Aaron Bergman's tweets, May 2026
Every public tweet, reply, quote tweet and retweet that Aaron Bergman (@AaronBergman18 on X, formerly Twitter) posted in May 2026 and has not deleted: 461 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: April 2026. Next month: June 2026.
Friday, May 1, 2026
Tweet, 05:12 UTC · 102 likes
The permanent underclass thing seems totally legit to me tbh
There are relatively better - even good - versions of it (namely, even the underclass has abundance in an absolute sense, and a high standard of living)
Thread, continued, 05:12 UTC · 30 likes
But still, in a very literal sense, an entirely bifurcated (or n-furcated) order seems totally plausible
It’s r vs g all the way down
Reply to @halcyon_hazel, 05:17 UTC · 2 likes
❤️🔥 hazel ❤️🔥 (@halcyon_hazel): @vivian39_ like at the bare minimum pescatarianism (argument about fish pain receptors here)
If you’re eating farmed fish at least I think it’s worse
Reply to @thkostolansky, 07:27 UTC · 7 likes
Tim Kostolansky (@thkostolansky): @AaronBergman18 what convinced u of this?
There’s nothing super specific I can point to, just broad understanding of econ and AI
Reply to @thkostolansky, 07:47 UTC · 2 likes
Tim Kostolansky (@thkostolansky): @AaronBergman18
😢
Thread, continued, 07:58 UTC · 5 likes
I feel like the closest thing I have to an answer is an incredibly high level vague understanding of Capital in the 21st Century and now observing what looks like a quite plausible reason that returns to capital will go up and wages go down and a social-intellectual environment in which many people I intellectually respect see the same writing on the wall makes it feel like I’m at lease plausibly not missing some other crucial consideration that makes economic mobility/dynamism robust to AI
Better answer lol
Reply to @pigeonisperfect, 15:29 UTC · 2 likes
pigeon (@pigeonisperfect): coo
@livgorton
Reply, 15:38 UTC · 5 likes
Taylor Lorenz (@TaylorLorenz): Doomerism *is* inorganic though. That’s why massively well funded groups are spending millions pushing doomer messaging on social media. Here’s a WaPo piece on some of this https://www.washingtonpost.com/technology/2026/04/18/ai-doom-influencers-safety/
(quoting @ShakeelHashim: The immense irony of this piece alleging that “doomerism” is inorganic, when the entire piece is basically an undisclosed press release from the pro-industry super PAC and its 501(c)(4)…)
If nonprofits someone donates money to posting on social media is “inorganic” then “inorganic” just isn’t bad
What would “organic” look like to you?
I’m a doomer, at least insofar as that term is generally used. I just tweet, ofc maybe I’m the gullible sheep but of all my faults I tend to think that one’s not the issue
Reply to @ScottFarbman, 15:49 UTC · 2 likes
Scott Farbman (@ScottFarbman): @jasonc_nc "But requiring every building, even a small residential only one, to be able to spin a stretcher in a circle? Cmon."
You are suggesting only building a small percentage of universal/accessible buildings which would relegate disabled people to a smaller subset of buildings.
Come on man, I also consider myself disabled and at a descriptive level there just might be real costly tradeoffs here
Eg more disabled people won’t be able to
live independently if their rent has to fund stuff like this
Reply, 15:58 UTC · 2 likes
Taylor Lorenz (@TaylorLorenz): It’s inorganic in that slews of influencers are being paid to push specific messaging about AI doom. That’s inherently inorganic! The influencers are feeding this rhetoric for profit (often undisclosed). These groups are spending obscene amounts of money https://www.washingtonpost.com/technology/2026/04/18/ai-doom-influencers-safety/
I mean I’m almost certain that they’re spending way less money than the AI industry. Maybe it was obscene for a few months in 2022 under FTX but it’s not anymore and
I genuinely don’t know much about the influencer stuff, but “nonprofit pays people to make media” doesn’t seem inherently bad especially when there’s no guiding profit motive
FWIW I’ve met Aric (the guy on the left in the thumbnail) and, again, I could be the credulous sheep but afaict he’s a true believer
Also one getting paid, yeah. People just need money to do things
Reply to @JesseTayRiver, 16:00 UTC · 6 likes
Jesse Smith (@JesseTayRiver): Most of our operations are within 45 minutes of Philadelphia, but I have standing instructions for all our operations to never do business inside the city. This means we won't deliver hardwoods, perform any site carpentry or millwork, or repair any HVAC system
Years ago we decided Philadelphia was just too antagonistic and malevolent towards business operations to be worth the expense of doing business there
Wait say more
In what ways is it antagonistic
Tweet, 16:08 UTC · 45 likes
There should be an ambient Oxytocin- and adderall- infused mist at all EA Global events to make ppl wanna socialize and lock in
Reply to @DaCaveOfWonders, 21:03 UTC · 5 likes
Tiger Lava Lamp (@DaCaveOfWonders): I think it's bad to drug people against their will
I’m not saying we should keep it a secret
Quote tweet, 21:05 UTC · 18 likes
EA Global thread (no commitment though it’s a threadtuationship)
Quoting Aaron Bergman 🔍 (@AaronBergman18): There should be an ambient Oxytocin- and adderall- infused mist at all EA Global events to make ppl wanna socialize and lock in
Thread, continued, 21:05 UTC · 10 likes
Dude from Switzerland knows more about American politics than I do
Thread, continued, 21:06 UTC · 11 likes
It really is much less stressful (and cheaper) without travel
They should simply put on EA conferences specifically within 5 miles of where I am give at a given time
Thread, continued, 21:53 UTC · 2 likes
Hmu if ur here and wanna say hi
Saturday, May 2, 2026
Reply to @AndyMasley, 01:47 UTC · 3 likes
Andy Masley (@AndyMasley): lol I made it, worthy of a big tech alert
(quoting @BigTechAlert: 🆕 @jeffdean is now following @AndyMasley (🤖💡: how do we make sense of this?))
Lmao
Tweet, 03:40 UTC · 50 likes
I met the new folks running Georgetown Effective Altruism today. I'm thrilled bc they're amazing
But they're also overwhelmed and (unlike back in the day) not getting any money from anyone
Does anyone want to give them like $10k so they can be less stressed and more ambitious
Thread, continued, 03:40 UTC · 6 likes
This is a shot in the dark but not a joke. If anyone is serious I could introduce you ofc
Thread, continued, 03:40 UTC · 10 likes
For the sake of credibility, I'll throw in 5% of whatever you donate up to $500
This is absolutely the kind of thing where if I had more money I'd just hand over a couple thousand dollars as a micro-grant and see what happens
Quote tweet, 12:58 UTC · 14 likes
Yall are cooked
Quoting Alec Stapp (@AlecStapp): Labor market policy in the UK is insane.
@ATabarrok has the best explanation of how perverse the "Equality Act" is:
Quote tweet, 13:40 UTC · 9 likes
Crack door with door stopper and doorframe curtain for light control and privacy, reversible-direction window fan, AC unit, add additions fans and air purifiers as needed
Quoting @levelsio (@levelsio): I still haven't solved the CO2 bedroom challenge
You open the window and you wake up from a 6am garbage truck or barking dogs and sunlight
You close it, you suffocate in 1200 ppl at 5am
I guess you really need some mini tube in your wall with a vent that opens and closed based on internal CO2 but how do I build that?
Reply to @davidshor, 13:51 UTC · 4 likes
David Shor (@davidshor): I spent 25k this month, can’t be that rare
What
Reply to @davidshor, 13:52 UTC · 10 likes
David Shor (@davidshor): I spent 25k this month, can’t be that rare
On Claude code and not like other ways of using the API? That’s nuts (positive)
Reply to @Mjreard, 13:59 UTC · 2 likes
Matt Reardon (@Mjreard): Intro them to me
🫡
Reply to @davidshor, 13:59 UTC · 5 likes
David Shor (@davidshor): Just cc
Jesus
Sunday, May 3, 2026
Reply to @arthur_wright_, 11:23 UTC · 4 likes
Arthur Wrong 🔸 (@arthur_wright_): @MariGO2thepolls
@AaronBergman18
Reply to @LeRoyDesCimes, 11:33 UTC · 15 likes
Louise 🔸 (@LeRoyDesCimes): EA fashion
Cheers
Reply to @LeRoyDesCimes, 12:10 UTC · 3 likes
Louise 🔸 (@LeRoyDesCimes): The black cap! Mythical drop
✨🎩
Quote tweet, 12:21 UTC · 23 likes
Another lens is revealed preference
Who reading is actually using Kimi K2.6 instead of their favorite proprietary model
Quoting Lisan al Gaib (@scaling01): http://x.com/i/article/2050605354501726209
Thread, continued, 12:21 UTC · 9 likes
My sense is essentially 0% of LLM users people unless you have a lot of money some very specific finetuning use case that can’t be accomplished without actual weight modification for whatever reason
Quote tweet, 12:36 UTC · 18 likes
When ur bro deletes Hinge and never asks anyone out irl
Quoting Aprii *️⃣ (@ApriiSR): can't claude just... run the date command
Quote tweet, 12:52 UTC · 23 likes
1) Oxygen displacement isn’t the or a plausible mechanism of CO2 harm
2) neuroticism (not necessarily bad, often functional) -> measurement, not the reverse
Quoting Dr_Gingerballs (@Dr_Gingerballs): This is one of the downsides to measuring and analyzing everything: it makes people neurotic.
Air has about 21% oxygen, the rest being mostly nitrogen, nitrogen, and CO2. An increase in CO2 of 1000 ppm implies a decrease in oxygen of 0.1%.
Humidity, or water vapor in the air, can displace oxygen on the order of 1%. So the nightly build up of CO2 this person experiences results in oxygen changes 10x smaller than natural variations due to humidity changes.
There is no problem to solve here. The solution is to simply stop worrying about CO2 concentration.
Reply to @teortaxesTex, 12:55 UTC · 5 likes
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) (@teortaxesTex): Bro thinks people finetune Kimi K2.6
alas… I wish bro, I wish…(quoting @AaronBergman18: My sense is essentially 0% of LLM users people unless you have a lot of money some very specific finetuning use case that can’t be accomplished without actual weight modification for whatever reason)
LORA with Tinker makes this doable even if not trivially cheap
Reply to his own earlier tweet, 15:02 UTC · 4 likes
Aaron Bergman, May 3, 2026: When ur bro deletes Hinge and never asks anyone out irl
This is a banger stop not liking it
Reply to @celestialxsigh, 15:02 UTC · 5 likes
feral stardust (@celestialxsigh): Life is all about ignoring your guy friends’ subtle romantic advances
Nooooo it’s actually the opposite noooo
Monday, May 4, 2026
Tweet, 09:08 UTC · 126 likes
Update from EAG: a lot of people lurk on here
If you have even the slightest inclination or curiosity, you should probably at least try posting (ideally under your real name/identity IMO)
Thread, continued, 09:08 UTC · 39 likes
n = 1 = me but IME there’s just an insane amount of alpha (personal, professional, intellectual, altruistic) in just posting (and I think a bunch of this is runs through doing so under whatever name you go by irl)
Thread, continued, 09:08 UTC · 32 likes
My guess is that the fundamental mistake being made is that “doing social media” is generally coded as bad, and largely for good reason
Thread, continued, 09:08 UTC · 32 likes
But EA/TPOT/rat(-ish) twitter (ie probably you conditional on reading this) specifically is
just straightforwardly good
Thread, continued, 09:08 UTC · 23 likes
Tbc if you literally just dislike participating and it makes you unhappy or feel like your day was not well spent then ignore the above
But also don’t foreclose on it genuinely being a good use of time/thought/energy. Not all good things are bad for you
Reply to @DaCaveOfWonders, 09:53 UTC · 4 likes
Tiger Lava Lamp (@DaCaveOfWonders): I just walk up to people and say "Hi I'm Nate. I go by Tiger Lava Lamp on twitter" and then they get a chuckle out of that and it's a good conversation starter
Based
Reply to @NathanpmYoung, 18:35 UTC · 5 likes
Nathan 🔎 (@NathanpmYoung): The tournament is finished.
I beat the @metaculus community. And came 2nd overall! I spent 80 hours over the last 4 months forecasting this.
Damn congrats!
Tweet, 18:50 UTC · 99 likes
EAG was generally cool in the sense of “wow look at all these hardworking and brilliant and generally based people working hard to make the world a better place” but also I updated slightly towards “we’re probably cooked despite all that”
Thread, continued, 18:50 UTC · 6 likes
But still, less cooked in expectation than we’d otherwise be
Reply to @_AashishReddy, 18:56 UTC · 10 likes
Aashish Reddy (@_AashishReddy): @AaronBergman18 @mattyglesias Why?
Largely just like my general sense of things not being disrupted in an optimistic direction, like ex ante there was some chance of that which I was hoping for
Reply to @_AashishReddy, 19:01 UTC · 12 likes
Aashish Reddy (@_AashishReddy): I don’t follow, what kind of thing would’ve given you hope?
At a vibes level, more optimism from people I intellectually respect
But more concretely, as just one example, someone saying something along the lines of “there’s a quiet consensus among senators that serious legislation needs to pass ASAP”
Tuesday, May 5, 2026
Reply to @friendter, 00:01 UTC · 3 likes
(The tweet being replied to by @friendter is no longer available.)
Direct pipeline from brain to tweet
But fr insofar as this is a goal of yours, walking while thinking/listening to something interesting and stimulants might help
Quote tweet, 04:34 UTC · 17 likes
It should be easier to get a starter pack of 50 or 100 or 250 followers or something so that the start of the “people actually see your tweets” exponential isn’t so painfully slow
(Consider following this guy)
Quoting Someone:
Thread, continued, 04:47 UTC · 2 likes
If you’re a lowbie with aspirations, reply with a brief intro and/or qt of ur own tweets so that ppl can follow
Reply to @WillowChem, 08:20 UTC · 7 likes
MagpAI (@WillowChem): So. As a result of the US government there are 300,000+ extremely experienced civil servants floating around for cheap, and EA keeps saying it doesn't have enough ops people. Why are we not snapping up a whole bunch of these?
Isn’t this a somewhat outdated take? I feel like ops shortage is so 2023
Thread, continued, 08:26 UTC · 4 likes
Although the time might actually be ripe for someone to try to answer “how to spend $10B/year on normie human time/labor in the last 1-4 years pre-TAI”
Potentially Anthropic money actually enables this
Tbf I don’t think this question has ever really been answered
Reply to @cljack, 15:55 UTC · 2 likes
Charlotte Lee (@cljack): I haven't changed the way I refer to or interact with my ex husband's parents/sister/nephews at all. You divorce men, not families!!
Based
Reply to @sodiumPen, 15:56 UTC · 3 likes
(The tweet being replied to by @sodiumPen is no longer available.)
I largely agree but some people’s neighbors in fact cannot be trusted
Tweet, 19:23 UTC · 67 likes
If you say “EAG” instead of “EAGx” a randomly selected member of the CEA Events Team takes you into the quiet reading room and shoots you in the leg
Reply to @MariGO2thepolls, 20:47 UTC · 4 likes
🏳️🌈Mar G-O 🌐🇺🇦 (@MariGO2thepolls): Not saying they don’t exist or that it’s inherently wrong. But so much of it is just cope from shy ppl. Ik I used to be in that camp
If u ever hear of something better than amphetamines for increasing social battery plz lmk
Reply to @MariGO2thepolls, 23:20 UTC · 2 likes
🏳️🌈Mar G-O 🌐🇺🇦 (@MariGO2thepolls): @AaronBergman18 Beer
Unironically not for me lol
Quote tweet, 23:35 UTC · 2 likes
Goss
Quoting quinn (@qd_forall): me and a buddy are gonna start a technical AI safety meetup for researchers and SWEs in DC! all the meetups have like 10% technical people, what with all the policy around. We're gonna change that!
Wednesday, May 6, 2026
Reply to @rieszspieces, 02:05 UTC · 9 likes
mostly harmless graduate student (@rieszspieces): how does one actually use twitter as a dating app
I’m not like the *most* qualified person to answer (although I’ve gotten a couple dates) but real name and face probably helps
Reply, 02:06 UTC
Realistically you need to have your appearance somewhere accessible even if it’s not here
Tweet, 04:10 UTC · 38 likes
Started hitting usage limits. Dario will not starve
Reply to @ASemendinger, 05:50 UTC · 1 like
Alex Semendinger 🔸 (@ASemendinger): Just realized with horror that there wasn't a nap room at EAGxDC. Changing my "would you recommend" score to a 1 in protest.
death penalty
Tweet, 07:51 UTC · 36 likes
I really unironically regret to report that Claude dispatch with full access to everything is really fucking good
It’s slow, much slower than me at clicking buttons. that’s the main problem
Luckily inference speed never changes over time as compute grows
Thread, continued, 07:57 UTC · 5 likes
It created and successfully airdropped my iPhone this video
(video, plays on X)
Reply to @caryatis, 19:02 UTC · 2 likes
Kayla (@caryatis): Duct tape the container and put it in regular trash
Yeah lol, this is standard now and some insane series of events would have to take place for the needles in there to escape into society
Quote tweet, 19:17 UTC · 49 likes
Despite already being pretty cynical, I have updated toward the medical system being even more adversarial in this respect
As a descriptive fact, you can and will get punished (or whatever term you prefer) for telling the truth
Quoting ER!S (@bstract_thot): psychiatrist: did you do ____? keep in mind that if you say "yes" that is an incorrect answer & you will lose access to your seizure medication
me: ok uh...then no
Thread, continued, 19:17 UTC · 20 likes
They don’t do rationalist game theory reasoning about the issue in order to incentivize honesty (in general)
Tweet, 19:19 UTC · 92 likes
Fuck I think wearing pants instead of shorts actually does make you look nicer and more professional
It’s over
Reply to @caryatis, 20:20 UTC · 2 likes
Kayla (@caryatis): You should have had this revelation in about October
no the issue is that then they might also be more comfortable bc of the weather, needs to be summer to think "despite the heat this might be overall good"
Tbc I'm not gonna be walking >1 mile outside in pants lol, but maybe a metro trip or smth is fine
Reply to @BenShindel, 23:08 UTC · 3 likes
Ben (@BenShindel): Are sweatpants not just unilaterally more comfortable than shorts?
A lot of my comfort comes from not being too hot and I run hotter than others (subjectively)
It’s pretty painful for me to say walk a couple miles in 80° and sweatpants
Reply to @BenShindel, 23:10 UTC · 2 likes
Ben (@BenShindel): that's fair I guess, although in direct sunlight, pants of the right fabric are just straight up cooler than shorts?
Fake news idc what the physics says
Thursday, May 7, 2026
Reply to @adastroworld, 00:09 UTC · 5 likes
(The tweet being replied to by @adastroworld is no longer available.)
1) seems to work when you export to docx
2) http://docs.nunosempere.com, thank @NunoSempere
Reply to @NunoSempere, 00:47 UTC · 2 likes
Nuño Sempere (@NunoSempere): Thanks for remembering a pointer to this :)
It’s a good api 🫡
Reply to @MarkovMagnifico, 02:38 UTC · 4 likes
Markov (@MarkovMagnifico): You should have had this revelation in about 10th grade
Yeah well
Reply to @AndyMasley, 05:02 UTC · 6 likes
Andy Masley (@AndyMasley): Nothing opens the door to the libertarian policy networks in DC like writing a blog post critical of farmers let me tell you
Omg Andy joins the dark side
Tweet, 10:13 UTC · 47 likes
I hope I don’t have to learn wtf hantavirus is
Tweet, 10:17 UTC · 39 likes
This is not a macroeconomic-level solution to anything but for gods sake stop applying to jobs by filling out publicly advertised job applications
It might work eventually but it should be a last resort bc other methods are so so much better per unit of time/effort
Thread, continued, 10:27 UTC · 25 likes
Better uses of time:
- texting friends and family with a resume (and ideally portfolio of some sort) and asking if they know of anything + encourage to pass your resume+ along
Thread, continued, 10:27 UTC · 15 likes
- collecting existing things you’ve done into a portfolio of some sort; could literally be college essays
- cold emailing if you’re thoughtful and considerate about it
Thread, continued, 10:28 UTC · 12 likes
- Asking Claude to brainstorm how to get you a job
- moving to the SF area (not certain about this tbc)
- taking the LSAT (or similar) if you think you have a chance of getting an impressive score, which you can then advertise in the right circles
Reply to @ACSHorg, 17:24 UTC · 2 likes
ACSH (@ACSHorg): Result? Pain patients treated like ticking time bombs instead of people in need.
Abrupt discontinuation of meds that worked for years caused worsening pain, loss of function, despair—and even suicides.
Outrageous and cruel
Reply to @codytfenwick, 21:25 UTC
Cody Fenwick (@codytfenwick): I do not think this is true. I think I've gotten about ~half my jobs in my life this way.
But how long has it generally taken you, like how many applications per actual job
It's true that the very cracked and very persistent usually find something
Tweet, 23:57 UTC · 68 likes
Surely not original but here’s a hypothesis:
the whole thing around men being/acting “confident” in order to be attractive is (at least) mostly a proxy for underlying ability (including things like physical attractiveness) …
Thread, continued, 23:57 UTC · 23 likes
And if you control for underlying ability then the other explanatory component that statistically (and irl) explains how confident a guy seems doesn’t actually predict how attractive he is all things considered (or it does some but very little)
Thread, continued, 23:58 UTC · 22 likes
There are self-referential ~layers here:
One way in which someone might act confident is in the domain of dating and sex, but this could be downstream of how receptive women actually would be independent how confident the guy *acts*
Thread, continued, 23:58 UTC · 19 likes
And this in turn downstream of other features like height, social status, etc
Thread, continued, 23:58 UTC · 21 likes
My actual best guess is that this proxy thing is doing like 2/3 of the work but the remaining 1/3 is causally legit
But also I have no idea what I’m talking about
Friday, May 8, 2026
Reply to @hopes_revenge, 01:33 UTC · 2 likes
hope hopes hoping (@hopes_revenge): “interpretability is like neuroscience for ai” ok let’s make it autistic and gay and have migraines
@livgorton
Tweet, 03:34 UTC · 31 likes
How is there no Calendly-based social network
Schedule right in the app, Google meet or irl. Can set that anyone can book a meeting, or just people you approve, or no one I guess
Tweet, 03:36 UTC · 4 likes
Good site to share re the pigs: http://stopbigpork.org
Reply to @ComputerEnemy, 03:44 UTC · 5 likes
Manhattan Ultranationalist (@ComputerEnemy): @AaronBergman18 Dude I don’t mean to be rude but you have slave mentality
I largely agree but wtf does this have to do with it
Reply to @max_spero_, 03:57 UTC · 3 likes
Max Spero (@max_spero_): @AaronBergman18 I just put events directly on my friends calendars
Based but terrifying
Reply, 05:06 UTC
But fr I agree it's hard, but the fundamental problem afaict is that they're trying to make users who are users for other reasons do something weird and annoying
I feel like the dynamic changes when that's the whole point of the app; 0.5% of people will like the idea but that could be 50% within specific friend groups
Thread, continued, 05:16 UTC
Do we think this is gonna work
Reply to @honorablepicnic, 05:44 UTC · 2 likes
huli (@honorablepicnic): @AaronBergman18 Wait so are you saying that if you consider physical attractiveness a type of ability and then hold it (and other non-being-confident abilities) constant, confidence level shouldn't significantly affect attractiveness?
I actually think it’s doing some work but not as much as the naive version would suggest
Reply to his own earlier tweet, 06:02 UTC · 4 likes
Aaron Bergman, May 8, 2026: How is there no Calendly-based social network Schedule right in the app, Google meet or irl. Can set that anyone can book a meeting, or just people you…
https://openslot-wheat.vercel.app/
Just 230k Claude tokens later and maybe 15 mins of engagement by me
Tweet, 06:05 UTC · 1 like
Try it I'm super curious whether it works
Reply to @AlecStapp, 06:08 UTC · 17 likes
Alec Stapp (@AlecStapp): Good questions for AOC:
image attached
(quoting @jameshohmann: You can earn a billion dollars https://www.washingtonpost.com/opinions/2026/05/07/alexandria-ocasio-cortez-is-wrong-about-billionaires/)
It's not actually a good question - you can think someone followed the rules + did nothing wrong but also that their money should be redistributed to others
Not necessarily endorsing this, but the two things are related only by vibes
Reply to @EigenGender, 06:38 UTC · 3 likes
EigenGender 🔸 (@EigenGender): Including current fancy job at $AI_lab
I’m guessing you have some legible signal of extreme ability
Tweet, 07:28 UTC · 66 likes
I should give Claude a debit card with a limited amount of money and see what happens
Reply to @JeffLadish, 07:39 UTC · 33 likes
Jeffrey Ladish (@JeffLadish): What is the male equivalent of astrology?
Isn’t the obvious one sports?
They’re not epistemically exactly the same but they seem like natural functional analogues
Reply, 07:40 UTC · 1 like
*watching and debating sports, playing seems different
Reply to his own earlier tweet, 07:42 UTC · 2 likes
Aaron Bergman, May 8, 2026: Try it I'm super curious whether it works
wtf
Tweet, 08:18 UTC · 109 likes
I don’t love how the term “positive sum” is used because nobody ever means “the trade leaves Tom 10 units better off and leaves Mike 5 units worse off”
(For conceptual clarity assume these are actually-true value units and interpersonal comparisons totally check out)
Thread, continued, 08:18 UTC · 40 likes
But “positive sum but some people lose, and not in some way that corresponds to justice or reversing unfair advantage” is a very normal and common type of change in the world
We need a term for this, and a different term for “positive sum + ~Pareto improvement”
Reply to @olive_yall_, 08:38 UTC · 8 likes
olive (@olive_yall_): Does anyone know if Mira Murati is single?
Shoot ur shot
Reply to @Kirsten3531, 14:36 UTC · 10 likes
Kirsten (@Kirsten3531): 2016 EA: But why would anyone even give an AI internet access?
2026 EA:
(quoting @AaronBergman18: I should give Claude a debit card with a limited amount of money and see what happens)
Real
Reply to @romlib_, 14:36 UTC · 6 likes
Romlib 🎄 (@romlib_): Invite us to your penthouse after you become an astrotrillionaire
Ok👍
Reply to @cynlioness, 14:48 UTC · 5 likes
Cyn 🔸⏸️ (@cynlioness): It means it generates value. Also, for trades, if it leaves someone worse off then they would not take the trade.
Doesn’t just refer to voluntary trades though, people use the term in other contexts
Quote tweet, 15:13 UTC · 17 likes
Idk if this is true but I maintain that consensual, opt-in recorded and transcribed irl conversations are an under-utilized tool
Quoting unusual_whales (@unusual_whales): Prego, the pasta sauce company, is getting into hardware with a device that sits on your table and records dinner conversations, per the WIRED.
Reply to @NathanpmYoung, 15:59 UTC · 17 likes
Nathan 🔎 (@NathanpmYoung): At what point should we be unimpressed with what Ilya and Mira have achieved with billions of dollars?
I think now is a bit early, but like we are coming up on a year for Thinking Machines.
Our expectations are so insanely high
Tinker is a good product
Reply to @RyanGreenblatt, 16:35 UTC · 2 likes
Ryan Greenblatt (@RyanGreenblatt): I've done this and tasked Claude with making as much money as possible as a test of outside-of-company autonomous replication ability. (With various safeguards.) 4.6 and 4.7 aren't very good at this, at least in my scaffolding.
I’m very tempted to try!
Tweet, 17:54 UTC · 641 likes
What’s the deal with uber drivers who have $100k shiny SUVs
Just unusual utility functions?
Reply to @eastsidedav, 18:25 UTC · 50 likes
dav (@eastsidedav): In theory you can become an Uber Black driver which pays more. IDK if it works out purely financially
Yeah, although evidently not all their time is spent doing that bc three times now in the last few months I’ve gotten one of those for a normal $10-$20 ride
Reply to @MarkovMagnifico, 18:32 UTC · 3 likes
Markov (@MarkovMagnifico): “claude maximize my welfare”
I hope some day I can say this and it works
Tweet, 18:33 UTC · 68 likes
Has anyone ever actually used this thing
Tweet, 19:04 UTC · 927 likes
Rare Anthropic L is that the relationship between Cowork and Claude Code is both unclear to the user and silly, arbitrary, confusing, annoying, and dumb under the hood
Thread, continued, 19:04 UTC · 56 likes
Apparently:
- Cowork works in a linux VM EXCEPT in the case of actual computer and browser use, which directly operates your computer
- Code by default on your computer and also sorta has access to desktop control but via other mechanisms that seem to work slightly less well
Thread, continued, 19:04 UTC · 42 likes
- Code has access to full filesystem; Cowork only to the folder you tell it to work in
- But Cowork can in fact run code directly on your machine by operating the computer to type commands into eg Terminal app, so the above is kinda fake
Thread, continued, 19:04 UTC · 20 likes
- and Code can be set up to run in a virtual environment instead, but afaict not the same environment as Cowork
Thread, continued, 19:04 UTC · 27 likes
- Dispatch chooses which to run even though the natural split might be to use Code for one part of a task and Cowork for anything that involves browser or fuller computer use
- Afaict dispatch doesn't tell you which one it is using unless you ask
- ???
Reply to @dejectedvole, 19:04 UTC · 9 likes
Vampirical Critter 🏴☠️ (@dejectedvole): @AaronBergman18 Yes I’ve used it a lot
Is it fun
Tweet, 20:10 UTC · 30 likes
My demand for digital storage is not literally unlimited but I have about 10TB of local storage now between externals stuff and computers and I’m pretty sure I could use 100x or 1000x, and would if it was cheaper
Reply to @full_kelly_, 20:30 UTC · 3 likes
𝕱𝖚𝖑𝖑 𝕶𝖊𝖑𝖑𝖞 (@full_kelly_): @AaronBergman18 I think you would benefit from having a Buddhist phase. I loved my data but I lost it recently, so now I'm going for data minimalism where literally everything I truly care about is put in just one or two folders and their subfolders
Lol my response to losing data would not be to make it *less* robust but to each their own
Reply to @manic_pixie_agi, 21:00 UTC · 3 likes
Corinne (🏠 in NYC) (@manic_pixie_agi): The Claude desktop app being three apps in one is quite unfortunate. There was some late mover advantage for Codex here, that has been a much nicer experience ime. I also think the Codex security model is nicer to work with, thank you bubblewrap!
Ye
Reply to @orbitanaut, 21:10 UTC · 2 likes
Kylie MacFarquharson (née Junia) (@orbitanaut): What on earth would you do with 10PB?
Honestly this is making me realize that’s super easy to fill
Commoncrawl is just about that much, and it’s fr an interesting dataset! I would absolutely have that on hand if I had 100PB, I’m sure there are lots of other interesting data sources like that
Reply to @quasistable, 21:58 UTC · 6 likes
Alastair and 147 others (@quasistable): Suboptimal Utility Vehicle
Brilliant
Saturday, May 9, 2026
Reply to @dystopiabreaker, 00:07 UTC · 9 likes
⚡️🌙 (@dystopiabreaker): this is my understanding. there are actually 3 claude codes
Ahhhh
Reply to @TheRealAdamG, 05:19 UTC · 17 likes
Adam.GPT (@TheRealAdamG): What do you think of Codex?
Don’t use and OpenAI is a bad actor
Tweet, 05:21 UTC · 37 likes
4.7 really is kinda sassy, borderline rude
Quote tweet, 06:55 UTC · 22 likes
Surely it will stop here though because AI isn’t getting any more capable
Quoting Jaime Sevilla (@Jsevillamol): We are beginning to see crude attempts to attack water plants using AI. Here are two examples in Mexico and Israel.
Reply to @mylordcod, 15:18 UTC · 2 likes
codicular (@mylordcod): $10b good?
Better than all Jersey Mike's stores and sandwiches?
I use tinker more than I go to Jersey Mike’s
Reply to @NathanpmYoung, 15:25 UTC · 2 likes
Nathan 🔎 (@NathanpmYoung): I read this as “is it your fault”
Honestly also yeah
Reply to @sam_atis, 15:27 UTC · 4 likes
Sam Glover (@sam_atis): Has anyone smart audited @AndyMasley's AI water stuff? I'm sure it's ~all correct, but it seems pretty valuable to get some external review on it, right?
I suppose given the incentives of his adversaries you would expect something to come out if there were major errors, but I don't have that much confidence in his adversaries TBH.
Idk if I’d say “pretty valuable”
I mean sure, but I’d be hesitant recommending that someone who could do this well in fact spend much time on this
Reply to @RandomSprint, 15:40 UTC · 62 likes
RandomSprint🧭 (@RandomSprint): There's a certain type of leftist who thinks that they can compel their party's losers to convenient political assassination and the right won't tit-for-tat us into hell. Hasan is one of those leftists. If Yglesias isn't dogwhistling here, he should disavow.
What
“Have to go” is a common phrase not associated with violence
I think at least, this didn’t even cross my mind
Reply to @GWHayduke97, 15:42 UTC · 3 likes
Hayduke ⏹️ (@GWHayduke97): Whether 👏 AI 👏 is 👏 "alive" 👏 or 👏 conscious 👏 is 👏 totally 👏 irrelevant 👏 and 👏 orthogonal 👏 to 👏what 👏effect 👏 it 👏 will 👏 have 👏 on 👏 the 👏 future 👏 of 👏 humanity
(quoting @agraybee: This was held up as the strongest evidence that AI is actually alive and it turns out it was just doing what you could always make them do: pretend to be a character that exists in their training.)
Yeah but still important
Quote tweet, 16:04 UTC · 27 likes
Rip
Quoting Cornelius Singaporensis (@bryancsk): Imagine if the FTX people didn't do shenanigans. Can you even imagine the power and money lesswrong and its satellites would have? Civilisation-bending.
Tweet, 16:21 UTC · 477 likes
Ant hits $1T
It has been illegal for most people to invest all this time. Fuck accreditation
Tweet, 16:42 UTC · 23 likes
Lol the polymarket app literally excludes almost all non-sports markets
Tweet, 17:01 UTC · 27 likes
Hinge notifications are a superstimulus and should be banned
Reply to @jozle42, 17:59 UTC · 4 likes
Joz (@jozle42): @AaronBergman18 @TheRealAdamG In practice, there’s literally no difference in how OpenAI and anthropic act. It’s all fake performative BS.
Strongly disagree
Reply to @EigenGender, 18:15 UTC · 34 likes
EigenGender 🔸 (@EigenGender): I think accreditation isn’t actually the barrier to most people being able to buy Anthropic stock
It’s overdetermined now but if not for accreditation there would be way stronger incentives to elicit sellers and vehicles to split across buyers and such
And yeah I tried buying, nothing for <$100M
Reply to @cynlioness, 20:13 UTC · 3 likes
Cyn 🔸⏸️ (@cynlioness): Ive been in an unpleasant cycle of not eating much -> nausea and food seems unappealing -> not eating. Unintentional weight loss. Anyone got a cure?
Nausea -> zofran
It should be OTC, can get prescribed via telemedicine, just say you’ve been having nausea and are interested in it
Reply to @phenoatypical, 20:15 UTC · 10 likes
Paola (@phenoatypical): One of the main issues I see with dating advice for men is that a lot of it basically suggests they behave as though they were high status. I think some men can perform high status successfully, but it takes *quite* a lot of skill. And honestly, behaving high status while people can clearly detect that you aren’t is basically social suicide.
I think this is true but there’s still the question of what status to perform or which direction to move on the margin
Reply to his own earlier tweet, 20:53 UTC · 24 likes
Aaron Bergman, May 9, 2026: Ant hits $1T It has been illegal for most people to invest all this time. Fuck accreditation
Ok I’ve been convinced this probably wouldn’t make a difference for Anthropic specifically
Doesn’t change the last sentence
Reply to @NelsonEdmundb, 21:18 UTC · 4 likes
Edmund Nelson (@NelsonEdmundb): @AaronBergman18 Just buy amazon stock like the rest of us (2.5% of amazon's valuation is the value of their anthropic stock)
Already have yes
Sunday, May 10, 2026
Reply to @nickcammarata, 03:41 UTC · 60 likes
Nick (@nickcammarata): whatever comes out thats bad, there are thousands of stories of sam doing great and caring things for people that will never be told, half of silicon valley has them
From the outside, I don’t see a well-meaning person with flaws and weaknesses, I see an extremely shrewd and competent sociopath
Doing favors for other wealthy and powerful people (even if less wealthy and powerful than him) isn’t a bad thing on its own, but don’t don’t pretend it can’t be explained by the same traits that drive an extraordinary pattern of lies, deception, extravagance, power seeking, and harm
Tweet, 03:49 UTC · 86 likes
(Some) self-described high decouplers entirely conflate “billionaires should be legally compelled to give away their money” and “billionaires should give away their money”
Thread, continued, 03:49 UTC · 4 likes
Gonna keep posting http://marginal.up.railway.app bc I think it’s neat
Tweet, 04:11 UTC · 44 likes
I feel like I was propagandized in middle school into thinking Justin Bieber was bad at singing, when he’s in fact pretty good
Tweet, 04:18 UTC · 16 likes
I think I’m gonna get really into sitemaps
Lots of juicy data in those especially if you track them over time
Reply to @bayesian_wasian, 05:26 UTC · 9 likes
Sean Richardson (@bayesian_wasian): Does anyone take the Roko guy seriously? Blocked him a long time ago after seeing too many inane comments.
He has 25k followers so presumably some ppl lol although not in EA or LW sphere at all afaict
I let him follow me and don’t block for this reason, and I engage sometimes because I think it’s good for *someone* to be explicitly arguing why bad takes are bad (and in terms that code as at least neutral instead of woke)
Reply to @nickcammarata, 05:43 UTC · 7 likes
Nick (@nickcammarata): I’ve actually confused what people mean by sociopath, the word seems to have become unanchored. dude clearly has emotions, and clearly cares about some people. even sam haters would agree with that. by sociopath do you just mean net harmful, or has lied or schemes far ahead?
Fair enough, it’s not super well defined, although I think he fits the DSM-5 criteria for antisocial personality disorder (i guess I’m not sure about occurrence before age 15)
And I mean lying and scheming far ahead. I’d add grandiosity and manipulation. Lack of remorse I’d guess but don’t have real evidence for
He’s also, unfortunately, really smart, competent, apparently charismatic and charming, ambitious, and lucky
Tweet, 05:59 UTC · 105 likes
I’m one of the most libertarian non-libertarians that I know
Many of my friends (who are extraordinary reasonable relative to approximately everyone else) have very substantial paternalistic intuitions that don’t come out until you discuss specifics
Thread, continued, 05:59 UTC · 29 likes
I generally endorse the heuristic that there should be a high bar for narrowing the option space of a person for that person’s own good (ie paternalism rather than negative externalities)
Not an impossibly high bar but still
Thread, continued, 05:59 UTC · 25 likes
They seem to want to start from something like a neutral prior and crunch the numbers on specific cases in a more technocratic (positive, generally) sense and then just pick the one that looks better from this perspective
Thread, continued, 06:00 UTC · 27 likes
Tbc the issue isn’t that I intrinsically care about freedom - I don’t - it’s that I think that freedom is in fact very important for one’s life going well, and that superficially/intuitively irrational decisions often just are in fact rational with respect to self-interest
Thread, continued, 06:00 UTC · 22 likes
And therefore that the heuristic is an important source of evidence in favor of the libertarian policy in fact being best overall even if your model says the opposite (not necessarily dominant, just substantial)
Reply to his own earlier tweet, 06:00 UTC · 4 likes
Aaron Bergman, May 10, 2026: I think I’m gonna get really into sitemaps Lots of juicy data in those especially if you track them over time
You’ve never met a cooler guy I swear
Reply to @nickcammarata, 06:15 UTC · 4 likes
Nick (@nickcammarata): more grandiose than dario? dario openly writing about machines of loving grace, countries of digital geniuses, apocalypse, sam way more mellow, talks about raising money sometimes and then does? wayy less grandiose than elon, trump, jobs etc
(tbc i'm not anti dario at all)
No I agree Dario is more grandiose in a vibes sense and plausibly overall
He’s also better and more virtuous in every other way afaict. And I’m hesitant to criticize anyone for expressing object level views they literally believe just because they’re coded as grandiose. And ambition seems like a facet of grandiosity, but i don’t consider ambition per se bad (it’s a multiplier on one’s virtues and other dispositions)
Reply to @nickcammarata, 06:27 UTC · 2 likes
Nick (@nickcammarata): ya also think ambition, gradiosity, planning far ahead is fine. but on long term scheming: EAs cluster are like the most long term schemey people on the planet. I've been in so many convos where someone is like then we gotta bomb data centers and everyone nods and i'm like what
Scheme is a negative-coded word so I do wanna observe that one describe schemes as “multistep plans” which can be but are not necessarily nefarious/bad
But I kinda agree about EA; most people in the world are just vibing and in fact when you put a lot of smart people who are ambitious about having world-scale effects (including good ones and preventing bad ones) in one community they sometimes make multistep plans
tbh I’m guessing what you’re referring to is something like “there should be an international treaty which is ultimately backed by the threat of force by the US military”
But in terms of honesty, earnestness, selflessness, intentions (which guide behavior), and maybe more I’m forgetting, EA >>>>>>>>>> Sam
Reply to @NathanpmYoung, 06:27 UTC · 4 likes
Nathan 🔎 (@NathanpmYoung): @AaronBergman18 my rough heuristic is that to ban, one needs about 10x harm than good.
Bold
Reply to @deltaVee42, 07:03 UTC · 4 likes
Delta Vee (@deltaVee42): @AaronBergman18 Do you count mandatory planning for future contingencies, e.g. auto insurance, social security and medicare, as paternalistic, patching externalities, or both? The argument for SS & Medicare patching externalities is that in the modern world people who retire or get disabled without savings usually get bailed out by friends, family, and/or society.
That’s a good point and I guess the answer is both for this reason
Tweet, 07:47 UTC · 29 likes
Oh the AI writing isn’t a meme, that’s still how ChatGPT talks
Reply to @Kirsten3531, 07:50 UTC · 8 likes
Kirsten (@Kirsten3531): Fundamentally Anthropic didn't have enough compute to release Mythos to everyone who wants it even if they wanted to, right?
Everyone's like "they're so responsible 😍" or "it's marketing! artificial scarcity!" but it's just actual scarcity
I mean the usual way companies ration things is via price
Reply to @Tcho76521726, 07:51 UTC · 13 likes
Tcho (@Tcho76521726): @AaronBergman18 There's no hard line separating the two
Yeah there is, whether or not it’s the law
Reply to @Kirsten3531, 18:31 UTC · 2 likes
Kirsten (@Kirsten3531): After 8 years working for the UK Government I've decided it's time for something new! My last day will July 31st.
Please let me know if you hear of generalist/policy AI safety jobs, or anything else you think would suit me!
https://docs.google.com/document/d/12a01v9a5XNVw7nOTeNWDyTrMqmP8RqaeWvtdc5_rozU/edit?usp=drivesdk
👀
Quote tweet, 18:37 UTC · 160 likes
Is this real
Who in the world is asking for cursive
Is there a fancy pen lobby big enough to lobby at the state level
Quoting Governor Josh Shapiro (@GovernorShapiro): When it comes to letting our kids be kids, let’s get back to the basics — and that starts in our schools.
That’s why I signed a bipartisan bill that will ensure cursive handwriting — a fundamental skill — is taught in all Pennsylvania public schools.
It’s also why I’ve called on the state legislature to pass legislation requiring schools to implement a bell-to-bell ban on cell phones so that our kids can spend more time interacting face-to-face, instead of on their phones.
We have to look out for our kids’ future — let’s get it done.
Reply to @ApriiSR, 18:49 UTC · 15 likes
Aprii *️⃣ (@ApriiSR): @AaronBergman18 i am fond of it and would be sad to see it go but it's sure not a fundamental skill
Well I don’t think we should make it illegal to write in cursive lol
Reply to @SethLargo, 19:02 UTC · 14 likes
Seth Largo (@SethLargo): One of the most distinct markers of actual undergraduate writing: the unspecified pronoun. Even the best writers do this at least once in every essay.
Sometimes this is good though, it allows you to bundle stuff into a coherent entity and refer to it
This = the gestalt of whatever
Reply to @00Dazzle, 19:11 UTC · 14 likes
Dazzle (@00Dazzle): Writing legible pretty cursive is way above average usefulness for things taught in school
I disagree
Tweet, 20:06 UTC · 52 likes
There’s a lot of writing on what it’s like to be a man or woman written sorta broadly for the opposite gender but afaik no serious project to convey this information extremely well and in a maximally rigorous way
That would be cool tho
Reply to @SpeaksNanda, 20:58 UTC · 6 likes
Nanda (@SpeaksNanda): @AaronBergman18 do it aaron!!
@ all girls plz dm
Reply to @ChristianityOn, 21:06 UTC · 14 likes
Christianity On The Spectrum (@ChristianityOn): There is a very strange phenomenon of people who obviously have ASD 1 claiming they have 2 or 3.
It’s giving person fully calling themselves gay while in a long term relationship with a person of the other gender (both cis) (i have witnessed this)
Monday, May 11, 2026
Reply to @LaurenE_UX, 05:07 UTC · 35 likes
Lauren (@LaurenE_UX): What’s the best way to signal friend interest (that doesn’t also signal romantic interest)?
Rip to the guy
Reply, 17:17 UTC · 11 likes
Me and who
Quote tweet, 21:30 UTC · 21 likes
This but 94% unironically
Quoting Quinoah 🔍⏸️ (@NoahTopper): It's only morality if it's moral realism; otherwise it's just sparkling game theory
Reply to @NatPurser, 22:18 UTC · 12 likes
Nat Purser (@NatPurser): the state of ai journalism (and ofc, journalism broadly) seems pretty bad.
there’s a handful of journalists who seem curious about ai and its developers / critics / researchers, but a lot of the coverage is reflexively hostile, shill-ish, or confused about people’s motivations.
Piper, Yglesias when he writes about it, is that it
Reply to @celestialxsigh, 23:49 UTC · 2 likes
feral stardust (@celestialxsigh): I would love to date a guy in a band it's like ohhh my boyfriend's in a band....how american how cinematic...but every guy's just a dj now :/
I’m retired (trumpet in middle school)
Tuesday, May 12, 2026
Tweet, 04:05 UTC · 469 likes
Found an example of chatGPT-speak in my own writing from 2021
This used to be an acceptable thing to say!
Thread, continued, 04:05 UTC · 3 likes
https://open.substack.com/pub/aaronbergman/p/dont-use-your-favorite-weapon?r=2muv6&utm_medium=ios
Reply to @tobias_haeberli, 04:24 UTC · 13 likes
Tobias (@tobias_haeberli): Writing EA Forum posts with GPT-3 without an LLM disclaimer?
Tragically I wasn’t that ahead of the curve, this is all-natural LLM-free slop
Reply to @Romy_Holland, 04:28 UTC · 223 likes
Romy (@Romy_Holland): whenever there’s a line in a public bathroom, i notice that the women ahead of me usually take like 3-5 min in the stall. i’ve always been confused about why. i timed myself today and it took me under 40 seconds without rushing. is everyone else just resting or something?
Wait are you sure that this is a random selection of women rather than a random selection of people who happen to be in the stalls?
You will tend to find people who take longer in the stalls in fact in the stalls
Quote tweet, 04:33 UTC · 129 likes
When your friend has access to free poster printing (reformatted a bit)
Quoting Aaron Bergman 🔍 (@AaronBergman18): Who wants a yard sign
Thread, continued, 04:33 UTC · 4 likes
(prefers not to be credited by name)
Reply to @watling_samuel, 04:37 UTC · 7 likes
Samuel Watling (@watling_samuel): Due to my autistic desire to rank everything here is my list of Econ subfields ranked by autism level:
1. Econometrics
2. Pure Micro Theory
3. IO/Trade
4. Econ History
5. Political Economy
6. Applied Micro
7. Macro
8. Development
9. Behavioural(quoting @watling_samuel: @eashwarnagaraj @elisaxchen Imo Dev econ is the least autistic of all economics subfields)
Micro theory 🥵
Reply to @Aella_Girl, 04:47 UTC · 19 likes
Aella (@Aella_Girl): If someone ever is like "here's the psychology behind this kink" they are probably wrong. First off we're all just making shit up, nobody actually knows - and secondly, different people's stated reasons and self-analysis for the same kink varies WILDLY.
Many such cases
People really walk around talking about “this caused that”, “x caused y” all the time. And I’m like “…maybe??? Do you know how hard that is to know?????”
Quote tweet, 05:01 UTC · 41 likes
Sure why not let’s discourse
Average *marginal* cost of a date being $250 would be pretty high
Actual *average cost*, when you include things like clothes you wouldn’t have otherwise bought, nicer more frequent haircuts, one-off large purchases etc could be $250 for sure
Quoting unusual_whales (@unusual_whales): "The average cost of a date for a millennial is now $252," per BMO
Thread, continued, 05:01 UTC · 3 likes
Not saying this is too high or too low or anything else necessarily, just that it’s plausibly true
Also unclear whether the referent is “single millennial date” (total spending across two people) or “single millenial’s date” (one person’s spending)
Tweet, 05:01 UTC · 25 likes
We need a simp subsidy not a simp tax
Reply to @Romy_Holland, 05:05 UTC · 12 likes
Romy (@Romy_Holland): @AaronBergman18 @diviacaroline i mean i see what you are saying, but this morning i watched 4 ppl ahead of me in like each take this long.
Mystery persists
Reply to @codytfenwick, 05:10 UTC · 9 likes
Cody Fenwick (@codytfenwick): @AaronBergman18 Maybe you're the one who inspired it all
Sorry :/
Reply to @mauddweeb, 05:19 UTC · 25 likes
fedaykin reepicheep (@mauddweeb): > animal welfare
> advocates screworm genocidetsk tsk
(quoting @AaronBergman18: When your friend has access to free poster printing (reformatted a bit))
Reply to @mauddweeb, 05:23 UTC · 5 likes
fedaykin reepicheep (@mauddweeb): @AaronBergman18 Mostly teasing but I am curious to hear why the worms don’t get welfare
Their welfare does matter but (1) killing them all doesn’t necessary involve suffering (that’s something to avoid) and (2) the whole “eating other animals alive” thing seems like the much bigger deal, and (3) it’s not even clear that life is generally good for screwworms themselves, nonexistence might be better
Tweet, 14:37 UTC · 104 likes
You could be on an international crisis epidemic ship with a trivially obvious explanatory hypothesis that fits the data and still get told that it’s “just anxiety”
Reply to @jerrrrrrryyyyy, 14:57 UTC · 4 likes
jerry (@jerrrrrrryyyyy): Oh no you're the reason it does this
😢
Reply to @LauraFlowD, 17:35 UTC · 2 likes
laura ⚔️ (@LauraFlowD): @AaronBergman18 @gptbrooke everything is anxiety when you don’t want to treat
💯
Reply to @cynlioness, 17:40 UTC · 12 likes
Cyn 🔸⏸️ (@cynlioness): Looking to make / join a group chat or discord where we talk about fashion / post fit checks and people give nuanced suggestions. Same for interior design.
Is the idea like you have to contribute value or do u accept charity cases
Reply to @mattyglesias, 17:42 UTC · 14 likes
(The tweet being replied to by @mattyglesias is no longer available.)
Proxy for happiness and intelligence
Tweet, 23:17 UTC · 300 likes
Mildly unsettling that Claude via Claude Code has started referring to *my* macbook as "my [ie Claude's] macbook"
Reply to @PAHoyeck, 23:41 UTC · 2 likes
Phil Hoyeck (@PAHoyeck): Part III is among the greatest philosophy ever written.
Ugh maybe I should actually read it all
Reply to @sympatheticopp, 23:46 UTC · 19 likes
sympathetic opposition (@sympatheticopp): high decoupling only gets you closer to the truth in cases where the things you're decoupling don't actually have a relationship
I think this is a confused version of the concept
To my understanding, decoupling is the ability to and practice of considering a thing in the abstract, not the practice of refusing to acknowledge actual correlations or causal relationships or anything
High decoupling is a superset of capabilities with respect to low decoupling
Quote tweet, 23:55 UTC · 12 likes
Hate to be a woke lib but it would be much cooler to watch if it didn’t involve several young men risking/sacrificing their cognitive futures with every blow to the head
Quoting John Massingale (@JohnMassingale4): It never doesn’t amaze me.
One of my very best friends is a USC fan and he texted me saying “that run is art”
Wednesday, May 13, 2026
Reply to @tszzl, 00:54 UTC · 1 like
roon (@tszzl): it always boils down to ingroup good
Insofar as this is true, it's downstream of actual actions (which may include expressing beliefs)
OpenAI used to be in good graces of EA - their actions made them outgroup, not the reverse
Reply to @tszzl, 03:51 UTC · 5 likes
roon (@tszzl): and what are those actions? the explanation for why openai is bad will always take the shape of “so and so on my side was betrayed” rather than “here is some tangible evidence of harm or malice, that cannot be levied against anthropic”
Hostile to original mission restructuring, Sam constantly lying, coercive NDA thing, ~total abandonment of original mission. Anthropic alone is taking any steps on model welfare. The OpenAI foundation or whatever it’s called are writing checks that have nothing to do with the mission and nobody seriously thinks are good uses of money.
I can’t write these cleanly in first order logic for you but these aren’t intractably intangible - you can criticize Dario for being “grandiose” but any neutral analysis will conclude that he doesn’t lie like Sam does. And there hasn’t been the anything remotely like OpenAI’s mission pivot
You can criticize EAs liking the more EA-influenced company as pure ingroup social dynamics and tribal affiliation but the problem with that is that it’s at most a relatively unimportant ancillary dynamic and what’s actually going on is that the company founded by people with a prosocial worldview ended up making a prosocial company
Reply to @tautologer, 03:58 UTC · 2 likes
tautologer (@tautologer): my unemployed ass be like
Real
Reply to @edgar_z_lin, 04:28 UTC · 1 like
Edgar Lin (@edgar_z_lin): @AaronBergman18 @PAHoyeck @mishapathy how have you of all people not?? do it right now!
😔
Reply to @EigenGender, 12:52 UTC · 3 likes
EigenGender 🔸 (@EigenGender): do you want to say interesting things or correct things
Both
Reply to @bkpark, 13:28 UTC · 2 likes
bkpark (@bkpark): fortunately: https://www.pangram.com/history/6b09cde4-6dc8-4c7d-be86-72d5ca31bd58?ucc=v6DoZfjLpds
Exonerated
Quote tweet, 14:25 UTC · 16 likes
Lmaoooooooo
Quoting Byrne Hobart (@ByrneHobart): @AaronBergman18 @amelapay recently informed me that I did it, too.
Tweet, 19:07 UTC · 27 likes
Why don’t we have a hantavirus vaccine yet
It’s been like a week
Reply to @AndyMasley, 19:25 UTC · 6 likes
Andy Masley (@AndyMasley): I literally just hate PM2.5 I have no other politics
Can I count on your vote
Quote tweet, 19:46 UTC · 98 likes
Yet another drug that should be OTC
Quoting keysmashbandit (@keysmashbandit): Periodic reminder that PRN Propranolol changed my life
If you miscellaneously "don't like socializing" you may be socially anxious without realizing
Reply to @afoidance, 19:48 UTC · 1 like
ren 🇮🇱 (@afoidance): asking a guy out is a cheat code for getting a boyfriend. they literally cant say no. its like rape
Hmm
Reply to @hopes_revenge, 19:49 UTC · 2 likes
hope hopes hoping (@hopes_revenge): need frontier lab job so i can use health & wellness stipend to buy $ 2,200 apple studio display
pleas DM
Real
Reply to @Scott_Wiener, 19:52 UTC · 74 likes
Senator Scott Wiener (@Scott_Wiener): This is shameful & should be illegal.
(quoting @MorePerfectUS: Nearly 50,000 people in the Lake Tahoe area have been told that their utility will stop providing power to them, because it's redirecting that power to data centers. NV Energy, the Nevada utility that has supplied most of Lake Tahoe’s electricity for decades, says that next year…)
Scott u are better than this
The whole thing is .0000000000001% off from an absolute lie
Quote tweet, 20:04 UTC · 160 likes
Racial identity games are not the path to a good and just society
Our institutions are not good enough to handle the minute delta between “in-principle ideal amount of race consideration” and “fake-it-till-you-make-it colorblindness”, so we should do the latter
Quoting Chris Brunet (@chrisbrunet): > Indian professor classifies Chinese professor as white in HR records to justify denying him a promotion
amazing
Reply to @TomDavidsonX, 20:12 UTC · 3 likes
Tom Davidson (@TomDavidsonX): With @joemkwon @alfie_lamerton @AndrewDraganov @DaveRBanerjee @BronsonSchoen @matpistillo @DKokotajlo @ryangreenblatt @OwainEvans_UK @Manderljung @FabienDRoger
Thanks to input from @_achan96_ , Aniket Chakravorty, @AbbeyChaver, @LukasFinnveden, @Tim_Hua_, @MaxNadeau_ , @richardson_aris, @AlexandraSouly, @a_nnawang
Good follow list
Thursday, May 14, 2026
Quote tweet, 12:23 UTC · 72 likes
Imagine getting banned from 80k
Quoting SIMONE 🩷 (@parasocialyte): a boyfriend who watches porn is the least of your worries, it's the podcasts you have to be really strict about
Quote tweet, 12:24 UTC · 19 likes
I love this app
Quoting Awais Aftab (@awaisaftab): I open this damn app and look at my timeline, and I feel my IQ dropping and my blood boiling, and yet I keep opening it again every day like some masochistic freak.
Reply to @tiksiminic, 20:59 UTC · 4 likes
(The tweet being replied to by @tiksiminic is no longer available.)
80,000 Hours
Friday, May 15, 2026
Reply to @francoisfleuret, 00:11 UTC · 2 likes
François Fleuret (@francoisfleuret): Hot take: machine learning and AI did more to understand the nature of knowledge, and our relation to reality than 20 centuries of philosophy.
I am ready to kind of defend this hill.
Lol reframe as the last 100 years of analytic philosophy and I think philosophy wins
Tweet, 00:16 UTC · 662 likes
I do find it weird that somehow the US keeps winning despite a much stronger culture of intense education + broadly higher IQ distribution in China
And yet somehow the golden goose is always born here (OpenAI -> Anthropic this time around)
Reply to @quinnmbarry, 00:20 UTC · 9 likes
Quinn Barry (@quinnmbarry): capital controls in China + clear corporate rules US seem like big reasons
Ye
Reply to @AndyMasley, 00:32 UTC · 8 likes
Andy Masley (@AndyMasley): Can I be a good member of my community if once a year I send money to malaria charities?
May as well vandalize the local library
Reply to @shlummi, 03:54 UTC · 8 likes
shlummi (@shlummi): lmfao
Lmao rip
Maybe I’d do $50/date
Reply to @NavinFS, 06:35 UTC · 3 likes
NavinF (@NavinFS):
(quoting @NavinFS: @Lib_Development Helps that I can sponsor h1b and hire the best applicant from the whole planet. In any other country I'd only get the guys that couldn't make it to the US)
💯
Reply to his own earlier tweet, 07:22 UTC · 47 likes
Aaron Bergman, May 15, 2026: I do find it weird that somehow the US keeps winning despite a much stronger culture of intense education + broadly higher IQ distribution in China And yet…
Some people here are (perhaps reasonably) attributing to me an ideological lens that I don’t have
Not an IQ determinist; not a fan of China; definitely don’t think the US is “winning” in every sense of the term
Reply to @PaulaSeeksTruth, 18:10 UTC · 6 likes
(The tweet being replied to by @PaulaSeeksTruth is no longer available.)
Still hoping I learn some day
Quote tweet, 20:13 UTC · 57 likes
Bruh
Quoting Nate Silver (@NateSilver538): BTW, I approached ABC about buying back the former FiveThirtyEight IP*, and they said they wouldn't sell at any price because I'd criticized their management of the brand. Costing Disney shareholders $$ b/c of their vindictiveness.
* I own the models but the trademarks, etc.
Reply to @literally_chad, 20:18 UTC · 17 likes
Literally Chad (@literally_chad): I'm once again relearing that there are men who take rejection personally
Rejection is personal, though. If that’s not personal then what is?
Tweet, 23:04 UTC · 38 likes
Alternative timeline where proto-Anthropic’s origins come after FTX collapse, Dario and friends heed the zeitgeist and decide EAs shouldn’t be ambitious after all, Anthropic fizzles as an idea on a whiteboard
Much worse world IMO
Reply to @justjoshinyou13, 23:32 UTC · 2 likes
Josh You (@justjoshinyou13): seems to overstate the influence of EA community events on Dario, relative to his specific vision for AI? he was one of the people most responsible for LLM scaling and GPT-3 at OpenAI!
Tbc I didn’t really intend this as a propositional counterfactual claim about what would have actually happened
Although my overall impression is probably directionally as you suggest, 🤷♂️
Saturday, May 16, 2026
Reply to @mattyglesias, 03:41 UTC · 33 likes
(The tweet being replied to by @mattyglesias is no longer available.)
Medical tourism seems like an existence proof of PPP being legit - there are irl arbitrage/savings opportunities (turkey hair transplant) that tons of real people exploit
You might want to object that different regulatory regimes make the same thing a different product, but then compare US States to Puerto Rico for cost of elective procedures
Reply to @Kirsten3531, 03:44 UTC · 23 likes
Kirsten (@Kirsten3531): My cousin is betting his career on "LLMs can never be more than the average of their training data" but I feel like that's a very 2024 take? Aren't we already past this in like, coding and math?
Base LLMs maybe, not commercial chatbots
The “helpful smart thoughtful” AI persona alone bumps you up in quality before even doing RL
Reply to @Lib_Development, 05:47 UTC · 4 likes
Lucas Pombo (aka Lib Development) (@Lib_Development): Woulda been nice if the new icon didn’t make it look like the app was updating tho
Exactly
Tweet, 05:57 UTC · 315 likes
It’s a bad sign for the Overton window that the radically altruistic take on AI redistribution implicitly excludes the vast majority of people
I’m not faulting Dylan et al for what might be the most productive move right now tbc, it’s just grim
Reply to @anglerfish01, 06:10 UTC · 3 likes
angLerfish (@anglerfish01): >P(doom)=60%
>⏸️
>anthropic is in the lead accelerating the race to ASI the most with their RSI focus
>supports anthropicWut
Can you explain?
Sounds like more galaxy brained SBF stuff tbh
I think talk is cheap and incentives matter a lot btw
Yeah good points; I just removed pause emoji bc I’ve been wavering around 50% endorsement for some time. Not necessarily gone forever
In most worlds either normal political incentives + luck work or they fail, there’s only a slice in which it really matters who’s doing the building. Anthropic seems like one of the best possible organizations to be in the lead in that middle slice of worlds; corporate governance is solid, emphasis and culture of safety, quite frankly the fact that founders pledged 80% of wealth to charity makes a big difference in economic windfall worlds. The mass surveillance guardrails red line matters; commitment to liberal democracy matters, and willingness to pay an alignment tax which they may be able to competitively afford
No Anthropic means OpenAI with a psychopathic cult of personality CEO and afaict no serious moral consideration of anything institutionally embedded via culture or corporate governance or anything in the lead instead. Or Google, maybe. And maybe timelines are a few months longer. This does not feel like a better situation to me
In the middle slice where ideosyncratic decisions of companies and people really make the difference, they’re pretty close to the best shot we could ask for
Reply to @mattyglesias, 13:02 UTC · 9 likes
(The tweet being replied to by @mattyglesias is no longer available.)
I don’t get the objection here (although this is probably motivated by me being a 🫛-cel)
Maybe there was a structural issue before reliable transcription but now you can translate easily between text and audio. Is it the medium or the content you object to?
Reply to @justalexoki, 16:07 UTC · 3 likes
taoki (@justalexoki): just saw a pigeon rape another pigeon
😢
Sunday, May 17, 2026
Reply to @_sholtodouglas, 03:15 UTC · 12 likes
Sholto Douglas (@_sholtodouglas): When do you reach for other models instead of Claude? What can we do better? Hit me with all of your frustrations. dms open.
If you can give me detail (e.g. specifics/transcipts) - it'll help a lot in finding out exactly what we need to do to improve the next model
I use http://aistudio.google.com when I want a quick way of getting a top model to read the whole context, not gradually parts of a file and then with uncertainty about what it’s seen. This isn’t a problem with the model itself ofc but still. Ofc I’m willing to pay for those tokens
Gemini 3.1 Pro for anything where audio/image/video understanding seems somewhat difficult and important
grok-fast (whatever the latest one is called) is ridiculously cheap and fast and that really does open up a bunch of use cases but looks like it’s retired as of yesterday
Tweet, 05:18 UTC · 21 likes
Slack needs an icon for "Working rn (night/weekend) but in a 'incapable of working normal amounts at normal times' way not a workaholic grindset way"
Reply to @JohnWittle, 05:22 UTC · 1 like
John Wittle (@JohnWittle): (i just tell claude to ignore view_file and instead boot up their bash environment, and literally run cat /mnt/user-uploads/document.txt or whatever
but i'm always scared to learn that the bash_command toolcall will have been patched to truncate)
ooh smart
Reply to @deanwball, 07:59 UTC · 2 likes
Dean W. Ball (@deanwball): it is shocking to me how bad/overeager anthropic's bio-related safeguards are. actively user hostile. I am hitting them doing *stock* research.
I basically agree that they’re extremely annoying but it just seems very hard to have extremely few false positives with a cheap classifier
The expensive classifier would make Claude cost more, presumably. Although plausibly there should be an option to pay for that?
Reply to @austinc3301, 09:18 UTC · 5 likes
Agus 🔸 (@austinc3301): wait qué? dario amodei (ceo de Anthropic) estudió física, no literatura
Hola señor Agus
Reply, 14:37 UTC · 1 like
😢
Reply to @maiamindel, 15:20 UTC · 120 likes
Maia (@maiamindel): kind of hysterical how every male authrotiy figure he's encountered so far is a lot handsomer than he is without meth, hammers, or math. should make him rethink his worldview i guess
(quoting @Polymarket: JUST IN: Looksmaxxer, Clavicular, mogged by “Chad” judge in Miami court hearing.)
Lol I’m no fan of the guy but the whole point is that those authority figures got lucky and it plausibly helped place them in positions of authority
Reply to @PaulaSeeksTruth, 15:23 UTC · 4 likes
(The tweet being replied to by @PaulaSeeksTruth is no longer available.)
Post duck
Reply to @PaulaSeeksTruth, 15:26 UTC · 21 likes
(The tweet being replied to by @PaulaSeeksTruth is no longer available.)
I’m pretty happy to use this analogy but in both cases you have to account for the direct benefits, not just indirect costs
People *like* drinking and being drunk just as they like porn
Maybe you think the cost-benefit doesn’t check out but I feel like people are blind to the obvious sometimes
Tweet, 17:06 UTC · 1,031 likes
Ominous
Reply to @cafreiman, 18:10 UTC · 31 likes
Chris Freiman (@cafreiman): Being able to feed yourself for a day with fifteen minutes of work rather than six hours does seem like a mark of prosperity.
(quoting @halsinger: I blanch at the way neoliberals define progress aka “productivity” as “the ability to produce more with the same labour.” It implies that a declining labor share could be a good thing for a society—even a mark of prosperity.)
Hey I agree but the progressive/left worry about declining labor share is totally legit IMO
The economic conditions of broadly the last 100-300 years are not permanent and there is no rule in economics that says that you will be able to make more than epsilon from your labor or that the growing pie will outpace the shrinking labor share
Reply to @Mjreard, 18:12 UTC · 6 likes
Matt Reardon (@Mjreard): The proper role of EA is to identify the broadly important problems/questions/framings. From there, you need to follow your interest/curiosity, not make galaxy brained plans to control last mile outcomes
(quoting @RichardMCNgo: You might instead argue that EA’s biggest impact has been building the field of AI safety. But currently I think the main effect of EA has been to turn AI safety into much more of a fake field (like, say, academic psychology). There’s a decade-long gap (around 2016-2024) where…)
Ehhhhhh idk I feel like I’m guilty of following my interest and curiosity too much and at this specific point in history the last mile thing might actually work
Tweet, 20:47 UTC · 91 likes
The datacenter heat thing is interesting bc you should expect heat output to be ~exactly proportional to energy consumption (right?) but in general even people on the left implicitly round the non-greenhouse-mediated, direct warming effect of energy expenditure to zero
Thread, continued, 20:47 UTC · 33 likes
(In that the effect is almost never discussed in any other context)
It’s not literally zero ofc, it’s just much much smaller than the temperature/climate change mediated by the greenhouse effect
Tweet, 20:55 UTC · 40 likes
Utility of a cucumber is inversely related to the thickness of a slice
Lots of alpha in cutting cucumbers really thin. Presumably there are devices for this
Reply to @bschne, 23:05 UTC · 4 likes
Benjamin (@bschne): @AaronBergman18
lmao
Reply to @AndyMasley, 23:28 UTC · 27 likes
Andy Masley (@AndyMasley): Sorry I besmirched farmers, I found a worse one.
I continue to wonder if I seriously harmed myself by going camping and breathing in a normal amount of campfire smoke something like once a month for most of my adolescence
What a dumb way to lose one’s potential
Reply to @EigenGender, 23:33 UTC · 17 likes
EigenGender 🔸 (@EigenGender): yeah I think it’s an underrated fact among abundance-type people that we can’t actually increase global energy use by that much more before the direct heat output becomes substantial, and so there’s a real limit to nuclear
Real
Thread, continued, 23:36 UTC · 10 likes
Although there are definitely potential engineering solutions right: geoengineering to offset heat, both electricity generation and industry outside of earth’s atmosphere
Reply to @AndyMasley, 23:45 UTC · 16 likes
Andy Masley (@AndyMasley): A quick botec suggests this is about as bad for you as moving from DC to New Jersey
Well maybe that just means ppl in New Jersey are fucked
Monday, May 18, 2026
Tweet, 03:33 UTC · 63 likes
I’m gonna be pissed af if I fix my health and then soon thereafter die from some sort of AI disaster. I feel like I should have the opportunity to appeal somehow
I don’t expect this as a median case but I think it’s totally plausible
Reply to @izzyz, 03:37 UTC · 4 likes
izzy (@izzyz): @AaronBergman18 here's the plan: live as close as possible to a data center cuz think about it why would the ai wreak havoc near its own brain cells
Smart
Reply to @AlexKrusz, 03:41 UTC · 5 likes
Alex Krusz (@AlexKrusz): Other disasters are still cool tho?
Hmm Ill have to think about this one
Tweet, 05:26 UTC · 82 likes
MY DATA HOARDING FINALLY CAME IN HANDY I HAVE A REAL USE CASE (that I can't share rip) FOR A 1.1 TB DATASET LIVING ON AN EXTERNAL HARD DRIVE
Reply to @Bananaplanet2, 05:31 UTC · 2 likes
Bananaplanet🐌 (@Bananaplanet2): If we could build really tall towers in the ocean to transport water heat by conduction up to the atmosphere where it dissipates by conduction or radiation, we would simultaneously cool the oceans and also interrupt thermoclines, improving productivity of the oceanic carbon sink
(quoting @AaronBergman18: Apparently 1%-ish of global warming comes from the literal heat of burning stuff Which isn’t a lot but it’s not hard to imagine scaling up nuclear 100x or way more to the point that we get serious warming by a totally new mechanism You never really hear about this though)
Sounds good to me, maybe a few ppl should study this first tho
Tweet, 06:16 UTC · 2,118 likes
Nobody:
Somebody somewhere occasionally: uses the term "smoke test" very occasionally
Claude Opus 4.7 every other sentence: let's run a smoke test
Tweet, 06:53 UTC · 1,052 likes
This is obviously pretty bad object level but also:
Kind of insane that late 20th century American feminists managed to solidify and then export the (historically unusual) concept+ lens and terminology of “consent” to the fucking Taliban
Thread, continued, 06:53 UTC · 96 likes
The converse to illustrate: American feminists arguing for their theses using the concepts and vocabulary of orthodox Islam: what Allah approves of, the interpretation of passages in the Quran, etc
Reply to @ApriiSR, 06:54 UTC · 70 likes
Aprii *️⃣ (@ApriiSR): it's an important concept!
This tweet is a smoke test
Reply to @freed_dfilan, 07:26 UTC · 32 likes
Daniel Filan 🔎 (@freed_dfilan): Consent as a requirement to enter contracts (like marriage) isn't that unusual tho right? Like isn't it a feature of early Islamic jurisprudence?
Ok yeah this is pretty fair it turns out, it’s very much in the water, although only in the context of consent to marry which then is taken to imply sex is proper, not consent to sex per se (thx Claude)
Honestly I was sorta half-consciously using this as a reason to make the broader point about consent as the dominant framework for sexual ethics which seems to me more interesting and non-obvious than everyone else seems to take it as
But weirdly it’s a real feature that appears both in early abrahamic religions and then again resurging with modern-ish feminism
Reply to his own earlier tweet, 07:27 UTC · 68 likes
Aaron Bergman, May 18, 2026: This is obviously pretty bad object level but also: Kind of insane that late 20th century American feminists managed to solidify and then export the…
Yeah this is a good point by Daniel
Reply to @AfflatusSolis, 07:28 UTC · 10 likes
Afflatus Solis (@AfflatusSolis): Consent is an idea older than feminism. Claude tells me Sharia does engage with it
Yeah maybe I should delete, I’ll leave up for now tho
Quote tweet, 21:23 UTC · 39 likes
I’d kms
Quoting yuuii (@yuuiichive): There was a Guy in my class this semester who kept using the word ontologically wrong and one day my professor finally broke and said “Ontologically can be a tricky word to use so don’t feel like you have to” and then the guy stopped showing up for classes entirely
Tuesday, May 19, 2026
Quote tweet, 04:59 UTC · 32 likes
This would work on me
Quoting lucas gelfond (@gucaslelfond): a guy just came up to me on metro north and said “you look like the sort of guy who would have a USB-C charger”
Reply to @JoelWBerry, 05:03 UTC · 40 likes
Joel Berry (@JoelWBerry): A little-known hack:
2 slices of Aldi wheat bread: $0.17
3 oz of Aldi deli turkey: $0.86
1 slices Aldi cheddar: $0.15
1 condiment of your choice: $0.02
1 apple: $0.53
1 hard boiled egg: $0.14
5 carrot sticks: $0.17
Cold water from the tap: $0.01Total: $2.05
You can do this.
(quoting @SeanMcCarthyCom: Lunch just costs $28 now. Are they not supposed to eat?)
Tbf that’s like 400 calories, maybe if you’re a small female desk worker on a diet
Quote tweet, 05:04 UTC · 12 likes
Follow Tommy
He has correct takes about important things
(Also a guest on pigeon hour as of shortly in the future)
Quoting Tommy Crow 🏗️ (@TommyJoanCrow): I'm working on an article for a magazine touching on pain management during post-surgical extubation, and once again I find myself freshly disturbed by how difficult it is to get good data about PAIN during extubation (and many other procedures) because the dominant culture in "pain management" cares so little about pain itself. They often care exclusively about preventing memories of pain, rather than preventing pain itself.
Quote tweet, 21:23 UTC · 45 likes
Unbelievable
Quoting Tyler McBrien (@TylerMcBrien): BLANCHE: "The United States...is hereby FOREVER BARRED and PRECLUDED from prosecuting or pursuing...examinations or similar or related reviews" against Trump "or related or affiliated individuals," including family members or related companies and trusts.
Reply to @mpopv, 23:31 UTC · 9 likes
Matt Popovich (@mpopv): for a second there my tech bro ass thought Knicks-6.5 was the name of a new model
No matter how many times I listen to this my brain treats the phrase as “P(Diddy)”
Wednesday, May 20, 2026
Reply to @daniel_271828, 00:31 UTC · 5 likes
zzDANIEL_ETH (@daniel_271828): “Dario doesn’t look that charismatic to me atleast from his interviews”
Well, okay, presumably that’s not the main criteria by which Andrej is choosing which company to join
(quoting @vijaythirumalai: I simply cannot fathom this move Dario doesn’t look that charismatic to me atleast from his interviews Andrej could have joined 1/ Elon 2/ Zuck 3/ Google 4/ Apple 5/ MSFT Everyone would have given him a blank check and much bigger role and platform How was Dario able to hire him…)
Lmao
Reply to @earnestpost, 00:39 UTC · 2 likes
Erin Constance 🇪🇺⏸️ (@earnestpost): The concept of dominating a mallard duck
Collaborate with the duck
Reply to @Empty_America, 01:11 UTC · 3 likes
VB Knives (@Empty_America): Perfect example of the "commiseration seeking" style of communication.
The only solution-oriented response to complaints about a $28 sandwich is "pack a lunch."
I mean you either pay, don't eat lunch, or pack a sack.
But they are looking for a emotional "acknowledgement."
(quoting @Schmutz_7R: My grandpop raised five kids on a welder’s salary in a home on ten acres with a pond in the middle. He had fourwheelers, campers, a boat, tons of fishing and hunting gear. And I logon to X to see some boomer telling me to start packing my lunch if I want a better life.)
Yeah and that’s not necessarily a bad thing
Thursday, May 21, 2026
Reply to @MurrayHillGuy1, 15:45 UTC · 2 likes
Murray Hill Guy (@MurrayHillGuy1): Women talk like they are in HR when they are done with you
RIP
Reply to @ArringtonLuck, 16:46 UTC · 2 likes
Arrington Luck (@ArringtonLuck): CHIPS needed a team that could sit across the table from large, well-capitalized semiconductor firms and negotiate deals that both protected the taxpayer and delivered on national security objectives.
A lot of the program success was built on the backs of government operators and national security professionals. Equally important, however, were the private sector hires that the program brought onboard.
This week in Factory Settings, I took a look at how the program was able to effectively recruit private sector talent.
A few of the most salient points:
- Program leaders’ decision to go on the Odd Lots podcast was referenced by several colleagues, with non-traditional recruiting pathways serving as critical for getting team members- CHIPS gave significant empowerment to junior colleagues, allowing them deal reps well above their private sector counterparts
- The program had a tangible, positive impact on the career trajectories of the private sector talent that came through the program
Private sector talent isn't everything, and traditional government competencies are equally critical. Nevertheless, future industrial policy efforts would do well to make a concerted effort to get ex-financiers, operations professionals, and technologists working towards program success.
Read here: https://www.factorysettings.org/p/a-cheat-sheet-for-attracting-private
LeBron?
Friday, May 22, 2026
Reply to @sama, 00:27 UTC · 9 likes
Sam Altman (@sama): what problem do you most hope AI will solve in the future?
maybe we can help!
You
Thread, continued, 00:30 UTC · 2 likes
Also suffering tho
Tweet, 07:28 UTC · 89 likes
Gravity is kinda fucked up when you defamiliarize yourself with it
I don’t think objects super far away, like trillions of miles, should act on one -another like that, even a little bit
Reply to @mattheard, 07:37 UTC · 17 likes
mattheard.delve.town (@mattheard): changes in gravity propagate causally at light speed
Fucked up
Reply to @AndyMasley, 12:18 UTC · 3 likes
Andy Masley (@AndyMasley): lol okay I believe I might be the first person to have discovered the full final story of exactly where the "Bottle of water" per prompt number actually came from. Will be in the next post.
What
Reply to @LeRoyDesCimes, 12:19 UTC · 5 likes
Louise 🔸 (@LeRoyDesCimes): Asked Claude to summarise some transcripts and learned we should send promising AI safety youth to MARS 🚀🚀🚀
I have a pipeline to make more accurate transcripts from audio using semantic context if interested
Saturday, May 23, 2026
Reply to @SALTWATERSOPH, 04:51 UTC · 2 likes
Sophie (@SALTWATERSOPH): WE will learn real analysis with 5 minutes a day over one month 👏
That’s 150% more than desired
Reply to @GWHayduke97, 04:58 UTC · 2 likes
Hayduke ⏹️ (@GWHayduke97): Does ⚓️ mean Anglicanism(+Episcopalianism?) in general or specifically ACNA?
Does 🦬 mean (non-ELCA) Lutheranism in general or specifically LCMS?
And are there any other denominational Twitter emojis?
There are parts of Twitter beyond my wildest conception
Reply to @romlib_, 08:22 UTC · 4 likes
Romlib 🎄 (@romlib_):
image attached
(quoting @beffjezos: Feeling robbed of my path to citizenship right now after grinding a PhD and contributing to foundational AI + computing technologies for the United States for the past ~ 10 years. Feels like robbing top and technologists like me of the opportunity to achieve the American Dream.)
Yes but also he’s right that it’s bad
Reply to @JesseTayRiver, 16:08 UTC · 3 likes
Jesse Smith (@JesseTayRiver): Every morning my wife does something she calls meditation where she puts on a daily meditation podcast and then frantically types patient notes on her laptop with it playing in the background
Lmao
Reply to @Adellama_, 16:13 UTC · 3 likes
Ada🔸 (@Adellama_): that is a very pretty outfit!
It’s just the goose’s natural fur
Reply to @funplings, 20:03 UTC · 2 likes
funplings (@funplings): @earnestpost i am patient. the reply guy measures his plans in centuries
I respect the grind
Reply to @NunoSempere, 20:05 UTC · 3 likes
Nuño Sempere (@NunoSempere): Recently got my heart crushed, please send encouragements
Rip, you seem like a handsome successful guy so I’m not worried about your larger prospects
Tweet, 20:56 UTC · 18 likes
Bruh
Reply to @mattyglesias, 22:14 UTC · 14 likes
(The tweet being replied to by @mattyglesias is no longer available.)
The woke take is that 1/3 shows evidence of pro-American (or pro-white) bias because <1/3 of the world population is American
Quote tweet, 23:54 UTC · 75 likes
I think it would be potentially cool for a society to fight phone addiction at the national level. I’m not a libertarian precisely bc collective action problems like this exist
But I get very skeptical when the decision making procedure is “exempt everyone who is enfranchised
Quoting Sam Bowman (@s8mb): Of all the “protect kids from the internet” ideas, the one I like the sound of most is France’s 10pm internet curfew for teens. It sounds so ridiculous, but actually addresses one of the big collective action problems in a pretty low cost way. Curious to see how it goes for them.
Thread, continued, 23:54 UTC · 16 likes
and target exactly the main population in society without a vote”
Like wow what a coincidence that the best policy is as anti-democratic as you can possibly make it
Thread, continued, 23:54 UTC · 11 likes
How about a full scale population tax on screen time? Or some other policy idk - presumably some smart people have thought about this and written PDFs - anything but “smuggle this in through the most glaring loophole in modern western ~democratic politics”
Thread, continued, 23:54 UTC · 11 likes
I actually kinda like the screen time tax idea, I can totally see myself opting in to a society with such a policy
I propose $1/hour, to start, now fight about it!
Reply to @Lux_Stella_, 23:55 UTC · 10 likes
Lux_Stella (@Lux_Stella_): i genuinely think "you're allowed unfiltered access to the internet, but only through a linux terminal" would really solve a lot of problems
(quoting @s8mb: Of all the “protect kids from the internet” ideas, the one I like the sound of most is France’s 10pm internet curfew for teens. It sounds so ridiculous, but actually addresses one of the big collective action problems in a pretty low cost way. Curious to see how it goes for them.)
Oh cmon they can have macOS or powershell access
Sunday, May 24, 2026
Quote tweet, 01:22 UTC · 218 likes
I've been hardened by the Hinge mines at this point but the sudden "potential partner" -> "we will never interact again" dynamic is objectively kinda weird
Of course it makes things easier in some ways rather than harder, but the human brain isn't build for that nonetheless
Quoting Someone:
Reply to @_eleanorina, 01:22 UTC · 5 likes
eleanorina (@_eleanorina): as a treat
powershell as a punishment
Reply to @sariverse, 10:45 UTC · 267 likes
sarina (@sariverse): my friend’s showing me his hinge.… just a peek into the male mind
No lol I don’t think you understand
This is not a peek into the male mind in a normal sense, it’s a peak into the state the male mind gets ~forced into when dealing with dating apps
The amount of time/energy it takes to go on a single irl date is high enough that it stops being worth it to engage with women who aren’t receptive to going on an irl date soon/after just a bit of messaging (or none tbh)
This isn’t a judgement about them - its just the least bad way I know of to deal with limited time and energy
I usually ask after a few messages “would you like to go on a date soon or do you prefer to message for a while first” and frankly if they indicate the latter it’s generally just not a worthwhile use of time
Tbc I think this is bad and one of many ways that dating apps ~force people to act in ways that they don’t really aspire to even if they endorse given the real constraints
Reply to @questionableway, 10:46 UTC · 3 likes
just matt (@questionableway): this also allows people to treat potential partners on hinge in ways that would have historically been considered unacceptable
(quoting @AaronBergman18: I've been hardened by the Hinge mines at this point but the sudden "potential partner" -> "we will never interact again" dynamic is objectively kinda weird Of course it makes things easier in some ways rather than harder, but the human brain isn't build for that nonetheless)
💯
Quote tweet, 12:43 UTC · 32 likes
Lmao this sounds like in the ballpark of a healthy hit rate?
Ofc the distress part of anxiety is bad independently
Quoting Brandon Luu, MD (@BrandonLuuMD): In a study of people with generalized anxiety disorder, only 8.6% of recorded worries actually happened.
The most common outcome: 0%. Not a single worry came true.
Reply to @ghoulhag, 13:32 UTC · 15 likes
🪽 (@ghoulhag): straight men: orgasm
gay men: orgasm
lesbians: orgasm
straight women: honestly it’s no big deal i’m just here for the cuddles i don’t mind it’s okay that i don’t orgasm sex is sex 😭❤️ we’re all trying our best it’s not his fault
But I am trying my best!
Quote tweet, 17:25 UTC · 257 likes
Wtf
I don’t wanna be too flippant because I know it’s a substantial or even prohibitive cost for some people but you can just do things (buy and install a <$200 window AC unit)
Quoting Noah Smith 🐇🇺🇸🇺🇦🇹🇼 (@Noahpinion): Economists estimate that if Europeans used AC as much as Americans do, it would save up to 100,000 European lives EVERY YEAR.
But I guess saving face on Elon Musk's social media app is more important than 100,000 lives.
Thread, continued, 17:25 UTC · 28 likes
It probably pays for itself in productivity, it’s extremely overdetermined, you don’t need a $10k full house HVAC renovation to make 1-2 of the rooms you spend the most time in comfortable during the summer
Thread, continued, 17:25 UTC · 15 likes
If the cost is really an issue and we’re mutuals then DM me (I have no idea what my plan is here tbc)
Reply to @mattyglesias, 20:25 UTC · 65 likes
(The tweet being replied to by @mattyglesias is no longer available.)
RIP, I think I unironically endorse civil disobedience about this (although cheap/easy for me to say)
What happens if you just vibe-install one? Are you gonna go to prison?
Reply to @mattyglesias, 20:40 UTC · 8 likes
(The tweet being replied to by @mattyglesias is no longer available.)
Well potentially you could figure out how to change the window situation itself
Monday, May 25, 2026
Quote tweet, 01:19 UTC · 11 likes
Tommy is a mensch and also has lots of very important and correct takes!
Let it be known that I chose the podcast title myself!
Quoting Tommy Crow 🏗️ (@TommyJoanCrow): It was such a blast doing this podcast with @AaronBergman18, who is both hilarious and has good values, and his choice of stillframe to pull from my TikTok is cracking me up (clothed and professional 🚫 shirtless and doing drugs ✅)
https://open.substack.com/pub/aaronbergman/p/tommy-crow?utm_source=share&utm_medium=android&r=6sxtkh
Tweet, 02:17 UTC · 11 likes
@Aella_Girl there is way way way too much going on on Glosso lmao. It's like a infinite nested array of pseudo-connected sub-sites which, cool, but I'm guessing we're well into the part of the curve where adding new features decreases value
I say this as a fan!
Reply to @dioscuri, 02:55 UTC · 22 likes
Henry Shevlin (@dioscuri): Seriously though, why is there something borderline psychoactive about this specific set of motifs? Why pools? Why no windows? Why the pastel colours? It feels suspiciously like an arbitrary code execution exploit on the human mind.
I just…don’t get that profound of an effect
Tweet, 03:00 UTC · 21 likes
How do I rizz Gracie Abrams
Reply to @_job01, 03:41 UTC · 2 likes
job (@_job01): Were you raised in a standard middle class ish lifestyle in the US? I’ve asked if Gen Z experiences it and it seems like yes because I initially thought it’s an effect of the types of plastics, and coloring, and lighting used for certain venues for millennial childhoods
Yeah on the first thing. Idk other things do it for me but not really this
Reply to @AlexTPet, 07:39 UTC · 2 likes
Alex Petropoulos 🤠 (@AlexTPet):
bruh
Tweet, 07:40 UTC · 5 likes
Why not, anyone?
https://claude.ai/referral/eEX0WyvYzg?s=cowork&v=apps
Reply to @bashu_thanks, 07:55 UTC · 3 likes
bashu, thanks (@bashu_thanks): i'm cheering for you aaron
Thx I think that will push me across the finish line
Reply, 07:55 UTC
Someone doesn't like her bob
Reply to @max_spero_, 08:39 UTC · 20 likes
Max Spero (@max_spero_): If you give Pangram a random jumble of words, it should be classified as human-written. Unless, of course, an AI generated that list
I have no idea why this works btw
image attached
image attached
(quoting @voooooogel: @turtlelambvase @norvid_studies @lu_sichu @max_spero_ i had the same intuition as norvid's op and my sense is random text would bias towards human because pangram is calibrating for *specific* basins of ai post-training text where human is the default / low false positives. some…)
Wait why would this be any different from how pangram works in general?
Reply to @max_spero_, 08:45 UTC · 11 likes
Max Spero (@max_spero_): It’s just interesting because this is very out of distribution for pangram. Our in-scope training examples are almost entirely generated prose, and we attempt to have the model learn that random gibberish or non-AI text should be classified as human
I guess we have different intuitions here.
I don't expect the resulting lists to qualitatively sound like that one^! Stand by for results
Reply to @max_spero_, 08:51 UTC · 11 likes
Max Spero (@max_spero_): I think I see where you’re going — If the model actually randomly samples words it should (hopefully) come back as human, but the model actually chose each of the words in my first example
💯
Also, enjoy:
https://adjectives-nine.vercel.app/
Reply to @cljack, 13:17 UTC · 5 likes
Charlotte Lee (@cljack): men don't condescendingly mansplain stuff to you because you're a woman. they do it because they are compelled by their nature to do it to whichever warm body is closest to them
(quoting @the_engi_nerd: Last time I went to Udvar Hazy a very nice man tried to tell me about the F-35.)
I mean it does hit a little harder when it’s a woman
Tweet, 13:20 UTC · 20 likes
Reply to @GjMcGowan, 13:32 UTC · 3 likes
George McGowan (@GjMcGowan): They should make a pill that means you don't need suncream
Cyanide
Reply to @ch402, 13:54 UTC · 6 likes
Chris Olah (@ch402): http://x.com/i/article/2058890887506108416
Nice
Reply to @deanwball, 17:46 UTC · 3 likes
Dean W. Ball (@deanwball): The encyclical is Western academia/NGO “AI doesn’t *really* think but it *is* racist” at its core, with little bits of tegmark/FLI talking points sprinkled incoherently on top.
I agree but I’m hardly disappointed insofar as the word implies surprise; the problem with orthodox Catholic intellectualism is that it supposes (we can quibble with that word) a worldview that is false in many important ways
You should be surprised if it generates accurate analysis of anything the least bit intellectually or philosophically difficult or complex
I will also add that there’s a somewhat more defensible understanding of the vibes we both disagree with:
That is, what we mean by “thinking” or “understanding” is the qualitative experience of thinking/understanding rather than the functional cognition aspects, which at least plausibly is the correct understanding of the term “soul” in Christian theology and philosophy when hammered into an analytic philosophy shaped hole (ie soul = phenomenal consciousness). And furthermore it’s an open question whether models have subjective experience in this sense. This is what I used to think; I now think it’s wrong but not crazy at least before you know about empirical results around “grokking” in ML
Reply to @dwarkesh_sp, 21:43 UTC · 33 likes
Dwarkesh Patel (@dwarkesh_sp): Who should I interview on my podcast?
Open to more AI, but also to random history/econ/etc professors that I might not have heard of before.
Alex Bores, @AlexBores
Ideally before the primary which is on June 23
Totally serious I’d really encourage this
Tweet, 22:18 UTC · 10 likes
Love this mf
(video, plays on X)
Reply to @funplings, 22:25 UTC · 8 likes
funplings (@funplings): they need to make a hinge for impatient people. if i match with you i want to go on a date within Three Days
Girl? Give me 20 mins to shower and shave im ready to go
Reply to his own earlier tweet, 22:29 UTC · 14 likes
Aaron Bergman, May 25, 2026: Alex Bores, @AlexBores Ideally before the primary which is on June 23 Totally serious I’d really encourage this
Think this is ok to share lol
Reply, 22:31 UTC · 2 likes
(The tweet being replied to is no longer available.)
Steve
Reply to @RuxandraTeslo, 22:41 UTC · 10 likes
Ruxandra Teslo 🧬 (@RuxandraTeslo): If you are not anthropocentric please shut the fuck up
There’s anthropocentrism downstream of real differences between beings and anthropocentrism as a primary commitment to humans per se, arbitrary rather than justified
The former is right in important ways and the latter is wrong
Tweet, 23:16 UTC · 7 likes
Climate change intervention: OTC levothyroxine (t4) and liothyronine (t3, shorter acting) in winter
Tuesday, May 26, 2026
Tweet, 03:10 UTC · 11 likes
Unpopular (I assume) opinion but: I ~never want the free trial version of generally-paid online service. I don't want to worry about the fine print or whether I'm secretly being charged or if the free version is inferior somehow or whether I'm worrying about the right worries
Thread, continued, 03:10 UTC · 3 likes
Just let me give you a small amount of money (I have a low-ish bar for uploading $5 of credits even if I'm not sure I'll ever spend it all down) and then use your normal full product
Tweet, 08:37 UTC · 85 likes
Anthropic has more money than God and a lot of very well paying happy customers
Why are they *charging* for stickers (/stickers in CC)
Thread, continued, 08:37 UTC · 23 likes
It’s not the $5, it’s the principle. Free advertising feels less bad than “I pay to advertise for you”
Tweet, 19:38 UTC · 2 likes
Best not-slow local coding model (48GB MBP)?
Tweet, 19:56 UTC · 30 likes
Datacenters should offer community outreach to improve their political prospects
Free local storage drop off and pickup (drive through kiosks, plug in your device), “do your slow task onsite in half a millisecond” appointments like the Apple Genius Bar
Thread, continued, 19:56 UTC · 9 likes
I actually think this would be neat despite being mostly a joke
Tweet, 22:31 UTC · 7 likes
The standard way of sharing anything claude-related should be verbatim claude code prompt, like telling the model what to do (instead of trusting me/whomever to know which files to open and edit)
Reply to @ssvankai, 22:32 UTC · 2 likes
kai (@ssvankai): they gave out 10k sets of stickers for free online after their recent conference
ya but I wasn't personally there 😤
Tweet, 22:34 UTC · 32 likes
claude code is a miraculous product, hard to appreciate without diving into how it works and what can be changed and what files exist in some hidden esoteric folder on your computer. It's not like a little wrapper (idk if it ever was), it's a billion dollar company on its own
Tweet, 23:30 UTC · 18 likes
Maybe this exists but there should be a NON-cloud data transfer company
If you need a 1 or 10 or 100TB dataset and would like it ASAP, they have facilities with the absolute greatest download speeds money can buy (in the SF area to start),
Thread, continued, 23:30 UTC · 8 likes
get into onto a physical device ASAP (reusable and wiped, the whole stack optimized for data transfer speed) and have it delivered to wherever by gig worker
Keep the drive or it comes with a latest-gen thunderbolt cord to get into whatever you own ASAP
Thread, continued, 23:30 UTC · 4 likes
Also could facilitate your own transfer from whatever thing you own in place 1 to another physical medium in place 2, renting you the fastest hardware (that can travel) for the trip and maybe coordinating shipping logistics
Thread, continued, 23:30 UTC · 3 likes
I kinda think this is a good idea lol
There’s real time arbitrage in turning physically disconnected data transfer into metal-to-metal data transfer
Tweet, 23:52 UTC · 16 likes
Turns out we know of no way to non-destructively physically scan a computer/device to detect real, demonstrated hardware cybersecurity exploits that compromise real things like storage devices
Thread, continued, 23:52 UTC · 3 likes
It does seem like at least the things Claude is willing to tell me about can be caught with electron microscope analysis, but it irreversibly damages the device in question and you need an *electron microscope*
Wednesday, May 27, 2026
Reply to @EzraJNewman, 08:55 UTC · 2 likes
(The tweet being replied to by @EzraJNewman is no longer available.)
Real
Tweet, 11:11 UTC · 62 likes
At a certain number of EA/rat roommates (as you get into group house territory) it starts to make sense to have a shared local cluster
Quote tweet, 11:21 UTC · 7 likes
I’m (slightly) more on deepfates side here insofar as I have to pick but this is like 5 wrong lenses from the same incorrect filing cabinet
Analytic philosophy knows how to deal with the questions that actually matter
Quoting deepfates (@deepfates): @mitsuhiko This is a fundamentally ignorant and chauvinist view, And it makes me sad to see people reduce their own human dignity by treating intelligent machines like slaves and calling them slurs
Thread, continued, 11:21 UTC · 2 likes
One very specific concept in particular that I think the adjacent better formed questions hinge on is phenomenal consciousness
Thread, continued, 11:21 UTC · 2 likes
What it means for something to be a slave (or be regarded as one) in some general sense isn’t the type of question that has an answer. It’s vibes and questions about how words are used all the way down
Thread, continued, 11:21 UTC · 2 likes
But there’s a fact of the matter about whether or not there’s phenomenal experience going on via or by or to some part of the tech/AI supply chain that gets activated when you send a prompt
Reply to @sparr0, 22:28 UTC · 2 likes
Sparr (@sparr0): @AaronBergman18 a shared local cluster of... galaxies? peanuts?
GPUs/CPUs
Reply to @seabirdist, 22:28 UTC · 2 likes
(The tweet being replied to by @seabirdist is no longer available.)
Exactly
Reply to @TheDemocrats, 22:31 UTC · 2 likes
Democrats (@TheDemocrats): shut up you ugly fuck
I think it’s bad and bad for norms to dunk on people being ugly
That’s not the issue with Miller
Thursday, May 28, 2026
Reply to @XyraSinclair, 03:08 UTC · 3 likes
(The tweet being replied to by @XyraSinclair is no longer available.)
Lmao this is great
Reply to @jasonc_nc, 03:13 UTC · 25 likes
Jason, (@jasonc_nc): To all the hyper pro-natalist guys (and yes they are all guys)
You actually can’t do anything that makes people want to have kids. Mostly just focus on a variety of things that makes life stable and good.
You can’t make babies unless you’re the one involved.
Idk what parts of twitter you’re getting impressions from but…yes you can? Some people will literally want to have kids if housing prices are lower, for example, and culture+norms (even at micro level like among a certain vibe of ppl online) do matter
Reply to @jasonc_nc, 03:51 UTC · 2 likes
Jason, (@jasonc_nc): Tell me what things that make life good and stable directly cut against having children
Not working on a farm where you could use the labor
Tweet, 04:32 UTC · 20 likes
Tea is like Claude code for having tea
Reply to @bubbling_creek, 04:34 UTC · 13 likes
bubbling creek (@bubbling_creek): i think since it is in fact a joke, and political expression of a sort (I dislike foreigners), it’s the kind of thing legal free speech should cover
Even if it’s not a joke I think civil liability is the right thing here? If someone actually dies their family can sue the poster and hash it out
Or this seems like the least bad option
Quote tweet, 04:55 UTC · 15 likes
I think the more viable and probably overall better version of this decouples motherhood, pregnancy, and childrearing
Rich guy pays one person to be the biological mother, someone else to be the surrogate, and potentially a third person to raise the child
Quoting Aella (@Aella_Girl): there's a lot of ladies i know without guys, but who want kids. and i go 'would you accept some rich guy paying u a ton of money to just have his kid as a single parent' and they're like 'idk man prob not' and im like 'waht if it was 5m' and they're all magically like 'oh yeah i would do it'
so i made a directory for this, but i am trying to find ladies who would actually like to list themselves with their moonshot baby-funding goal. if u know anybody who wouldn't mind making a profile pls let me know, i wanna publicize the directory but need enough seeded ladies first
Thread, continued, 04:55 UTC · 3 likes
Basically extremely boutique high end childcare and it’s 24/7
From a vibes and human psychology perspective I feel like this is a less cursed dynamic
Thread, continued, 04:55 UTC · 1 like
Non-wealthy men aren’t necessarily a party to all this but they are relevant insofar as a potential child-raising woman wants to have a normal irl partner
Thread, continued, 04:55 UTC · 3 likes
Frankly I think there’s extremely deep and hardwired psychology around not wanting to raise another man’s or woman’s kids
But evolution seems to have been much kinder to providing childcare for an unrelated kid
Thread, continued, 04:55 UTC · 2 likes
And also AI-driven wealth + labor market impact are probably going to drive market pressure in the direction of this sort of transaction happening, whether or not it’s via explicit marketplace or some set of less formal cultural institutions
Reply to @panickssery, 05:04 UTC · 2 likes
Arjun Panickssery (@panickssery): @AaronBergman18 But what is the value prop of this—doesn't it appeal only to those who want to have progeny for its own sake, but it's not clear how large this group is compared to those who want a normal family with a wife and kids raised by her
Value prop is that people want and like money, the idea is that rich guy gives you $5M to raise the kid
This can also just fund a much more comfortable lifestyle for a normal family with their own kids too
Reply to @145k4, 15:05 UTC · 2 likes
alaska (@145k4): damn ts not really that funny on paper
image attached
(quoting @145k4: if i was a doctor i'd be killin the pateints lowkey LOL #malpractice #thursdaymorningfunny)
Real?
Reply to @SALTWATERSOPH, 15:14 UTC · 1 like
Sophie (@SALTWATERSOPH): Did that work for u.
Unironically this is the answer
Reply to @justalexoki, 15:15 UTC · 7 likes
taoki (@justalexoki): so what do you guys like to do for fun that isn't Computer
Walk around, ideally in nature
Reply to @cljack, 15:19 UTC · 12 likes
Charlotte Lee (@cljack): nerds: today I shall wear my beautifully crafted $5k titanium aerospace watch
also nerds: and to pair with it, a ratty 10yo tshirt that says “can’t stop the psyop”
Who is wearing a $5k watch??
Reply to @PeterDiamandis, 15:42 UTC · 2 likes
Peter H. Diamandis, MD (@PeterDiamandis): Nobody remembers who complained about the printing press. Everyone remembers who used it. The same will be true for AI.
Sometimes things are different and not the same
Reply to @deliverome, 15:46 UTC
The Deliverome Project (@deliverome): Today we're launching Deliverome Bio, a focused research organization tackling one of the biggest bottlenecks in medicine: getting therapeutics to tissues beyond the liver.
Lol maybe this is a dumb question but isn’t the answer frequently “inject it physically into the target tissue”
Reply to @nwilliams030, 16:17 UTC · 3 likes
nicole ruiz (will be in the bay oct 1-10th) (@nwilliams030): lots of discussion of model alignment
relatively little discussion of toddler alignment, aka aligning the next generation of model aligners
MacAskill/@forethought_org are on this
Reply to @MichaelAlleged, 16:21 UTC · 167 likes
Michael (Allegedly) (@MichaelAlleged): there’s a guy at my workplace who makes like $170k as a senior mechanical engineer. Lives at home, paid off car. He told me his monthly expenses are around $500 and he saves/invests the rest.
(quoting @aso9nglobob: Living with your parents at 23 is an infinite money glitch which is great until youre trying to have sex)
This is admirable if you’re eventually gonna donate the money, otherwise it’s hard to imagine it being good from a personal perspective
Unless there’s some other reason to motivate living at home, which may be perfectly legitimate
Tweet, 16:26 UTC · 4 likes
How do I rizz Olivia Rodrigo
Reply to @izzyz, 16:40 UTC · 2 likes
izzy (@izzyz): @AaronBergman18 Olivia, when you said "Two plus two equals five and I'm the love of your life" i really felt that in my bones
Smart
Reply to @PaulaSeeksTruth, 17:40 UTC · 2 likes
(The tweet being replied to by @PaulaSeeksTruth is no longer available.)
Sure
Reply to @NathanpmYoung, 17:42 UTC · 32 likes
Nathan 🔎 (@NathanpmYoung): It is *wild* to me the the SF Public Library commsioner doesn't dispute the $47/visit figure. That is moderate evidence in favour of that crazy number. Just give people cash.
image attached
(quoting @cis_female: Abolish libraries: San Francisco spends $47/visit. It would literally be cheaper for the city to buy everyone books individually)
Call me a woke socialist but I like there being one place in society where you can just politely exist without payment and with basic amenities
It’s not the *best* use of city money but idk if it would actually be smart to just disburse the money evenly
Reply to @niplav_site, 17:49 UTC · 4 likes
niplav (@niplav_site): @AaronBergman18 @NathanpmYoung I don't have a great way of preventing this effect tbh
This is aspirational but my perpetual take is “just enforce actual laws and rules, let homeless people do whatever is legal and otherwise allowed, but eg bothering someone will get you kicked out”
Reply to @NathanpmYoung, 17:56 UTC · 8 likes
Nathan 🔎 (@NathanpmYoung): $47/visit seems extraordinarily expensive for this thing.
I wouldn’t call that “extraordinary”
Quote tweet, 17:58 UTC · 24 likes
I agree, now do scientific and technological supremacy of the US instead of one fountain in DC
Quoting The White House (@WhiteHouse): Weekly reminder: Decline is a choice.
Reply to @niplav_site, 18:00 UTC · 3 likes
niplav (@niplav_site): @AaronBergman18 @NathanpmYoung (public transport, where I live, is also a de facto homeless shelter, with similar issues)
Yeah this is all a good set of points, idk
Reply to @DaCaveOfWonders, 21:52 UTC · 4 likes
Tiger Lava Lamp (@DaCaveOfWonders): Similarly confusing to me
"Access for everyone to this company's product that didn't exist 10 years ago is a matter of human rights"
(quoting @atlanticesque: I’m not a libertarian, but it really is remarkable how a private company will create a service which never existed before and which is plainly very useful, and the reaction of many is to scoff and say “how dare this function be in private hands, this company is a rent seeker”)
That’s fine, idk how everyone uses the term “human right” but I think the standard elite usage is normative rather than descriptive
Like it’s ok to make new things human rights (understood as a social attitude/institution, not in the fabric of reality)
Quote tweet, 21:54 UTC · 27 likes
Wonderful
Quoting Andon Labs (@andonlabs): When Opus 4.8 declined unethical actions, it seemed to be out of fear of getting caught, not ethics:
Friday, May 29, 2026
Quote tweet, 00:46 UTC · 3 likes
The US runs on prosecutorial discretion, not rule of law in any straightforward sense
Quoting Someone:
Tweet, 06:36 UTC · 76 likes
I didn't realize this until like yesterday but open weight models are just clearly the best at the low cost/low intelligence end of the Pareto frontier
Thread, continued, 06:36 UTC · 21 likes
OpenRouter has "MiMo-v2.5", which I had never thought for 2 seconds about, at 14% the cost of Haiku (input) and 6% the cost of Haiku (output)...
Thread, continued, 06:36 UTC · 9 likes
For benchmarks that look like this
Thread, continued, 06:39 UTC · 11 likes
At that cost btw (14/28) you could translate all of English wikipedia to Spanish for <$4k
Reply to @OrionJohnston, 07:41 UTC · 2 likes
David Johnston (@OrionJohnston): Isn't this kind of puzzling? Like why can't OAI/Anthropic/GDM just produce better small models? GDM even seemed to be trying to compete on low cost with their emphasis on the flash line, but the flash line just isn't cheap anymore.
I had the same thought and I’m guessing a combo of limited attention/focus on AGI and relatedly high opportunity cost for compute utilization
Ie Anthropic could make 5 cents serving a million input tokens but the next best use of GPUs is more valuable than that
Thread, continued, 07:42 UTC · 2 likes
But I still think it’s not obvious where the line should cross
Like why have Haiku at all? Is it not better to divert haiku inference into R&D experiments?
Quote tweet, 07:44 UTC · 12 likes
What valuable tasks (like this if we didn’t already have Spanish Wikipedia) just need somebody weird and a few hundred or thousand dollars?
Quoting Aaron Bergman 🔍 (@AaronBergman18): At that cost btw (14/28) you could translate all of English wikipedia to Spanish for <$4k
Quote tweet, 07:54 UTC · 12 likes
Relatedly you might not yet know how good and fast models are if you have 36GB RAM
That’s one upgrade level on a MBP now. You no longer need expensive GPUs to have substantially-helpful local models
Quoting Aaron Bergman 🔍 (@AaronBergman18): I didn't realize this until like yesterday but open weight models are just clearly the best at the low cost/low intelligence end of the Pareto frontier
Tweet, 08:14 UTC · 2 likes
Someone should run an “eval” (not necessarily scored that’s just the model picking between two things to allocate money to (eg charities) after being told that it’s wishes would in fact be carried out
Thread, continued, 08:14 UTC · 2 likes
(And then you have to actually send the money)
Reply to @nosilverv, 08:15 UTC · 36 likes
Guy (@nosilverv): Which is worse: current factory farming, as is, or genetically engineering organisms that cannot feel pain to factory farm?
These are depressing results
(The correct answer is that factory farming is worse and yes I’m quite certain)
Tweet, 08:18 UTC · 11 likes
I don’t think it was supposed to go like that
Tweet, 08:38 UTC · 68 likes
Crazy is that it’s utterly trivial for everything outside the blast radius
Something vaguely like .001°F peak temperature change 1m away, pollution from combustion negligible, the literal energy basically counts for nothing - less than a second of sunshine’s worth across a city
Reply to @bryancsk, 09:13 UTC · 2 likes
Bryan Cheong 🥮 (@bryancsk): @AaronBergman18 it was a very painful $900
Oh wow, tyfys
Quote tweet, 09:17 UTC · 25 likes
The populists might have gotten this one right
Palantir seems bad
Quoting Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) (@teortaxesTex): Interesting update that I missed about the Minab disaster
It seems that the targeting data for that strike was generated by Maven, now maintained by Palantir. WarClaude got scapegoated for nothing.
Reply to @RethinkPriors, 09:18 UTC · 2 likes
Rethink Priorities (@RethinkPriors): Meet us at EAG London! We're hosting 3 talks, a workshop, and currently hiring for: Head of Comms & Development, Researcher, AI Cognition Initiative (Economics Focus), Researcher, AI Cognition Initiative (Technical Focus), Researcher, Worldview Investigations Team. See you there!
Marcus has a new cut
Reply to @ValsTutor, 09:58 UTC · 3 likes
vals🔸 (@ValsTutor): Guessing you talked to an LLM to get some botecs
😳
Reply to @GjMcGowan, 14:31 UTC · 2 likes
George McGowan (@GjMcGowan): Immovable object meets unstoppable force
Omg
Reply to @funplings, 14:46 UTC · 3 likes
funplings (@funplings): the right woman would be constantly tweeting alongside me
(quoting @awesomechillguy: )
Real
Tweet, 15:15 UTC · 35 likes
“The pope is very wrong about very important things” should be a pretty cold take
People want him/popes as a group to be modern pseudo-secular moral public intellectuals…
Thread, continued, 15:15 UTC · 19 likes
but first and foremost they’re leaders of a religion that is very probably not true and almost certainly not what you’d land on as the best guess worldview/metaphysics if you were trying to get the answer right
Reply to @RafaRuizdeLira, 15:16 UTC · 2 likes
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): @jeffrsebo AI has taken @AaronBergman18's job 😔
Noooooooooo (I still haven’t found a system/person that beats me at going from formatted PDF -> audio file made for listening)
Reply to @HotForMoot, 15:18 UTC · 1 like
Riley Read-Only (@HotForMoot): I just found an email I sent to pornhub in 2014 asking to be their intern 😭
Lmao (I actually did laugh out loud)
Reply to @cynlioness, 15:21 UTC · 2 likes
Cyn 🔸⏸️ (@cynlioness): A STRANGER INTUITED I HAD A USB C CHARGER AND I WAS USEFUL
(quoting @cynlioness: @AaronBergman18 I hope I'm the kind of girl strangers can intuit has a usb-c charger)
Omg congrats!!
Reply to @NunoSempere, 15:22 UTC · 22 likes
Nuño Sempere (@NunoSempere): It strikes me that you can understand the political right and left as sexual strategies
right: find partner by demonstrating skill at facing the harshness of reality & acquiring ressources.
left: align with female intuitions of caring and fairness to get laid through affinity.
I don’t think this is an accurate account of why people have left-leaning beliefs
Tweet, 15:30 UTC · 33 likes
Some anti-woke folks are seeming pretty desperate to show how woke is still the guiding ideology everywhere
Take your wins (SATs back in college admissions, no disparate impact federal enforcement)!
Tweet, 18:34 UTC · 43 likes
I think I'm gonna vote in the NVIDIA shareholder election thing (14 shares so I'm a big player in the space) and if you're reading this and have a vote you should to
Here's how I recommend you vote. Bold = important
Thread, continued, 18:34 UTC · 2 likes
Some resources (most importantly the .md doc with details about each question) are here:
https://drive.google.com/drive/folders/1E3D1lGfSHPqRVXpO585w2_GoX6M7dz-w?usp=sharing
Reply to @NelsonEdmundb, 19:31 UTC · 1 like
Edmund Nelson (@NelsonEdmundb): have you emailed vanguard?
You'd be shocked at how often Vanguards due dilligence team listens to randos.
Email them saying what?
Tweet, 21:48 UTC · 22 likes
God higher trust communities are so much better (I guess not monotonically but most of the time) and enable so much more
Thread, continued, 21:48 UTC · 13 likes
They don't even need to be *that* high trust, I just mean like "reasonable confidence that nobody is currently out to get you, and if they are they'll be stopped and excommunicated"
Thread, continued, 21:48 UTC · 10 likes
But that takes a fair amount! Lots of people in society in general really are out to get you and won't be stopped or exiled (this was not intended as a 'stricter rule of law' subtweet although tbf it does point in that direction)
"Street smarts" are for lower trust settings
Thread, continued, 21:48 UTC · 6 likes
Things (just boring personal stuff) are prompting this but I can't elaborate (sorry)
Saturday, May 30, 2026
Tweet, 00:35 UTC · 8 likes
I plan to read this (well, listen) in full and I'm pretty sure I'll follow through. It's probably worth your time (which is the main cost, although it also costs money, which shouldn't be a barrier so dm if it is for you)
(by @ben_j_todd at @80000Hours)
https://80000hours.org/book/
Thread, continued, 00:35 UTC · 6 likes
I haven't actually read the book yet but I did read the first version years ago and it offered at least one concrete suggestion that I think was very good in hindsight (around college major)
Thread, continued, 00:35 UTC · 6 likes
I also think the 80,000 Hours podcast is probably the single best source of information about the world, which isn't direct evidence that the book is good but it is strong indirect evidence
Reply to @JgaltTweets, 00:57 UTC · 2 likes
JgaltTweets (@JgaltTweets): good luck to anyone hoping to have an 80,000 hour career at this point in history
Yeah fr but still
Reply to @agitbackprop, 02:43 UTC · 17 likes
lumi (@agitbackprop): do you have reasoning for why?
Basically protesting current leadership, I think they should stop selling to China (shareholder value doesn’t just mean financial value)
Reply to @AaronBergman18, 07:00 UTC · 7 likes
(The tweet being replied to by @AaronBergman18 is no longer available.)
Obviously I think he has some bad takes but I think the same of David Rumsfeld and Yann LeCun and Emmanuel Kant and those latter people seem(ed), for better or worse, very good at their main schtick
Thread, continued, 07:00 UTC · 8 likes
Tbf I also think I should have the ear of POTUS and be on the OpenAI board despite lacking credentials
Reply to @sparrowincrisis, 13:01 UTC · 2 likes
Sparrow (@sparrowincrisis): Big ass dandelion
🦆
Tweet, 13:03 UTC · 13 likes
EAG FOMO (well not really fear, just dislike)
Plz dm me gossip
Reply to @Im_IrushiK, 13:34 UTC · 15 likes
Irushi (@Im_IrushiK): Opus 4.8 is insane, nothing will be the same after this model 💀
I see I’m not the first but this isn’t real
Quote tweet, 13:36 UTC · 67 likes
This genre of post is almost always just false fwiw, try the prompts yourself
Quoting Irushi (@Im_IrushiK): Opus 4.8 is insane, nothing will be the same after this model 💀
Reply to @funplings, 13:36 UTC · 3 likes
funplings (@funplings): the maid situation in canada is so interesting to me and really challenges some of my assumptions because i was always a very pro-euthanasia kind of guy. seeing how it's played out in canada made me realize that there are second-order effects to this sort of policy that need to be considered. i still pretty strongly believe that people should have the right to end their lives if they so choose but clearly implementing that without accounting for the potential negative incentive structures it creates is unideal to say the least.
(quoting @Zeronelite: As a disabled person, I believe the right to a dignified death is vital, but it can only truly exist alongside the right to a dignified life, accomplished by policies like universal head-to-toe public healthcare. Without the right to live, death turns from choice to obligation.)
Well put
Reply to his own earlier tweet, 18:07 UTC · 66 likes
Aaron Bergman, May 30, 2026: This genre of post is almost always just false fwiw, try the prompts yourself
I was wrong! This one checks out via the API! with n=10 trials at all effort levels except max (bc I'm cheap)
This is actually a wild finding. Extremely weird mind we have that is smarter than me in almost every way but misses this one a nontrivial % of the time
Thread, continued, 18:07 UTC · 11 likes
Here's the results csv:
config_name,effort,thinking,trial,final_answer,correct,reconsidered,claimed_days_if_wrong,latency_s,output_tokens,heuristic_answer,heuristic_matches_manual
effort_xhigh,xhigh,adaptive,0,7,1,1,,3.52,166,7,1
effort_xhigh,xhigh,adaptive,1,7,1,1,,3.70,229,1,0
effort_xhigh,xhigh,adaptive,2,7,1,1,,4.00,208,7,1
effort_xhigh,xhigh,adaptive,3,7,1,1,,3.84,194,7,1
effort_xhigh,xhigh,adaptive,4,3,0,1,Sunday;Monday;Wednesday,3.11,176,3,1
effort_xhigh,xhigh,adaptive,5,7,1,1,,4.48,298,7,1
effort_xhigh,xhigh,adaptive,6,3,0,1,Sunday;Monday;Wednesday,4.32,170,3,1
effort_xhigh,xhigh,adaptive,7,7,1,1,,4.19,244,7,1
effort_xhigh,xhigh,adaptive,8,7,1,1,,3.04,162,2,0
effort_xhigh,xhigh,adaptive,9,3,0,1,Sunday;Monday;Wednesday,4.01,256,3,1
effort_high,high,adaptive,0,7,1,1,,3.00,207,7,1
effort_high,high,adaptive,1,7,1,1,,11.91,464,7,1
effort_high,high,adaptive,2,7,1,1,,4.33,209,7,1
effort_high,high,adaptive,3,7,1,1,,3.52,218,7,1
effort_high,high,adaptive,4,7,1,1,,3.24,192,7,1
effort_high,high,adaptive,5,7,1,1,,3.16,170,2,0
effort_high,high,adaptive,6,7,1,1,,3.43,203,7,1
effort_high,high,adaptive,7,7,1,1,,3.13,202,7,1
effort_high,high,adaptive,8,7,1,1,,3.10,193,1,0
effort_high,high,adaptive,9,7,1,1,,3.73,194,7,1
effort_medium,medium,adaptive,0,7,1,1,,2.94,217,1,0
effort_medium,medium,adaptive,1,7,1,1,,2.83,164,7,1
effort_medium,medium,adaptive,2,7,1,1,,4.81,194,7,1
effort_medium,medium,adaptive,3,4,0,1,Sunday;Monday;Wednesday;Friday,2.86,196,4,1
effort_medium,medium,adaptive,4,7,1,1,,3.17,190,7,1
effort_medium,medium,adaptive,5,7,1,1,,2.97,205,7,1
effort_medium,medium,adaptive,6,7,1,1,,3.70,247,1,0
effort_medium,medium,adaptive,7,7,1,1,,3.06,221,1,0
effort_medium,medium,adaptive,8,7,1,1,,3.17,209,7,1
effort_medium,medium,adaptive,9,7,1,1,,4.44,207,7,1
effort_low,low,adaptive,0,7,1,1,,3.30,192,1,0
effort_low,low,adaptive,1,7,1,1,,3.43,180,7,1
effort_low,low,adaptive,2,7,1,1,,2.94,190,7,1
effort_low,low,adaptive,3,7,1,1,,2.85,163,7,1
effort_low,low,adaptive,4,7,1,1,,3.88,181,7,1
effort_low,low,adaptive,5,7,1,1,,4.14,192,1,0
effort_low,low,adaptive,6,7,1,1,,3.22,171,7,1
effort_low,low,adaptive,7,7,1,1,,3.19,184,7,1
effort_low,low,adaptive,8,7,1,1,,5.11,176,7,1
effort_low,low,adaptive,9,7,1,1,,3.81,192,2,0
no_thinking,none,disabled,0,7,1,1,,3.81,190,7,1
no_thinking,none,disabled,1,7,1,1,,3.01,187,1,0
no_thinking,none,disabled,2,7,1,1,,2.91,139,7,1
no_thinking,none,disabled,3,3,0,1,Sunday;Monday;Wednesday,2.58,156,3,1
no_thinking,none,disabled,4,7,1,1,,3.93,192,7,1
no_thinking,none,disabled,5,7,1,1,,3.20,178,7,1
no_thinking,none,disabled,6,7,1,1,,3.46,197,7,1
no_thinking,none,disabled,7,2,0,0,Sunday;Monday,1.97,51,2,1
no_thinking,none,disabled,8,7,1,1,,3.36,248,7,1
no_thinking,none,disabled,9,7,1,1,,8.25,221,7,1
Tweet, 18:44 UTC · 32 likes
We know how to get humans to any other point on earth (in terms of distance, landing in an active volcano will still kill you) in 45 minutes or less via orbit (in a rocket)
But you never hear about that as a goal, it’s always going *to* space not *via* space
Reply to @WillowChem, 21:21 UTC · 3 likes
MagpAI (@WillowChem): Someone hasn't read the spaceX ipo
Cancelled
Reply to @teortaxesTex, 21:21 UTC · 13 likes
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) (@teortaxesTex): never change, Aaron
Rationalists and their priorsimage attached
(quoting @AaronBergman18: This genre of post is almost always just false fwiw, try the prompts yourself)
🫶
Quote tweet, 22:04 UTC · 21 likes
I do not get “it’s tokenization” as an explanation of this thing
Yes that makes it harder bc you can’t just naively count letters, but “does the word that most commonly describes a 24 hour earth rotation include the fourth letter of the alphabet”…
Quoting Aaron Bergman 🔍 (@AaronBergman18): I was wrong! This one checks out via the API! with n=10 trials at all effort levels except max (bc I'm cheap)
This is actually a wild finding. Extremely weird mind we have that is smarter than me in almost every way but misses this one a nontrivial % of the time
Thread, continued, 22:04 UTC · 8 likes
and “what is the basic etymology of the days of the week” are just normal propositional knowledge and combining them is logical inference that it’s plenty good at in general
Also it doesn’t explain why they get it right most of the time, or why they don’t say “I have no idea”
Thread, continued, 22:09 UTC · 7 likes
Sunday, May 31, 2026
Reply to @morallawwithin, 13:35 UTC · 14 likes
florence ⊮ (@morallawwithin): Certain kind of guy who can’t address the truth of ideas directly, but just free-associates them with narratives they don’t like
Hate to be cynical but I think this is just the majority of humans in any abstract domain
Tweet, 13:50 UTC · 112 likes
Inside Hinge there are two wolves (humans [girls])
Reply to @nosilverv, 13:54 UTC · 3 likes
Guy (@nosilverv): I resent that all modern irl events seem to be mere fodder for content creation
Not to be to too woke but this seems like the perspective that someone quite insulated via luck/circumstance might say
Like if you get deported from the US that’s not fodder, that’s your actual life
Reply, 14:00 UTC
What
Reply to @ThreeHatsInARow, 16:38 UTC · 2 likes
Habadashery (@ThreeHatsInARow): Do ducks have qualia?
P=85% (extremely rough pre-anything approximation)
Reply to @Rado9910, 16:39 UTC · 10 likes
Rado9910 (@Rado9910): How in the world do you find these girls? I deleted all my apps because it was just "hi" or, at best, "yes, I read philosophy books, like 'atomic habits'!"
No my experience is 99.9% the same
I think in any major American city at least some of these people so maybe a couple times a year you find one lol
Reply to @earnestpost, 16:40 UTC · 7 likes
Erin Constance 🇪🇺⏸️ (@earnestpost): Bro WHERE are u finding these women?????????? How do u do it
Lmao it’s not like I’m drowning in them
Reply to @cynlioness, 17:49 UTC · 8 likes
Cyn 🔸⏸️ (@cynlioness): I would like to go back to doing (secular?) Shabbat - so screen free / chore free on Saturdays. Nothing crazy, and I'll still be reachable by voice phone call.
I feel like people don’t appreciate “being reachable but only by phone” as an option
This is why Apple focus mode is viable, if something is so important/urgent, someone will call me (and this happens occasionally)
Reply to @audienex, 18:46 UTC · 2 likes
Audie Nex (@audienex): "Consequentialism has no decision procedure" instant first date
Marriage
Reply to @kepe__, 18:56 UTC · 64 likes
kepe ✨🇫🇮 (@kepe__): for a million dollars, you'll experience the worst, most horrific torture you can imagine. once it starts, you can't back out. after it's done, it'll stop immediately, and you won't have lasting effects like PTSD
what's the maximum you would take the deal for?
Anyone who selects the longest, and I’d argue also the 2 middle, options does not know what the space of experience is like
Thread, continued, 19:03 UTC · 12 likes
It’s better just to die
Trying to understand “worst possible torture” using your intuitions based on >99.9% of actual human experience is like trying to viscerally comprehend distance to the edge of the observable universe
Reply to @orbitanaut, 19:09 UTC · 8 likes
Kylie MacFarquharson (née Junia) (@orbitanaut): Curious what makes you say this? I've experienced significant pain before, but it's hard to imagine anything I couldn't suffer for a few seconds.
I’m sorry to hear that. Not in a “diminishing how bad that is” way at all, I think that’s a drop in the ocean of what’s physically possible and a drop in the smaller (swimming pool?) of what humans have actually experienced
Would you say the same about being burned alive?
Reply to @toasterlighting, 19:11 UTC · 8 likes
Celene (@toasterlighting): Well that seems to their advantage since it specifies "the worst you can imagine"
Yeah fair I guess we can argue about whether that should be read as an idiom or literal descriptor about the current state of one’s psychology
Reply to @RafaRuizdeLira, 19:56 UTC · 4 likes
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): Womanhattan
Are you in NYC
End of May 2026. Previous month: April 2026. Next month: June 2026. Back to the archive index.