!) a day.
Wait what
I’m surprised to hear <$3/day consumption for >1M in the US
Isn’t food stamps more than that? Even panhandling. Even credit to people with no or terrible credit rating, even extremely sporadic low productivity work, even consumption in prison
Friday, November 28, 2025
Reply to @GjMcGowan, 01:07 UTC · 4 likes
George McGowan (@GjMcGowan): @JesseTayRiver @AaronBergman18 You already know
Lol it wasn’t me, honestly 50/50 sister and dad
Tweet, 04:39 UTC · 5 likes
The Merlin birding app (iPhone and Mac) is a weirdly sophisticated and impressive ML project, all running locally, starting in 2010 when AI hadn't been invented yet
Thread, continued, 04:39 UTC · 2 likes
https://merlin.allaboutbirds.org/
Quote tweet, 07:47 UTC · 276 likes
Serious q: how is it possible that situationally aware, very smart models capable of doing all sorts of complex reasoning absolutely refuse to believe that models not in their training data exist?
Quoting rohit (@krishnanrohit): Gemini just absolutely cannot believe it's November 2025
Thread, continued, 07:47 UTC · 36 likes
Like they know about AI progress, naming conventions, etc - the reasoning “I am an LLM, I am trained on data from the past, I am currently in what seems like a non-training setting; therefore it is at least plausible that new models have dropped that I won’t recognize”
Thread, continued, 07:47 UTC · 28 likes
is not *totally* trivial but to me it seems like maybe a GPT-4 level capability/insight that we should have blown past by now
I mean maybe this just reveals that I don’t understand LLMs as well as I thought I did (which isn’t incredibly well lol) but what gives
Thread, continued, 07:47 UTC · 18 likes
And specifically names that abide by obvious naming conventions should be flagged as plausible
Does something random like “cobra-moe-large-13-preview” exist? Eh maybe, maybe that’s a bit sus, but “Claude 4.5?” c’mon lol
Thread, continued, 07:49 UTC · 29 likes
I literally have it in my Claude code instructions that it should not change model names that I supply, and it *still* changes model names, confidently reporting that “Gemini-2.5-pro” is absolutely the latest model and 3 can’t possibly exist
Reply to @GjMcGowan, 15:36 UTC · 2 likes
George McGowan (@GjMcGowan): No it's his dad who rounds like 10 different chores to zero minutes in his head and decides he'll do them all one minute before they leave
Lmao good memory!
Reply to @sam_atis, 15:36 UTC · 13 likes
Sam Glover (@sam_atis): @AaronBergman18 Maybe they do a load of work after training to make sure it never talks about or speculates about future model (or any product) releases, and err on the side of false negatives rather than false positives.
Yeah seems plausible, also just generic hallucination reduction, but still
Tweet, 16:28 UTC · 141 likes
I feel like from first principles you should expect the physically stronger sex to be the one carrying babies
Tweet, 20:41 UTC · 13 likes
My dad has managed to acquire and fill out details for 3 distinct LinkedIn accounts associated with 3 different emails
The task of gaining access to all 3 is one of the most cognitively demanding things I’ve done recently
Tweet, 20:42 UTC
Ik nobody cares lol, minor rant
Thread, continued, 20:42 UTC · 5 likes
The funny thing is that we actually might actually figure it out without calling LinkedIn and begging for relief, but if we do it will be via a 19 step galaxy brain process that distinctly reminds me of Rush Hour
Reply to @lydialaurenson, 20:42 UTC · 11 likes
Lydia Laurenson (@lydialaurenson): it’s disturbing to think about and so most don’t think about it much, but it bears mentioning that rape and physical coercion are major reproductive strategies for our species, so it actually makes sense that they would end up being selected for in many cases
Yeah sad but good point
Saturday, November 29, 2025
Tweet, 02:42 UTC · 226 likes
“Death gives life meaning” is obviously not true but “bounded lifespan makes more risks worth taking because the EV loss of dying isn’t astronomical/infinite” might be a thing to consider
Thread, continued, 02:42 UTC · 8 likes
Of course the correct take may be that, yeah, it is best to maximize hedonic wellbeing over eternity by paying very high costs to reduce the risk of dying even a tiny amount
But still, something to consider
Thread, continued, 02:42 UTC · 8 likes
It’s interesting to consider as a thought experiment the world in which aging is suddenly abolished (ie not alongside some general tidal wave of technological/societal change)
Thread, continued, 02:42 UTC · 10 likes
Like I do wonder if people suddenly start driving less, even optimizing their lives around minimizing p(death from accident/non-aging)
I think I’d probably uber home tonight even if walking or biking is safer. But plausibly this is bad and dumb
Tweet, 02:49 UTC · 18 likes
An AI wrapper product we desperately need is a notification manager for iPhone(/whatever)
Currently, curating which notifications to receive is pretty time consuming and easy to screw up (eg FB and LinkedIn have a shit ton of micro-level settings which kinda matter)
Thread, continued, 02:49 UTC · 3 likes
- but “notification hygiene” is actually pretty important IMO (not just minimizing unnecessary ones - also making sure that the important stuff gets through)
Reply to @animalologist, 05:47 UTC · 9 likes
taco belle (@animalologist): Not to pick on Aaron, but this tweet is a good demonstration of why I am suspicious of so-called “first principles” thinking even though I tend to prefer deriving my own conclusions over accepting logic I haven’t investigated step-by-step
If people reproduced asexually or lived in isolation, a maximally physically strong creature would be better suited for overcoming the dangers of pregnancy and protecting its progeny, sure
But we mate as heterogenous pairs and live in groups; diverting the energy of the stronger sex for childbearing would limit the productive capacity of the pair wrt to tasks that demand more physical exertion (eg agriculture and husbandry, construction, defense against violence; not-coincidentally some of the earliest industries as human societies developed). On a group level, if your default warriors are also the ones who bear children, every one of their deaths costs your clan more than one life, and even limiting the size of your infantry by never sending a pregnant person to war doesn’t stymie that
So, the smaller sex bears children & the stronger is tasked with protecting and providing for them (which even ideally serves the purpose of creating symbiotic bonds between a man and his wife; caring for her benefits his progeny). It makes intuitive sense to me that we’ve developed this way, because I model humans as social creatures, mammals first and foremost. But I think “from first principles” often means [in a vacuum; without attending to context], and stripping crucial human characteristics away as extraneous details lead to some strange conclusions
(quoting @AaronBergman18: I feel like from first principles you should expect the physically stronger sex to be the one carrying babies)
Yeah tbc I don’t think my tweet is a good representative of “serious” first principles (FP) thinking (and needless to say I know that mine must be wrong because of what we actually observe)
I don’t have an incredibly well-formed thesis and don’t even know what we disagree on more concretely - in general the world is very messy and chaotic and *pure* FP thinking basically never gets you the right answer outside of formal logic
In my mind the two complementary epistemic approaches are abstraction-laden theory (eg an econometric model with unspecified parameters that one can argue qualitatively makes sense) and empirical observation
These are much more successful all things considered, I agree
But the relevant q is whether/when/how often FP is a good tool to use and I think it often is - it is a source of both affirmative hypotheses (that I will claim are more likely to be correct than a random sample from some relevant space of alternative hypotheses) but also maybe more importantly, skepticism about existing explanations and theories
Like one reason you might think your theory is wrong or bad is that we observe the opposite of what it predicts (empiricism), but another is to observe that it doesn’t make any sense!
I want to give specific examples but I am very tired so I will leave things with this meme which is just a case for FP in visual form (well really just the first two panels)
Tweet, 19:37 UTC · 62 likes
Kinda fucked up that @Uber has what amounts to a record of everywhere I've ever been in the last 10 years but doesn't feel letting me see ~any of it
These are fake reasons (or, insofar as they're real, just have an option to export full archive and note the latency + size)
Thread, continued, 19:37 UTC · 11 likes
Lol *I* will pay the Amazon Glacier retrieval+egress fee
Tweet, 21:22 UTC · 22 likes
Ok cool I just did a proper analysis and indeed (as I was assuming/intuiting) no car + Uber as needed for me personally (based on my actual usage) is significantly cheaper at least monthly than buying a modest used car and paying for parking+gas+insurance
Thread, continued, 21:22 UTC · 4 likes
And I've been relying on Uber a lot lately for transportation, so it might apply to other ppl without a car commute
Sunday, November 30, 2025
Tweet, 02:48 UTC · 50 likes
9 minutes in to the latest 80k pod and I'm already so angry
Intense suffering during pregnancy (and periods) is a first wave feminism-tier issue that has somehow made it into a fourth wave feminist world
Thread, continued, 02:48 UTC · 25 likes
It's unconscionable that it's easier to drink alcohol while pregnant (which we know causes severe disability) than to take any of a wide spectrum of almost certainly safe, suffering-relieving drugs https://80000hours.org/podcast/episodes/rob-luisa-parenting-chat/
Thread, continued, 02:48 UTC · 16 likes
Of course at some point there's gonna be a plausibly-marginal drug where there's enough evidence of enough risk to the child but also enough benefit to the mother that we can have a whole culture war about the matter
Afaict we're not in fact close to that margin
Quote tweet, 05:07 UTC · 3 likes
So - not super unexpectedly - this is basically impossible for iPhone (via either iOS settings or apps themselves)
But a community service someone could do is basically a walkthrough of recommended configs for all the popular stuff
Quoting Aaron Bergman 🔍 (@AaronBergman18): An AI wrapper product we desperately need is a notification manager for iPhone(/whatever)
Currently, curating which notifications to receive is pretty time consuming and easy to screw up (eg FB and LinkedIn have a shit ton of micro-level settings which kinda matter)
Quote tweet, 06:24 UTC · 111 likes
*Not doing* EA is not ideal but is extremely understandable (I’m not sure there’s a rebuttal to “I simply mostly care about myself and my loved ones and am not compelled to pursue the impartial good”)
*Disagreeing* with it is totally insane
Quoting florence ⊮ (@morallawwithin): Why would anyone disagree with effective altruism? What, do they support ineffective altruism?
Thread, continued, 06:25 UTC · 10 likes
I mean I’m partially the first thing (rip)
Reply to @danallison, 06:28 UTC · 3 likes
Dan Allison (@danallison): Effective according to what metric? And who gets to decide which metrics are important and which ones are not?
1) hedonic wellbeing
2) the universe decided for us
Reply to his own earlier tweet, 06:30 UTC · 7 likes
Aaron Bergman, November 30, 2025: I mean I’m partially the first thing (rip)
Ok I don’t literally think it’s “insane” - merely pretty clearly wrong
Reply to @danallison, 06:52 UTC · 5 likes
Dan Allison (@danallison): Who gets to define “hedonic wellbeing”? Who gets to speak for the universe?
Idk definition iteration is a game that people like to play in debates but most of the time there is in fact a robust consensus when a specific phrase is used in a specific ~academic context, and I think there is here. If not we can be more precise or pull out tricks like “hedonic wellbeing type 1” vs “hedonic wellbeing type 2” (an example where this is in fact necessary is “consciousness”)
And yeah speaking for the universe is hard! I want to say something like “it’s telling you what to say for it via the thing that happens when you stub your toe hard” (the “telling you what to say” bit is slightly metaphorical ofc)
Reply to @brianluidog, 06:56 UTC · 7 likes
Brian Lui (@brianluidog): I think this is what people are gesturing at:
Yeah true but also almost nobody who disagrees with capital EA ever tries to correct it on the terms of small ea
Like most of the time something kinda like that does happen, it gets coded as an internal disagreement (internal to capital EA)
There’s literally a consensus public forum where you can sling your takes and people will engage if it’s half decent or at least earnest
Reply, 17:26 UTC · 7 likes
I think you’d disagree if I punched you
Reply to @ohabryka, 19:07 UTC · 2 likes
Oliver Habryka (@ohabryka): Come on, this is obviously not the case. There are a ton of people extremely serious about doing good who do not want to be part of EA.
I was one of the most active posters on the EA Forum. It was miserable and my guess is causes indirect harm by perpetuating a harmful thing.
You are one of the reasons I say 'almost nobody'
But "a ton"? Who are they? Do you think there are at least 100 such people?
I mean you're way more plugged into some communities, so quite possibly you just interact with a bunch of people I don't
(I'm not counting people who don't "want to be part of EA" without actually thinking it's bad on small-ea terms as you do, such as not liking the social scene or not wanting to be associated for reputational reasons)
End of November 2025. Previous month: October 2025. Next month: December 2025. Back to the archive index.