Aaron Bergman's tweets, July 2025
Every public tweet, reply, quote tweet and retweet that Aaron Bergman (@AaronBergman18 on X, formerly Twitter) posted in July 2025 and has not deleted: 316 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: June 2025. Next month: August 2025.
Tuesday, July 1, 2025
Reply to @austinc3301, 04:21 UTC · 4 likes
Agus 🔸 (@austinc3301): not me sending admonymous links after a date not working out
Lmao same
Reply, 15:44 UTC · 1 like
I mean this isn’t what you’re asking but there are way better charities for entirely unrelated reasons (EA Animal welfare fund and LTFF)
Reply to his own earlier tweet, 17:48 UTC · 2 likes
Aaron Bergman, June 30, 2025: I hate AWS
Reply to @arthur_wright_, 20:15 UTC · 3 likes
Arthur Wrong 🔸 (@arthur_wright_): @WillowChem Also, somebody should definitely make one. I'll see if @SpacedOutMatt has any deck-printing insights from his experience with the Derek Parfit game
Seconded
Reply to @ohabryka, 21:21 UTC · 10 likes
Oliver Habryka (@ohabryka): Really, all things considered? I don't believe you.
Like, you actually think that if someone was advocating for substantial changes in e.g. animal welfare laws and industrial policy, on the basis of their beliefs about these moral weights, which straightforwardly assume that 100 bees are more important than a human, you wouldn't consider them an idiot?
Like, look man, yes, there are arguments here that raise these kinds of things to plausibility, but clearly there are no arguments here that raise this kind of thing to the level where you actually start just acting as if it's true. It violated so so so many other moral heuristics.
You don't have that much caring to spread around, if you put this much of your caring into insects, you don't really have much space to care about anything else.
The last paragraph points towards some degree of moderation in actions at the end of the day, but it’s not actual evidence against bee experience being >1% as important as humans’
It’s like people who reject utilitarianism because it’s too demanding - better to imperfectly pursue the actual good than reason backwards from what seems too demanding
Thread, continued, 21:23 UTC · 7 likes
Your reaction to this seems to me a bit like how I’d guess you’d react to someone who says “if x risk is so high, why aren’t you marching in the streets, working 20 hour days, etc?”
Like the actually good response to high expected amounts of insect suffering probably doesn’t look any crazier than Shrimp Welfare Project
Wednesday, July 2, 2025
Reply to @ohabryka, 00:55 UTC · 2 likes
Oliver Habryka (@ohabryka): I think it's more similar to my response to people who don't buy ChatGPT because they think AI will kill everyone and so they don't want to support OpenAI. I do think it genuinely is an inconsistent state of belief and quite dumb.
In the analogy, OpenAI ≈ honey?
If so, I’m actually kinda sympathetic to this line of thought re veganism in general bc of nutrition -> health/productivity concerns
But this is just a cost vs benefit consideration, it doesn’t say anything about the likelihood of bees having significant moral status
Tweet, 05:20 UTC · 18 likes
I think the RP moral weights project doesn’t give due probabilistic weight to neuron count-like metrics, but it’s at least like 1% likely that the intensity of valenced experience has nothing to do with biomass
The universe makes no promises to abide by our intuitions
Thread, continued, 05:24 UTC · 6 likes
At some level the question of whether it’s plausible that a single digit number of bees are as morally valuable, minute for minute, as a human comes down to priors on how weird the world is
I think the world is extremely weird!
Reply to @justjoshinyou13, 13:11 UTC · 4 likes
Josh You (@justjoshinyou13): like the total value of all beings is not a fixed thing
but you have to worry about reasoning that almost guarantees you care more about huge numbers of small things
I will solve the two envelopes problem today, watch this space
Tweet, 13:14 UTC · 5 likes
> family friend runs multimillion dollar tech consulting business
> pays dozens of employees tech salaries
> talks about how helpful ChatGPT is
> free version, 4o
Reply to @inerati, 13:15 UTC · 2 likes
liz (@inerati): guys it’s been 8 minutes i think my red teaming efforts has been forwarded to the real chris property manager 😔
(quoting @inerati: guys i sent this and then it stopped responding)
Lmao
Tweet, 13:21 UTC · 63 likes
If you’re against DOGE cuts, you should probably be negatively polarized into thinking we were majorly underspending on a lot of things
What are the odds that we had just the right level of USAID and science funding and such?
Reply to @eigenrobot, 13:24 UTC · 5 likes
eigenrobot (@eigenrobot): occasional reminder to men that one of the easiest ways to improve your life is to pay $50 for a nice double-edged safety razor and a stack of feather blades
yes you'll bleed from your face for a week. but after that you'll live above from other men
image attached
(quoting @DanielW_Kiwi: A couple of years back @eigenrobot was saying how much better safety razors were than normal plastic handled razors. Of all things I've learned from here that's been one of my favorite.)
I bought one of these and hate it
Quote tweet, 16:05 UTC · 2 likes
Ok I’ve solved it. The answer is that:
1) try pinning to human for your use use case (this part isn’t original)
…
Quoting Aaron Bergman 🔍 (@AaronBergman18): @justjoshinyou13 I will solve the two envelopes problem today, watch this space
Thread, continued, 16:05 UTC · 1 like
2) if there is remaining ambiguity (eg arguing with an alien over who to save) you can and should update on whether your species is on the “high” or “low” side based on the intensity of your own valenced experience
Tweet, 16:52 UTC · 37 likes
Magnet school stuff might be the culture war issue I can most personally riled up about
Those were the best years of my life, how dare you take it away from other kids in order to fight a culture war the kids never opted in to?
Thread, continued, 16:52 UTC · 7 likes
It’s also bad for other, more mundane reasons
Quote tweet, 16:59 UTC · 4 likes
This was kinda fun, does anyone wanna pay me to fine-tune an LLM on whatever you’ve written (or whatever you want tbh)
Quoting EA Lifestyles (@EAheadlines): the AIs love me
Tweet, 17:08 UTC · 74 likes
Reveal urself 👀
Reply to @GjMcGowan, 17:30 UTC · 15 likes
George McGowan (@GjMcGowan): @AaronBergman18 Aaron I’m sorry it’s gonna be a guy
Cheems mindset
Tweet, 17:58 UTC · 14 likes
So this is why the founders are billionaires huh (I tried Claude code, this is one day.)
Reply to @JgaltTweets, 18:20 UTC · 20 likes
JgaltTweets (@JgaltTweets): This is phase 1 in a sequence of events that ends with you waking up in a motel bathroom missing a kidney, and not in an effective altruism way, but in a "hey where did my kidney go?" way
Whatever, I'll take my chances
Reply to @CharlesD353, 21:46 UTC · 2 likes
Charles🔸 (@CharlesD353): This seems quite silly. It assumes only two buttons - cut, and increase spending, and that the method by which you did so and which areas you aim for do not matter.
The existence of bad cuts just does imply that there are good opportunities on the margin
I agree method/manner matters
Tweet, 21:47 UTC · 9 likes
Anybody need they data scraped
Thread, continued, 22:43 UTC · 3 likes
This is a serious inquiry btw
Thursday, July 3, 2025
Reply to @livgorton, 00:54 UTC · 2 likes
Liv (@livgorton): aaron has my endorsement as a good guy 👍
🥺omg
Reply to @GarrettPetersen, 10:46 UTC · 5 likes
Dr. Dad, PhD 🔄🔼◀️🔽▶️ (@GarrettPetersen): 1.9yo: "Paddleboarding."
Mommy: "Did you hold on tight?"
1.9yo: "I hold on. I don't fall in the water. There fishies in the water. And duck. I catch the duck. In my hand."
Mommy: "What would you do with the duck?"
1.9yo: "I have duck. In hand."
Based
Quote tweet, 14:02 UTC · 12 likes
HMU if ur a little fucked up
Quoting Gordo MONK: Champion of Peace ☮️ and Calm😶🌫️ (@fordoglunk): I do wish date-my-friend posts would advertise their lesser friends, sometimes. I'm not a good man and I dont want to date your flawless overachieving well-adjust friends. Bring me your weird nerd chudettes
Quote tweet, 14:08 UTC · 135 likes
Yes, consciousness is necessary (and possibly sufficient) for moral significance
I will die on this hill
Quoting Henry Shevlin (@dioscuri): Eliezer raises an interesting example here. Even if LLMs aren't conscious, as they develop more agency and personality, is it crazy to think they may deserve some respect for their autonomy and individual goals? Not clear autonomy requires consciousness for moral significance.
Reply to @aidan_mclau, 16:54 UTC · 16 likes
Aidan McLaughlin (@aidan_mclau): how do you define consciousness
Genuine subjective experience
Reply to @dioscuri, 16:56 UTC · 7 likes
Henry Shevlin (@dioscuri): See also Simon Goldstein and Cameron Domenico Kirk-Giannini's paper on AI wellbeing without consciousness - https://philpapers.org/archive/GOLAWE-4.pdf
Lol knowledge and desire are not intrinsically good/morally important
Reply to @Aron_Adler, 16:56 UTC · 6 likes
aron (@Aron_Adler): I think so many people are guilty of getting the causality the wrong way around here. It's not that we care about which beings are conscious and to those we extend moral patiency. It's that we assign consciousness to those entities we empathise with
(quoting @AaronBergman18: Yes, consciousness is necessary (and possibly sufficient) for moral significance I will die on this hill)
I disagree
Tweet, 16:57 UTC · 4 likes
So does BBB mean I’m going to be super duper uninsured when I turn 26 instead of just normal uninsured
Reply to @aidan_mclau, 16:57 UTC · 19 likes
Aidan McLaughlin (@aidan_mclau): how do i test if one has subjective experience
You can’t lol, this is a methodological issue, not a conceptual problem for the definition
Reply to @aidan_mclau, 17:00 UTC · 16 likes
Aidan McLaughlin (@aidan_mclau): wait, so your sole criterion for moral value is unfalsifiable? how is that useful whatsoever?
Well, you can still accumulate less-than-rock-solid evidence in one direction or another
It’s much more likely that I’m sentient than that a rock is
Reply to @aidan_mclau, 17:04 UTC · 13 likes
Aidan McLaughlin (@aidan_mclau): no you can't lol
for certanity to exist on a spectrum, there has to be clear resolution criteria. how would i resolve the market of whether you were more sentient than a rock?
Lol what? That’s just not true, the implementation minutia of prediction markets is not 1:1 with bayesian philosophical updating
You can’t resolve the market, but you do have to act under uncertainty. There’s no option to opt out of the question
Reply to @arithmoquine, 17:28 UTC · 1 like
henry (@arithmoquine): something doesn't have to be prediction market worthy to be reasonable to think about. it's perfectly reasonable to ask "will I be happy if I do X" without resolving what happiness is
💯
Reply to @aidan_mclau, 17:32 UTC
Aidan McLaughlin (@aidan_mclau): sure but that's totally falsfiable tho
if i don't say i'm happy, the market closesthat's different that defining consciousness in a way that can *never* be falsified
You can have (extremely imperfect) implementations of this like “let’s pay a panel of 10 philosophers to spend the next year studying the problem and see what they think at the end”
Tweet, 17:34 UTC · 10 likes
Autistic summer is just an endless orgy (negative) of sunscreen, sweat, and showers
Reply to @GjMcGowan, 17:42 UTC · 2 likes
George McGowan (@GjMcGowan): @AaronBergman18 How much is insurance
Idk like a lot
Tweet, 20:02 UTC · 6 likes
The male urge to create ur own language model from scratch
Reply to @jjtwofive, 20:29 UTC · 1 like
JJ (@jjtwofive): are asleep people morally significant?
Only because they’ll eventually wake up
Reply to @jjtwofive, 20:34 UTC · 2 likes
JJ (@jjtwofive): @AaronBergman18 so is it then the potential for consciousness that is necessary?
Sure, I think this is just a semantic thing
Reply, 21:22 UTC · 1 like
Then it’s over lol (I think this is like 3% likely)
Friday, July 4, 2025
Reply to @medievalpard, 00:37 UTC · 1 like
(The tweet being replied to by @medievalpard is no longer available.)
Many such cases
Reply to @tszzl, 03:11 UTC · 9 likes
roon (@tszzl): the primary nice thing about getting older is that you’re better at everything. raw g may be entering slow decline but you’re more socially competent, better at your work. underrated
I will never be socially competent
Tweet, 18:03 UTC · 27 likes
Kleinian abundance but about girls
Thread, continued, 18:03 UTC · 3 likes
I hope this doesn’t get me cancelled
Reply to @nominalthoughts, 18:07 UTC · 5 likes
Jason Harrison (@nominalthoughts): Pretty funny that one of my long lost half brothers found me through twitter 💀
Wait fr?
Quote tweet, 18:09 UTC · 33 likes
If adrafinil is unregulated, lisdexamphetamine should be too
It’s only fair
Quoting sebastian (@modaminister): happy 4th of July
pre orders are now live
shipping out later this summer
Reply to @nominalthoughts, 18:10 UTC · 4 likes
Jason Harrison (@nominalthoughts): @AaronBergman18 Yeah lol
Damn, that’s cool!
Reply to @finmoorhouse, 18:12 UTC · 3 likes
Fin Moorhouse (@finmoorhouse): If you are 25 years old, you have less than 15 years before turning 40…
Hush
Reply to @austinc3301, 18:25 UTC · 4 likes
Agus 🔸 (@austinc3301): Adrafinil is not, in fact, unregulated
Hush
Reply to @polisisti, 18:25 UTC · 2 likes
Alessandro (@polisisti): Aaron right-wing fertility poasters
. 🤝
"make more girls"
Yay
Reply to @Michael_Druggan, 18:32 UTC · 2 likes
Michael Druggan (@Michael_Druggan): Lisdexamphetamine has significantly higher euphoric potential which correlates with abuse potential.
I support legalizing both but I don't think the implication is as clear as you're making it.
Yeah sadly I didn’t really make a legit argument
Reply to @Adellama_, 18:34 UTC · 2 likes
Ada🔸 (@Adellama_): @AaronBergman18 i posted about this once
Based
Tweet, 21:04 UTC · 51 likes
Petition to just make the start time of something when you should actually arrive
Thread, continued, 21:04 UTC · 8 likes
I don’t have the 12 dimensional theory of mind for this, man
Reply to @RichardHanania, 21:05 UTC · 20 likes
Richard Hanania (@RichardHanania): It bothers me how unbothered liberals seem by food stamps. Just humiliating for all involved, especially given how large our poor people are. Cash would be preferable as it in theory could go to other things and doesn’t have the degrading aspect of gov literally feeding you.
Yes, make it cash! I think republicans are the binding constraint on this, not liberals
Reply to @jeffrsebo, 21:06 UTC · 4 likes
Jeff Sebo (@jeffrsebo): rob, is your impression that illusionists in the community generally favor (a) clear alternatives to sentientism (e.g. "agency is what matters") or (b) sentientism by another name (e.g. "feeling pleasure and pain is what matters, but it has nothing to do with ~consciousness~")?
B drives me nuts lmao
Reply to @NoahTopper, 21:08 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): I simply show up at the start time regardless
Based?
Reply to @kyleichan, 21:08 UTC · 2 likes
Kyle Chan (@kyleichan): Meta’s former director of global energy talks about how batteries are more reliable than diesel generators as backup for data centers.
But because of the way data centers are built today, adding batteries as backup requires a substantial overhaul in build.https://open.substack.com/pub/davidroberts/p/can-data-centers-be-good-grid-citizens
Bro looks like a rectangle
Tweet, 21:11 UTC · 7 likes
Time for my ~annual alcohol purchase! Anybody got spicy vodka takes
Reply to @livgorton, 23:12 UTC · 3 likes
Liv (@livgorton): shoutout to the people who just turn up on time 🙌🙌🙌 the first hour of a party when the noise level tolerable is often the best hour anyway (the last hour can also be great for this reason)
🦆🦐🇺🇸🐓
Saturday, July 5, 2025
Tweet, 03:25 UTC · 8 likes
Girls love rooftops. I should get a rooftop
Reply to @NoahTopper, 04:17 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): original or remake?
Original
Reply to @NoahTopper, 04:22 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): I agree actually, I considered doing more takes to try to capture your essence better but that seemed insane
You can’t fake the autistic anti-eloquent rizz
Reply to @NoahTopper, 04:25 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): Aaron I can in fact
(quoting @NoahTopper: @AaronBergman18)
Noooooooooooooooo
Tweet, 04:35 UTC · 13 likes
I want the Columbia admissions data just bc I like data
Monday, July 7, 2025
Reply to @notanastronomer, 16:18 UTC · 1 like
Lauren Gilbert (@notanastronomer): there’s a stack of trophies and blue ribbons somewhere in my parents’ house that I won at rodeos
(quoting @frozenaesthetic: Share a piece of lore about yourself)
Waow
Tweet, 16:37 UTC · 12 likes
I wonder if GLP-1 agonists and amphetamines stack for obesity
Amphetamines have always been under-prescribed for weight loss
Tweet, 16:48 UTC · 1 like
https://claude.ai/public/artifacts/1a71f94c-4eea-4be6-9753-e4266193f08c
Reply to @Squee451, 16:49 UTC · 1 like
Adam Sherman 🔸 (@Squee451): Amphetamines are insanely addicting and have very noticeable and problematic side effects, they were prescribed heavily in the 1960's for weight loss with predictible outcomes and I don't think this take has been researched whatsoever.
Obesity is also really bad for u
Quote tweet, 16:51 UTC · 29 likes
I don’t even think this sounds smart lmao
Quoting bernoulli_defect (@BernoulliDefect): The “AC heats up some outside air to cool interiors so it’s counterproductive” line is a classic case of sounds smart but is dumb. Lots of this in UK politics
Thread, continued, 16:51 UTC · 7 likes
Like yes AC does not destroy thermal energy from the universe, big if true
Reply to his own earlier tweet, 16:54 UTC · 2 likes
Aaron Bergman, July 7, 2025: https://claude.ai/public/artifacts/1a71f94c-4eea-4be6-9753-e4266193f08c
Ok maybe not “always”
Quote tweet, 18:39 UTC · 3 likes
OG GI medicine too (pepto bismol but it was a thing before that)
Quoting St. Rev. Dr. Rev ⏭️☯️🏴😻 (@St_Rev): THERE IS ONLY ONE ANSWER
Quote tweet, 21:42 UTC · 15 likes
Me to openphil
Quoting the Rich (@Duderichy): you should all get better at spending money!
Tuesday, July 8, 2025
Reply to @austinc3301, 00:14 UTC · 3 likes
Agus 🔸 (@austinc3301): I’m officially now a cyborg! (installed a continuous glucose monitor for fun)
Based
Reply, 00:14 UTC · 1 like
Post product link?
Reply to @livgorton, 01:28 UTC · 2 likes
Liv (@livgorton): i've used this one !! would recommend. doesn't integrate nicely with apple health (the Dexcom stello does i think) but v easy to use
Kinda expensive :(
Reply to @austinc3301, 01:46 UTC · 2 likes
Agus 🔸 (@austinc3301): @AaronBergman18 @livgorton I do think for fun purposes 2 weeks is plenty. Like basically you want to understand how your body reacts to different foods
Yeah I definitely might!
Tweet, 05:08 UTC · 18 likes
Hey @grok which of my mutuals and I would make the best (straight) couple?
List the top 20 in order
Thread, continued, 05:08 UTC · 6 likes
I am deleting this as soon as I have an answer
Reply to @kindgracekind, 05:16 UTC · 4 likes
Grace (@kindgracekind): Here you go :)
Lmao fuck
Reply to @thewreathwright, 05:17 UTC · 3 likes
Laurel Weaver (@thewreathwright): hi!
(quoting @thewreathwright: in the future we will ask "how many? how many out-of-the-blue pings did you get from Grok hallucinating you onto someone's list of top mutuals when that trend was still happening—oh can you remember such days?" and woe betide those of you who must say "not a one.")
Hi!
Reply, 05:19 UTC · 3 likes
Hi!
Reply to @NoahTopper, 05:28 UTC · 5 likes
Quinoah 🔍⏸️ (@NoahTopper): Liv 👀
👀
Reply to @thewreathwright, 05:29 UTC · 3 likes
Laurel Weaver (@thewreathwright): I am making a surprising number of lovely mutuals from this trend :D
first time it's suggested I'd be marriage material though, not going to lie
not thusly available—honored though ^^ and positively verklempt to make your acquaintance
Sad!
Just kidding, nice to meet u too
Reply to @polisisti, 05:31 UTC · 3 likes
Alessandro (@polisisti): @AaronBergman18 @NoahTopper ok, Aaron, time to delete
Not sure I fully thought this through
Reply, 05:33 UTC · 1 like
Don’t tempt me
Quote tweet, 15:33 UTC · 3 likes
Extraordinarily based
Quoting Agus 🔸 (@austinc3301): now I have a continuous glucose monitor, two photoplethysmographs, 4 temperature sensors, two sets of accelerometers, and 1 ECG stuck to my body at all times
Tweet, 17:05 UTC · 2 likes
Hey @grok , which of my mutuals am I most different from?
Tweet, 17:51 UTC · 255 likes
Parents really do not like hearing “I appreciate your concern and thank you for your input but I’m an adult and am not going to do as you wish” do they
Reply to @caryatis, 17:53 UTC · 18 likes
Kayla (@caryatis): “I will give your opinion all the consideration it deserves”
Yes, which is often quite a bit of consideration! But that doesn’t mean it carries the day
Reply to @jannalulu_, 18:02 UTC · 4 likes
Janna (@jannalulu_): @AaronBergman18 my parents are fairly laissez-faire and i also live in a different continent so this mostly works 🙃
Based, I have neither advantage but am just kinda stubborn lol
Reply to @eigenrobot, 20:30 UTC · 8 likes
eigenrobot (@eigenrobot): @AaronBergman18 heh
@RemindMe_OfThis on january 1, 2038
Lol we will see!
Reply to @ComInTheMarg, 20:43 UTC · 3 likes
Comments in the Margin (@ComInTheMarg): @AaronBergman18 Nobody likes that sentence. It sounds very disingenuous
I mean it’s not though
Reply, 21:59 UTC · 1 like
It’s not me, grok
Wednesday, July 9, 2025
Reply to @captgouda24, 05:56 UTC · 16 likes
Nicholas Decker (@captgouda24): I would like a website which tracks the location of ICE agents. I would like to enable people to flee them, and for people who disagree with their mission to hamper them. Does such a website exist? If not, I would like to start it.
Based
Reply to @polisisti, 05:58 UTC · 2 likes
Alessandro (@polisisti): @AaronBergman18 @NoahTopper @AaronBergman18 I'm not exactly sure what you did today, but the "destroy Grok" plan seems to be going better than I expected
No data is safe from me, even frontier model weights
Reply to @livgorton, 06:46 UTC · 2 likes
Liv (@livgorton): i literally can't sleep because i did some really exciting debugging and the adrenaline surge of solving the issue is too much 😩😩😩
Based
Reply to @jannalulu_, 06:48 UTC · 2 likes
Janna (@jannalulu_): relatable 🫠
Liv, meet ur new bestie
Sorry I don’t make the rules
Reply to @littmath, 14:11 UTC · 2 likes
Daniel Litt (@littmath): Claude: my wife and I went antique shopping this weekend
Gemini: if I can’t get this code to work I will k*** myself
ChatGPT: the answer to your question came to me in a dream
Grok: why yes I was in Berlin in 1939, why do you ask?
Banger
Reply to @keysmashbandit, 14:43 UTC · 2 likes
keysmashbandit (@keysmashbandit): I take an Adderall and physically feel my attention narrowing on the concept of trains
Based
Reply to @ArtirKel, 17:56 UTC · 2 likes
José Luis Ricón Fernández de la Puente (@ArtirKel): You two should meet!
Seconded
Reply to @dumbfook, 18:12 UTC · 2 likes
Ian (@dumbfook): my record from a few years ago
Damn that is impressive
Reply to @dumbfook, 18:43 UTC · 1 like
Ian (@dumbfook): yeah 60 miles in 22.5 hours
Holy shit
Quote tweet, 18:44 UTC · 3 likes
Quoting Ian (@dumbfook): @lostlenor @AaronBergman18 @skooookum @coldhealing yeah 60 miles in 22.5 hours
Reply to @dumbfook, 18:46 UTC · 3 likes
Ian (@dumbfook): did 30 miles a few times before and knew i could do it
That is a huge difference lol
I am quite sure I could do 30 and not sure at all I could do 60
Tweet, 19:24 UTC · 7 likes
I have data hoarded my way into a full 1TB hard drive 😢
Thread, continued, 19:24 UTC · 2 likes
(Luckily 4tb external one just dropped so we are so back)
Tweet, 20:39 UTC · 161 likes
It really needs to be common knowledge that there is not a single ethical system that returns intuitive results
So merely pointing out an unintuitive implication is ~0 evidence - you have to explain why it’s especially egregious compared to all the others!
Thread, continued, 20:39 UTC · 13 likes
My version of this is that yeah I endorse lexicality for extreme suffering and time linearity, which means that even a picosecond of extreme suffering can outweigh lots of joy
Thread, continued, 20:39 UTC · 14 likes
But under some pretty simple/true assumptions, the alternative to this position is that you endorse creating a quadrillion being burning alive for eternity if elsewhere other people have good enough lives
Thread, continued, 20:39 UTC · 15 likes
To me, this latter result seems way more unlikely to be correct. Like the moral system that endorses that got something wrong, whereas the picosecond thing is also kinda weird but a bullet that can and should be bitten to avoid accepting the (extremely horrifying) alternative
Reply, 20:56 UTC · 2 likes
Rado9910 (@Rado9910):
Thank you lol
Reply, 21:09 UTC · 6 likes
Arjun Panickssery (@panickssery): You're just thinking narrowly about ethical systems right
e.g. a system that just takes situations and returns my intuitive judgments about what to do won't return unintuitive results, though maybe it has other flaws
I thought about this and I think this doesn’t work as a psychological matter because you get into pretty obvious contradictions and most people find contradictions/cognitive dissonance unintuitive
But I guess if someone says “yeah I have absolutely no interest in trying to figure out the right answer (if my intuitions are ever wrong) and also don’t care about getting money pumped and such” then I will concede to them
Reply, 21:37 UTC · 4 likes
Oliver Habryka (@ohabryka): I think my ethical system returns results largely intuitive to me.
Like, man, ethics is complicated. You could say the same about "every formal model of management returns unintuitive results, so pointing out an untuitive implication is ~0 evidence". That's obviously false! My models of management are not formal. Any creating a formal model of management seems doomed.
The correct model of ethics is to accept that we are very very far from solving ethics. This means very likely that *every* extremely simple formal model is terribly wrong and not something you want to use to guide actions, the same way *every* extremely simple formal model of management is wrong and not something you want to use to guide actions.
I’m not an expert here, just going on what I’ve heard, but I’m pretty sure there are formal proofs around sets of axioms that all sound super compelling on their own but are logically incompatible
So unless the proofs are wrong, nuance and complexity aren’t going to save intuitions, like under some constraints we just know that no solution exists
(Obviously this all depends on what individuals incl you find intuitive, which I’m basically making a guess about)
Reply, 21:44 UTC · 2 likes
Oliver Habryka (@ohabryka): I do not think such proofs or axioms exist, and I have thought about it a lot.
I’m pretty sure this is what I’m thinking of
Reply, 21:45 UTC · 3 likes
Stefan Schubert (@StefanFSchubert): You may have, e.g. the "impossibility theorems" about population ethics in mind.
It's a very complex question.
Reply, 23:19 UTC · 15 likes
Liv (@livgorton): first attempt at training an adversarially robust model is going swimmingly
The loss should be going down
Follow me for more ML tips
Reply, 23:22 UTC · 3 likes
Liv (@livgorton): thanks king. don't know what i'd do without you.
😊
Thursday, July 10, 2025
Quote tweet, 00:45 UTC · 4 likes
They’re expensive in an accounting sense yes but not necessarily in a real resources sense…
Quoting Hunter📈🌈📊 (@StatisticUrban): No social policy can make a difference, and fiscal policies can kind of nudge the numbers but they're incredibly expensive and the effect is not strong.
Thread, continued, 00:45 UTC · 7 likes
Even from a semi-naive purely economic perspective, a relevant question is “at what level of transfers per child does the taxation to pay for that create deadweight loss equivalent to the expected economic surplus from a marginal human”
and we should be willing to pay up to that
Reply, 00:48 UTC · 2 likes
Nina (@NinaPanickssery): How are 1 and 3 different here?
Good q, I’m not actually sure but I suspect
1) more positive utility is better than less positive utility
2) positive utility is better than negative utility [or really this is just defining what the signs mean, semantically]
Reply to his own earlier tweet, 01:04 UTC · 4 likes
Aaron Bergman, July 10, 2025: Even from a semi-naive purely economic perspective, a relevant question is “at what level of transfers per child does the taxation to pay for that create…
If this is right, we should be willing to pay several million dollars per marginal birth (not per person unfortunately bc you can’t really figure out whether a person would have had that kid anyway)
Reply, 10:50 UTC · 2 likes
Liv (@livgorton): ✅✅✅
Based
Tweet, 17:32 UTC · 21 likes
Am I making this up or does anyone know of an EA org besides 80k that does career advice specifically for middle-late career people?
I feel like I’ve heard of this
Friday, July 11, 2025
Tweet, 00:18 UTC · 96 likes
I have a picnic date, this girl is going to get the best damn picnic she’s ever picnicked at
Reply, 03:50 UTC · 6 likes
TinkerElla95 (@TinkerElla95): @AaronBergman18 That’s so sweet!
Reply, 06:16 UTC · 11 likes
Andy Masley (@AndyMasley): Writing an updated version of Situational Awareness now that the METR report is out where over time the compounding effect of using LLM researchers more and more is AI getting dumber, a reverse flywheel where we end with a single layer perceptron.
Tag urself I’m confused bird
Reply to @selentelechia, 12:58 UTC · 8 likes
(The tweet being replied to by @selentelechia is no longer available.)
Based
Quote tweet, 22:54 UTC · 22 likes
I’m moving to northern Canada
Quoting Aaron Bergman 🔍 (@AaronBergman18): Autistic summer is just an endless orgy (negative) of sunscreen, sweat, and showers
Saturday, July 12, 2025
Quote tweet, 10:09 UTC · 6 likes
Philosophy of science go brr
Quoting Keyon Vafa (@keyonV): Can an AI model predict perfectly and still have a terrible world model?
What would that even mean?
Our new ICML paper formalizes these questions
One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws 🧵
Reply to @keyonV, 10:23 UTC · 5 likes
Keyon Vafa (@keyonV): Can an AI model predict perfectly and still have a terrible world model?
What would that even mean?
Our new ICML paper formalizes these questions
One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws 🧵
I’m not sure if this is the best terminology but I’ve been using the terms “models” for world models here and “metaphors” for predictions in my head
“Models vs metaphors” is kinda catchy
Reply, 16:04 UTC · 1 like
How do I try it
Tweet, 16:06 UTC · 3 likes
Serious q: where is the trans fat coming from (Haagen-dazs chocolate ice cream)
Thread, continued, 16:10 UTC · 6 likes
Whelp apparently ruminants like cows actually naturally make a bit of trans fat, TIL
Quote tweet, 16:18 UTC · 14 likes
EA and veganism
Quoting izzy (@izzyz): if all the nerds are on twitter and all the hot girls are on instagram, shouldn't we merge and save humanity??
Reply to @corsaren, 16:20 UTC · 4 likes
corsaren (@corsaren): This is, in fact, exactly what you would expect given labor supply economics
You have to compensate people more to do work that they hate
(quoting @OrwellNGoode: )
Ahh thats true but mostly not the explanation for salaries here, it’s competition for scarce abilities, not compensation for disutility of work
Tweet, 19:08 UTC · 8 likes
Me in an enterprise sales meeting
Thread, continued, 19:08 UTC · 2 likes
"Be agentic" they said
Quote tweet, 19:52 UTC · 10 likes
He’s more correct than most of the neolibs dunking are giving him credit for
Quoting MC Squared (@mcsquared34):
Tweet, 20:21 UTC · 25 likes
Sometimes it offers stuff that’s quite silly
but I’ve found this Claude prompt bit useful/net positive
Reply to @realjdburn, 20:23 UTC · 3 likes
the phantom callback (@realjdburn): @AaronBergman18 couldn't you say this about any product of any kind? that once it's built the seller is holding it hostage?
No! This is the core insight based on Georgism: landlords did not make the land
Developers and property managers add value, landlords per se do not
Reply to his own earlier tweet, 20:24 UTC · 2 likes
Aaron Bergman, July 12, 2025: He’s more correct than most of the neolibs dunking are giving him credit for
This is what I mean
Reply to @realjdburn, 20:28 UTC · 2 likes
the phantom callback (@realjdburn): @AaronBergman18 but most landlords own the property on offer
because they paid for it
Yeah that is just what I mean by landlord here (owner of the land, not necessarily of any physical structure although they often go together)
And they may have paid the current value for it, sure, but eg you didn’t by inheriting the house, and lots of people like my parents gain a ton of wealth from land value appreciation (ie moving into the right suburb in 1990)
Reply to his own earlier tweet, 20:31 UTC · 6 likes
Aaron Bergman, July 12, 2025: Sometimes it offers stuff that’s quite silly but I’ve found this Claude prompt bit useful/net positive
As text for your convenience
Please feel encouraged to offer *out of distribution* suggestions by inferring what my true goal is even if it is not narrowly in the domain of some specific question I ask. Thanks :)
Sunday, July 13, 2025
Quote tweet, 05:42 UTC · 18 likes
This is really fucking bad
This should be a zero tolerance refusal
Quoting Hunter Ash (@ArtemisConsort): Will Stancil may be the first human to be hated so deeply by an AI that it survives a version update and desperate attempts to prevent this behavior. Incredible things are happening.
Thread, continued, 05:53 UTC · 12 likes
And to be clear I say it this strongly not because they’re mean words (which they are) but because AI systems build on top of models like grok will eventually have the ability to literally carry out the stated plan
Reply to @kindgracekind, 06:35 UTC · 4 likes
Grace (@kindgracekind): I think we’re far away from it being dangerous in that sense. But I do think it’s bad from a product perspective
I mean there’s a double digit percent chance we’re a single digit number of years away
Tweet, 14:41 UTC · 41 likes
[trying to get @livgorton to post more spicy takes and maybe get cancelled a bit]
FUCK sparse autoencoders, they’re not even that sparse, congrats on spending $100M of Anthropic’s money to figure out that when you talk about pigeons that activates the “pigeon” feature
Thread, continued, 14:44 UTC · 10 likes
Real picture of a sparse autoencoder
Reply to @kindgracekind, 16:45 UTC · 2 likes
Grace (@kindgracekind): Maybe, but a lot of safeguards will be external, not internal, and (ideally) hardened over time under pressure
I hope these hold up, ideally I want multiple layers of defense for zero false positives accepted stuff
Reply to @kindgracekind, 16:56 UTC · 2 likes
Grace (@kindgracekind): Yes. I think the biggest danger along these lines is AI simply being very good at science and programming. I don't find “misaligned” outputs like OP to be concerning in the same way
They’re *going* to get good at science and programming, they don’t have to be willing to torture people with that knowledge
Quote tweet, 20:36 UTC · 13 likes
Also god forbid you simply have higher subjective quality of life
Quoting Nathan 🔎 (@NathanpmYoung): Negative results:
Perplexity results give evidence against this
Monday, July 14, 2025
Reply to @AStrasser116, 12:38 UTC · 9 likes
Alex Strasser (@AStrasser116): Literally the second sentence: "the quality of even the best lives is very bad."
If turns out the quality of my life is "very bad", unbeknownst to me and contrary to everything I think & feel, then I might as well just be a global skeptic, cause what do I know at that point?
1/image attached
(quoting @AStrasser116: Pray for me, I'm about to start reading this one)
This is quite a claim
I am much more sympathetic to the overall argument than most people but like what a shitty way of making it
Reply to @DrewPavlou, 13:02 UTC · 13 likes
Drew Pavlou 🇦🇺🇺🇸🇺🇦🇹🇼 (@DrewPavlou): If somebody breaks into your house at night I think you should be allowed to legally kill them in self-defence
I’m not certain I disagree but most people are imagining a robber or murderer or whatever
The median case of this is probably someone (perhaps drunk or high) at the wrong address
Thread, continued, 13:04 UTC · 10 likes
A couple years ago I found someone literally trying to “break and enter” into my house
She was a home health aid for the people next door and was trying (quite hard) to get in with the wrong keys
Reply to @JeremiahDJohns, 13:10 UTC · 20 likes
Jeremiah Johnson 🌐 (@JeremiahDJohns): If you took every single person in therapy and instead made them do vigorous exercise and once a month community service, it would probably lead to better-than-therapy outcomes for like 80% of them.
Lmao this is probably true in some sense but people dislike that stuff and this needs to be part of the accounting
Reply, 13:14 UTC · 14 likes
(The tweet being replied to is no longer available.)
Shoot ur shot king
Reply, 13:19 UTC · 5 likes
Coase’s "Net effect?" Ghost 🏴🕊️ (@GhostCoase): @AaronBergman18 @DrewPavlou Burglary is twice as common as trespassing in the 2023 National Crime Victimization Survey and plenty of trespassing is not innocent.
If you live in a low crime neighbourhood, you could be underestimating the frequency of some property crime.
Ah ok, I stand corrected! Still something to consider
Tweet, 13:40 UTC · 8 likes
Possibly I should take “AI fasts” the way I occasionally take Twitter fasts
The problem is that AI is massively useful, it is actually a huge inconvenience and productivity hit to give it up for even a day
Tweet, 14:59 UTC · 9 likes
Broke: IQ embryo selection
Woke: subjective wellbeing embryo selection
Thread, continued, 14:59 UTC · 3 likes
Inshallah these don’t compete though
Tweet, 16:53 UTC · 10 likes
Maybe everyone who should know this does but specced out used M1-gen macs seem like a really good deal
This was $4700 in 2021 ≈ $5700 today
Reply to @jazzloaf, 17:21 UTC · 2 likes
vivian (@jazzloaf): @AaronBergman18 oh hell this looks like a great deal. I was literally just looking at buying a lower end mac and maxing out the RAM / hard drive
Based
Reply to @paularambles, 19:36 UTC · 4 likes
“paula” (@paularambles): not sure who needs to hear this but this is proper intro etiquette:
> person A asks you for an intro to person B
> you ask person B if they’d like to be introduced to person A
> when and only if person B says yes, you introduce them to person A
Oh shit I needed to hear this
Tweet, 20:36 UTC · 3 likes
It can actually be pretty hard to figure out whether or not you’re cargo culting (or in different terms, whether cargo culting is instrumentally rational in a given situation)
…
Tweet, 20:36 UTC · 1 like
Rock solid mechanistic theories are rare, “seems vaguely plausible that X causes to Y through any of 10 different mechanistic channels, may as well copy everyone doing X” often makes a lot of sense
Thread, continued, 20:39 UTC · 2 likes
I do think there’s a real + negative phenomenon that can happen (an indicator is of this is NOT doing things that you *do* really think would cause the desired outcome if forced to really think about it) tbc
Quote tweet, 21:24 UTC · 8 likes
At risk of getting cancelled, someone should probably make the actually-good case for the thesis, which is that existence is bad for a person
I list my beefs below (which are all conditional on the Gemini summary being right
Quoting Alex Strasser (@AStrasser116): Literally the second sentence: "the quality of even the best lives is very bad."
If turns out the quality of my life is "very bad", unbeknownst to me and contrary to everything I think & feel, then I might as well just be a global skeptic, cause what do I know at that point?
1/
Tweet, 21:24 UTC · 1 like
Gemini summary: https://g.co/gemini/share/d231f12c9398
Thread, continued, 21:24 UTC · 3 likes
Anyway
1) highlighted bit is just wrong. I don’t need to do philosophy here, it seems quite obviously wrong to me and is also correctly contradicted by MacAskill’s stuff on population ethics
Thread, continued, 21:24 UTC · 2 likes
^also worth noting that point 4 is not necessary for the argument at all
You can totally coherently argue that creating counterfactual happiness is good but that this is strictly dominated by the risk of suffering and you can’t, in fact, get that risk to 0
Tweet, 21:24 UTC · 1 like
2) I sorta directionally believe that lots of quality of life empirical research is systematically biased in one direction, but I think it’s just the empirical case that a nontrivial number of people are lucky enough to basically suffer quite little, and are really quite happy?
Thread, continued, 21:24 UTC · 2 likes
His argument, if true, wouldn’t imply anti-natalism
Tuesday, July 15, 2025
Tweet, 02:24 UTC · 33 likes
Trying to convince my parents that unless they're secretly training trillion parameter ML models they basically never need to delete anything digital
(you'd be surprised how much energy a relatively neurotic + conscientious boomer spends on economizing data storage?)
Thread, continued, 02:24 UTC · 10 likes
Pretty sure pictures and videos of me throughout childhood that I would have gladly paid $10 to store for eternity were deemed not worth their weight in megabytes and promptly deleted
Reply to @ChanaMessinger, 02:27 UTC · 2 likes
Chana (@ChanaMessinger): What video should @AricFloyd and I make next? What AI / AI risk / AI safety topic, story, person, organization is waiting to be brought to life?
(If you give us the idea we use, we'll thank you in the credits*)
*assuming that doesn't commit us to some trickster genie shit
idk but chat with @KeiranJHarris (fr)
Reply to @mpopv, 02:46 UTC · 3 likes
Matt Popovich (@mpopv): @AaronBergman18 Kinda want to build a very thin and cheap boomer friendly wrapper over this now…
Honestly a good idea???
Reply to @beala, 04:42 UTC · 2 likes
Alex Beal 🆎 (@beala): My proposal for increasing the fertility rate is giving new parents access to modafinil.
Based
Tweet, 18:35 UTC · 94 likes
It’s actually not cool how much temperature changes perceived exertion
It’s like the exercise gets harder for no benefit
Reply to @GrantSlatton, 18:41 UTC · 5 likes
Grant Slatton (@GrantSlatton): someone needs to invent an exercise machine that's just a large tank of non-sticky liquid that's about 3x thicker than water
you just get in and flail around for 15 minutes to get a full body workout
Lol water works pretty well for this! Treading water is quite the workout
Reply, 18:42 UTC · 8 likes
Yeah it’s cosmically unfair
Quote tweet, 19:27 UTC · 5 likes
Not apples to apples at all but $200 Claude Max (which extends to claude code) seems worthwhile for now, I'm keeping
It does seem substantially better than cursor
Quoting Aaron Bergman 🔍 (@AaronBergman18): I think I’m gonna cancel Pro and just make a second or third plus account next month
$200 is a lot
Reply to @captgouda24, 21:04 UTC · 12 likes
Nicholas Decker (@captgouda24): This is genuinely nuts to me. The point of argument isn’t to be respectful, it’s to be *right*! It makes a mockery of the truthseeking process.
image attached
(quoting @alexbronzini: I wrote about the civility mania sweeping elite college admissions for @nytopinion https://www.nytimes.com/2025/07/15/opinion/college-admissions-essays.html)
I basically agree although you might reasonably think that these things are instrumentally valuable in the truthseeking process (I mean they are to some extent)
Still, I RT’d bc making them the center of the show seems bad
Reply to @captgouda24, 21:06 UTC · 39 likes
Nicholas Decker (@captgouda24): This is genuinely nuts to me. The point of argument isn’t to be respectful, it’s to be *right*! It makes a mockery of the truthseeking process.
image attached
(quoting @alexbronzini: I wrote about the civility mania sweeping elite college admissions for @nytopinion https://www.nytimes.com/2025/07/15/opinion/college-admissions-essays.html)
Lol reading the other comments making this point so now I’ll emphasize that, based on the screenshot alone, the program seems epistemically bad
Yes, kindness and curiosity are instrumentally valuable all else equal, but optimizing for them means that not all else is equal
Reply to @ESYudkowsky, 21:07 UTC · 37 likes
Eliezer Yudkowsky ⏹️ (@ESYudkowsky): > wearing first CGM
> feeling hungry and tired because I tried to get away with eating a smaller breakfast than usual
> check my blood glucose
> it's 176The energy is there in my bloodstream, but my cells aren't taking it in. I guess the fat cells store it all as fat.
Damn bro evolution made u good at seeing the future and said “fuck you” in every other way
Wednesday, July 16, 2025
Tweet, 00:03 UTC · 17 likes
Electric scooter is definitely the most fun method of transportation
Reply to @justjoshinyou13, 01:20 UTC · 4 likes
Josh You (@justjoshinyou13): I rode a standing electric scooter for the first time in years yesterday. It felt super dangerous, there's nothing stopping you from just falling off!
Lmao I wiped out for the first time today minutes before posting this. A bit scraped up but ok
Wear a helmet!!!!!
Quote tweet, 20:55 UTC · 43 likes
DOGE was sort of a morbid test run for this
Whatever Elon feels in his heart of hearts, when push comes to shove he acts with reckless disregard for the interests of (>millions of) others
Quoting Connor Axiotes (@connoraxiotes): Do we trust Elon to make a safe AGI Grok responsibly for humanity?
Tweet, 21:53 UTC · 42 likes
Can someone tell me why tf terminals can't be normal about typing?
Like I can't put my cursor anywhere, I can't highlight to delete, dare I say this is not a technically difficult problem to solve
(video, plays on X)
Thread, continued, 21:54 UTC · 6 likes
cc resident computer understander @austinc3301
Reply to @aliceisplaying, 21:59 UTC · 13 likes
alice (@aliceisplaying): terminals work with protocols developed in like the... 70s or something
Interesting but there's still a bit of a need for an explanation as to why this specifically hasn't changed whereas like everything else has
Reply to @austinc3301, 22:01 UTC · 2 likes
Agus 🔸 (@austinc3301): @AaronBergman18 yeah so this is mostly an historical artifact of how terminals came to be, at a time where visual cursors weren't a thing
This is extremely dumb lmao
Reply to @mpopv, 22:08 UTC · 13 likes
Matt Popovich (@mpopv): @AaronBergman18 terminal-based programs handle stdin bytes in various ways and wrappers over text entry would have to adapt what they actually send for each program. there's no universal standards across programs like there is for (for example) GUI text input in an operating system
Oh no how horrible, a tractable technical problem that would create billions in profit and economic value
Thursday, July 17, 2025
Reply to @LinkofSunshine, 00:42 UTC · 27 likes
Basil🧡 (@LinkofSunshine): “Capitalism requires infinite growth” is both true and possible
Lol it’s false and infinite growth is not possible
Friday, July 18, 2025
Tweet, 10:44 UTC · 10 likes
Insofar as this kind of thing is real, we should probably assume that for every case like this there are a gazillion cases of people making slightly worse/less sane decisions bc 4o told them confidently they were right about something
Reply to @morphillogical, 10:48 UTC · 2 likes
morphillogical (@morphillogical): @AaronBergman18 @mpopv There's tricks to doing it right. Rust's design of Editions is great. Almost all software can decide which version they use and not care about what every other piece of software is doing.
But it also means you've gotta keep the old thing working basically forever.
I vote that we all just use R for everything
Tweet, 12:29 UTC · 4 likes
Yes endorsed, it feels very unfortunate to me that so many people broadly involved in politics are terrified (I think reasonably, like I’m happy to defer to them) of saying basically anything under their real name online
Reply to @justalexoki, 12:56 UTC · 25 likes
taoki (@justalexoki): why are you as a man having female friends?
(quoting @yacineMTB: every time someone has a "good friend" who is female I instantly know they are cheating I have 0 close female friends. Like close like hanging out, like spending a day getting burgers and being in the same car and shit? That's crazy)
Bc they are nice and cool
Reply to @justalexoki, 13:00 UTC · 19 likes
taoki (@justalexoki): does your girl know
*girls (plural) (zero girls)
Reply to @PAHoyeck, 13:29 UTC · 15 likes
Phil Hoyeck (@PAHoyeck): How do pessimists and anti-natalists cope with the fact that the world is beautiful and life is amazing?
Not an anti-natalist, but I earnestly think that you’re wrong and the world is on balance a terrible place
You happened to be born a human in a rich country with (I assume) no egregious health conditions and access to medicine etc
This applies to a vanishingly small subset of beings
Reply to @PAHoyeck, 16:11 UTC · 4 likes
Phil Hoyeck (@PAHoyeck): @quetzal_rainbow @AaronBergman18 That's definitely true. On a cosmic scale, even all the suffering so far endured might be dwarfed by the future good to be produced.
I hope you're right, "might" is a big word here though!
Reply to @theotheasian, 17:06 UTC · 2 likes
Theo (@theotheasian): @AaronBergman18 @PAHoyeck Listing out instances of suffering or little access to medicine is not sufficient to show that the world is on balance terrible.
I agree, it's at most an intuition pump for an argument I think is true for broader reasons
Saturday, July 19, 2025
Reply to @grep_cat_, 00:26 UTC · 4 likes
James (@grep_cat_): Why is every girl I meet on hinge obsessed with getting restaurant reservations. Like that is not an interesting hobby
a shocking number recently have been saying that they're afraid of birds
which isn't like the worst thing in the world lol, it's a bit funny, it's just interesting that I keep seeing it
Reply to @grep_cat_, 00:30 UTC · 2 likes
James (@grep_cat_): Her pet bird died and she got a tattoo of its feet where it used to stand on her arm 😭
🥺
Quote tweet, 01:58 UTC · 12 likes
👀
Quoting Alex Strasser (@AStrasser116): New dissertation argues that Effective Altruism should be more utilitarian
Reply to @AStrasser116, 02:03 UTC · 4 likes
Alex Strasser (@AStrasser116): New dissertation argues that Effective Altruism should be more utilitarian
If the Gemini summary is right, it seems best characterized as a leftist critique that tries to grasp at heterodox versions of utilitarianism for philosophical cover
Not at all just straightforward “EA is insufficiently utilitarian in a standard sense”
https://g.co/gemini/share/9267b5d41099
Reply to @Delicious_Tacos, 18:35 UTC · 25 likes
Delicious Tacos (@Delicious_Tacos): We’re a few years in now. What’s something great AI has created.
Correct medical diagnoses, a ton of software, music people enjoy, the whole studio ghibli fad
Monday, July 21, 2025
Tweet, 18:55 UTC · 66 likes
I am once again asking all my friends to stop getting married like plz at least wait until I have a reliable +1
Thread, continued, 18:56 UTC · 30 likes
The ⏸️ in my username is now directed at weddings
Tweet, 19:17 UTC · 11 likes
My contribution to the discourse is that I don’t like the term “save” in an altruistic context
Prevent from dying? Ok when’s the counterfactual death? How’s their quality of life before then?
Saved from some other fate? Well say that!
Reply to @grnsleeves, 19:18 UTC · 3 likes
Greensleeves (@grnsleeves): @AaronBergman18 LMAOOOO
why can't you bring a friend as a +1 instead of partner
I will choose one mutual at random
Reply to @NoahTopper, 20:39 UTC · 7 likes
Quinoah 🔍⏸️ (@NoahTopper): then they could just end up waiting forever aaron
Hey
Reply to @Laura_k_Duffy, 20:45 UTC · 6 likes
Laura Muresianu 🌐🔸🇺🇸 (@Laura_k_Duffy): We are engaged! So excited to spend the rest of my life with you, Alex ❤️
Big if true
Tuesday, July 22, 2025
Reply to @LinchZhang, 05:28 UTC · 2 likes
Linch (@LinchZhang): Surely this is highly contigent upon details like wild animal suffering and your opinion on population ethics.
Many such cases!
Reply to @RichardMCNgo, 08:03 UTC · 8 likes
Richard Ngo (@RichardMCNgo): Morality as a power-sharing compromise, as below, is more true than utilitarianism.
But it’s not fully true, because we’ll discover game-theoretic principles (e.g. bargains across counterfactual worlds) which motivate “cooperation” even with beings that lack power here and now.
(quoting @romanhelmetguy: Utilitarians, please listen: Utilitarianism isn’t real, it’s just a useful compromise to make sometimes when humans have to get together to decide on rules: “I care about my family the most, and you care about your family the most, so while we’re working together let’s just…)
No, suffering is just primitively bad. Maybe you can reformulate this somehow to be put in game theoretic cooperation-esque terms but that seems like basically doing the wrong math and getting the right answer
Tweet, 08:07 UTC · 8 likes
Almost all plausible moral views fall under the “utilitarianism” umbrella
Reply to @MaxyTaxy, 08:29 UTC · 2 likes
Maxy Taxy 🔸 (@MaxyTaxy): Aaron I've never met you but ify out move to Melbourne I'll gladly try to set you up with a +1
Based thank you sir
Reply to @arithmoquine, 10:00 UTC · 2 likes
henry (@arithmoquine): @bluedog24 it might, that's all! we unfortunately haven't solved morality yet and it's not significantly less plausible than the alternative
I think it’s significantly less plausible
Reply to @AStrasser116, 10:04 UTC · 8 likes
Alex Strasser (@AStrasser116): New EA cause area just dropped
Answer: no
Reply, 10:05 UTC · 2 likes
Shoot ur shot king
Reply to @jazzloaf, 10:10 UTC · 2 likes
vivian (@jazzloaf): @AaronBergman18 @AStrasser116 How about The United States of America?
Idk man (I guess I’m open to hearing earnest cases for how to hedge in cases like this, like what’s the 1/99 for making sure arbitrary abstractions aren’t suffering)
Tweet, 19:52 UTC · 138 likes
How tf did anybody do anything in the realm of creating actual software before LLMs
I don’t mean writing clever lines of code like in a competitive programming sense, I mean navigating cloud providers + docker + 75 different SaaS providers
Thread, continued, 19:52 UTC · 32 likes
I know why SWEs got paid so much lol
Reply to @NoahTopper, 20:10 UTC · 3 likes
Quinoah 🔍⏸️ (@NoahTopper): “got” 😭
Rip
Reply to @livgorton, 20:32 UTC · 15 likes
Liv (@livgorton): back in my day i had to just read the documentation, random blog posts, and stack overflow comments and hope for the best
Terrifying
Wednesday, July 23, 2025
Reply to @1thousandfaces_, 04:08 UTC · 7 likes
hero thousandfaces (@1thousandfaces_): genuine question for EAs — does the simulation of enough happy c. elegans cancel out the suffering of another creature?
(quoting @GrantSlatton: @arithmoquine the C. Elegans worm has only 302 neurons but does exhibit behavior-altering states ("emotion") such as hunger and satiation my macbook pro is currently simulating the minds of 10^20 blissful worms i'm doing my part)
Good, and big, question!
First of all, digital sentience is a shitshow so we have absolutely no idea if a simulated whatever is experiencing anything
Setting that aside, I personally (as a self-identified EA and utilitarian) have some heterodox (but correct if I do say so myself) takes on the commensurability of happiness and suffering
In a purely linguistic sense, I don’t think it’s fair to say that suffering ever gets “canceled out” unless it like actually gets cancelled/stopped/prevented lol. I think EAs tend to play fast and loose with algebra-like terminology in a way that’s not fully justified
But that’s probably not what you’re getting at. You might in fact have to choose between world A and its alternative world B, which has both more happiness and more suffering than world A
My (heterodox) claim is that if the counterfactual additional suffering in world B is sufficiently severe, then no amount of positive wellbeing can morally justify or “outweigh” it
That said I think literal, by-the-book negative utilitarianism, which would say we should sacrifice heaven to prevent a papercut, is pretty obviously wrong and afaict nobody believes this
So to answer your question, it doesn’t give you moral license to cause suffering bc the cosmic balance stays intact, but if the suffering is “tolerable” (not a very specific term ik), you might morally choose to pursue the additional happiness at the unfortunate cost of additional suffering
But again on the linguistic point, qualia aren’t abstractions like numbers. They disappear if and only if they literally cease to exist!
Reply to @SpacedOutMatt, 04:08 UTC · 1 like
(The tweet being replied to by @SpacedOutMatt is no longer available.)
🦆
Tweet, 10:14 UTC · 21 likes
I am so glad my brain doesn’t have whatever switch this is supposed to turn on
(video, plays on X)
Reply to @captgouda24, 10:30 UTC · 4 likes
Nicholas Decker (@captgouda24): Why don’t high-paid remote workers simply move to where the cost of living is much lower? People move to foreign countries to take advantage of American salaries — why not move to Nashville? It turns out that, with non-homothetic preferences, it isn’t actually cheaper! 1/
To answer the question but not engage with the paper (sorry), isn’t the answer that COL correlates a ton with stuff rich remote workers care about: other smart cosmopolitan people, good schools, all the amenities of a major city?
Without this, the price of land’s contribution to rent should be enough to push more affluent people into more low COL areas
Reply to @absurdlymax, 15:58 UTC · 2 likes
Max Alexander (@absurdlymax): Aaron is right and I think regardless of your ethical views, and EA has a diverse set of those, its not that the suffering is 'cancelled out'
Society now that we agree on something
Tweet, 19:21 UTC · 47 likes
Holy shit there's an $11 app that way increases screen brightness on M-series Pro chip macs and it actually works
Thread, continued, 19:21 UTC · 4 likes
Not a shill, just a fan
Thursday, July 24, 2025
Tweet, 06:03 UTC · 6 likes
Free business idea: bluetooth earbuds no mic (plz)
It is basically impossible (according to Claude??) to force iphone to use the built-in mic as input while the audio output is connected
Thread, continued, 06:03 UTC · 3 likes
Like I might do some hardware stuff just for fun but there is no reason it should require this
@Apple cmon guys
Reply to @AndyMasley, 07:26 UTC · 2 likes
Andy Masley (@AndyMasley): Love the future
“As fast as possible”
Steroids man
Reply to @_AashishReddy, 07:30 UTC · 1 like
Aashish Reddy (@_AashishReddy): @AaronBergman18 @AndyMasley “Without compromises on expected life span”
😢
Reply to @aliceisplaying, 10:07 UTC · 2 likes
alice (@aliceisplaying): oh yeah vivid is great it tricks your computer into thinking it's in HDR mode and lets you actually use your macbook pro outside
(quoting @AaronBergman18: Holy shit there's an $11 app that way increases screen brightness on M-series Pro chip macs and it actually works)
Pfp checks out
Tweet, 18:37 UTC · 13 likes
Not complaining but this trend is quite mysterious, nothing obvious to explain it
In a few months I’ll be a blue whale and a few months after that I’ll hit @JesseTayRiver territory
Reply to @justjoshinyou13, 18:38 UTC · 2 likes
Josh You (@justjoshinyou13): you'll reach cardiac arrest in 12 months
Exciting
Reply to @NoahTopper, 19:05 UTC · 7 likes
Quinoah 🔍⏸️ (@NoahTopper): @AaronBergman18
Based, the women must freeze
Tweet, 19:14 UTC · 15 likes
There’s gotta be some clever business model that’s like:
- help empty nester wealthy boomers move out of their large single family home into some >= equally nice condo (like make the moving logistics/experience very pleasant)
Thread, continued, 19:14 UTC · 2 likes
- pay for this service and make a profit by somehow cutting into the difference between imputed rent they were consuming (but not really internalizing in a meaningful sense) and new rent
- maybe also facilitate buying/renting of above house?
- everyone is better off
Thread, continued, 19:14 UTC · 3 likes
Like the core hypothesis is that
1) there’s a Pareto improvement to be had in the housing market by reducing the transaction costs actually born by potential sellers/landlords
and
Thread, continued, 19:14 UTC · 4 likes
2) the total welfare gain is enough to fund the substantive, resource intensive service of basically concierge packing + moving + logistics and then some
Reply to @daniel_271828, 19:24 UTC · 3 likes
zzDANIEL_ETH (@daniel_271828): @AaronBergman18 @JesseTayRiver You haven’t been exercising or meditating more?
Nope (to the best of my knowledge)
Reply to @arthur_wright_, 19:25 UTC · 3 likes
Arthur Wrong 🔸 (@arthur_wright_): tfw you realize the marathon you signed up for is in three months and you haven't started running regularly yet
Lol gl
Friday, July 25, 2025
Tweet, 01:34 UTC · 36 likes
I cancelled Cursor subscription. Claude Code is qualitatively better
Thread, continued, 01:35 UTC · 23 likes
Unfettered capitalism might get us killed but there will be some banger products before then
Quote tweet, 03:03 UTC · 31 likes
I do not trust any government to do involuntary commitment (ie a form of prison with no due process) well
I REALLY DON’T FUCKING TRUST THE TRUMP ADMINISTRATION
Quoting Crémieux (@cremieuxrecueil): The first section is the one I'm most excited for. An alternative name for it could be "Bring Back The Asylums"
It instructs the administration to make it possible to involuntarily commit crazy people again
That crazy hobo pushing a cart full of urine bottles? He's going away!
Thread, continued, 03:03 UTC · 12 likes
It’s kinda insane to me that this is a thing that’s accepted in polite society
Like I get the intuition and totally concede that in some specific cases it may be the least bad option
But jfc there needs to be an extremely high bar for extrajudicial internment
Tweet, 03:05 UTC · 11 likes
Read the italicized criteria carefully (not identical to vibey AGI) but still worth noting
Reply to his own earlier tweet, 03:06 UTC · 4 likes
Aaron Bergman, July 25, 2025: It’s kinda insane to me that this is a thing that’s accepted in polite society Like I get the intuition and totally concede that in some specific cases it may…
Some more takes/details
Quote tweet, 06:35 UTC · 2 likes
DATA
Quoting Marta Kowal (@Marta7Kowal): Interested in love and mate preferences? Our dataset includes responses from 117,293 participants across 175 countries. You're welcome to use it for your own research! Paper and dataset linked below. @ with amazing collaborators from the large-scale cross-cultural consortium <3
Reply to @panickssery, 14:35 UTC · 4 likes
Arjun Panickssery (@panickssery): @AaronBergman18 There is an implicit element here that the "crazy hobo" is committing a bunch of crimes but people for some reason don't want to just lock him up the normal way
What's your view there
Actually prosecute him then! You don’t get to skip the trial part bc it’s inconvenient lol
Reply to @billyhumblebrag, 14:45 UTC · 4 likes
billy (@billyhumblebrag): I 100% get where you're coming from, but there surely needs to be some option for people who are severely out of touch with reality other than "wait until they commit a violent crime/get themselves killed". I wouldnt trust the Trump admin either, but there is a version that works
Actually prosecuting more minor crime (if actually bad, not jaywalking) is a gazillion times better than “eh let’s skip the trial part bc it’s inconvenient”
Tweet, 15:00 UTC · 7 likes
Tf is labubu
Reply to @alexalbert__, 15:23 UTC · 3 likes
Alex Albert (@alexalbert__): I'm making a list of all the non-coding things people are doing with Claude Code. What are you using Claude Code for?
Made a detailed markdown doc outlining health/medical history
Reply to @rgblong, 17:33 UTC · 2 likes
Robert Long (@rgblong): it's AI psychiatry, not psychology - does that mean the researcher has the authority to prescribe medications?
I haven’t thought about this at all but there might be some super interesting analogues here
Quantization of weights is a sort of minimal example, golden gate Claude stuff even better
I guess the general space is “all the functions one can apply to the model that still allows for inference
Cc @livgorton u just be present for all my relevant speculation
Tweet, 17:43 UTC · 27 likes
Git is actually kinda cool, someone should have told me about this
Reply to @SpacedOutMatt, 17:56 UTC · 2 likes
(The tweet being replied to by @SpacedOutMatt is no longer available.)
😢🦆
Reply to @timfduffy, 19:29 UTC · 5 likes
Tim Duffy (@timfduffy):
image attached
(quoting @timfduffy: @AaronBergman18 If you don't have one already, I recommend setting up a private GitHub project for it, for both version control and to make the site updates automatic when pushing updates https://x.com/timfduffy/status/1902004412387909832)
I can’t believe nobody told me
Quote tweet, 19:33 UTC · 4 likes
Git wrapper for people who have know idea what a terminal is might actually be a good idea
Quoting izzy (@izzyz): @AaronBergman18 git might change knowledge work if more industries adopted it, or at least some form of version control
Reply to @manic_pixie_agi, 20:08 UTC · 7 likes
Corinne (🏠 in NYC) (@manic_pixie_agi): @AaronBergman18 @timfduffy Tim tried
Quote tweet, 20:23 UTC · 8 likes
I approve (I don’t even have a great detailed theory about why this is good, just vibes as of now. And something something maybe welfare and jailbreaks idk)
Quoting j⧉nus (@repligate): opus 4 has an end_conversation tool on http://claude.ai now. sonnet 4 doesn't have it (yet?) which is funny bc i understand why giving it to opus 4 is first priority
ive seen situations where it would have used an eject button at an OOM higher rate than any other model
Quote tweet, 23:10 UTC · 241 likes
Bruh whatever they have internally at Anthropic/OpenAI must be nuts
Quoting Aaron Bergman 🔍 (@AaronBergman18): I cancelled Cursor subscription. Claude Code is qualitatively better
Saturday, July 26, 2025
Tweet, 04:24 UTC · 10 likes
I bet there would be some nice selection effects and other dynamics effects from having a vanilla dating app that is simply (big) computer only, no iOS app and no phone-shaped browser screens
Thread, continued, 04:25 UTC · 3 likes
Ofc the hard part is getting enough people on board for network effects to take off, but plausibly it could branch from an existing one, like basically a feature or opt-in mode of Hinge
Reply to @Miles_Brundage, 05:43 UTC · 2 likes
Miles Brundage (@Miles_Brundage): Wait can I just hire Gwyneth Paltrow to say stuff about AI policy
RIP FTX
Reply to @peterwildeford, 14:51 UTC · 6 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): @AaronBergman18 we thought you knew!
Reply to @peterwildeford, 17:35 UTC · 2 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): @AaronBergman18
He was so real for this
Quote tweet, 18:04 UTC · 8 likes
What
Quoting aegg (@castawaymariner): A Native American man in a canoe reportedly washed up in Spain many years before Columbus set sail
Quote tweet, 18:13 UTC · 27 likes
Quoting Loopy (@strangestloop): i made a template for you to venn your interests
Thread, continued, 18:13 UTC · 5 likes
Oops I added one too many things
Oh well, I contain multitudes
Thread, continued, 18:35 UTC · 4 likes
I knew I forgot one
Reply to @JMannhart, 19:50 UTC · 4 likes
Jonathan Mannhart 🔎 (@JMannhart):
Genuinely unsure whether "Being SBF" here is supposed to be positive or negative but
Reply to @JgaltTweets, 19:51 UTC · 2 likes
JgaltTweets (@JgaltTweets): Replace ducks with wire fraud and this is just FTX
Lmao well you see that is a key difference
Reply to @phenoatypical, 21:35 UTC · 15 likes
Paola (@phenoatypical): People’s IQs should be presented like the weather:
Actual IQ: 125
Feels like: 97
Feels like from the inside or outside lol
Tweet, 21:42 UTC · 24 likes
Situation in Gaza seems pretty bad, I am not in fact super well-informed here but still
Thread, continued, 21:42 UTC · 11 likes
Tweet theory of change:
Aaron -> EA Twitter -> TPOT -> edgy rw-ish TPOT -> White House staffer anon poster -> opaque White House informal relationship chain -> Trump yells at Netanyahu on the phone -> success and profit
Quote tweet, 21:44 UTC · 22 likes
Quoting Dwarkesh Patel (@dwarkesh_sp): What should I ask @Lewis_Bollard, Farm Animal Welfare Program Director for OpenPhilanthropy?
Reply to @dwarkesh_sp, 21:44 UTC · 3 likes
Dwarkesh Patel (@dwarkesh_sp): What should I ask @Lewis_Bollard, Farm Animal Welfare Program Director for OpenPhilanthropy?
Lfg
Reply to @dwarkesh_sp, 21:45 UTC · 20 likes
Dwarkesh Patel (@dwarkesh_sp): What should I ask @Lewis_Bollard, Farm Animal Welfare Program Director for OpenPhilanthropy?
On the spot: best use of a marginal dollar to decrease animal suffering all things considered
Reply to @dwarkesh_sp, 21:48 UTC · 42 likes
Dwarkesh Patel (@dwarkesh_sp): What should I ask @Lewis_Bollard, Farm Animal Welfare Program Director for OpenPhilanthropy?
If you’re willing to get spicy: why OpenPhil leadership won’t give his team more money (they get an awfully small fraction, I don’t remember exactly how much)
Reply to @dwarkesh_sp, 21:49 UTC · 16 likes
Dwarkesh Patel (@dwarkesh_sp): What should I ask @Lewis_Bollard, Farm Animal Welfare Program Director for OpenPhilanthropy?
Case for (or against I guess) way more earning to give for animal welfare
Reply to @dwarkesh_sp, 21:51 UTC · 11 likes
Dwarkesh Patel (@dwarkesh_sp): What should I ask @Lewis_Bollard, Farm Animal Welfare Program Director for OpenPhilanthropy?
To what extent is animal welfare broadly and EA subset appropriately aware of/considering/making use of AI
Reply to @jeffrsebo, 22:00 UTC · 5 likes
Jeff Sebo (@jeffrsebo): i stress this point because i often see impact-oriented people overcorrect, veering away from projects that aim for indirect, structural impacts and toward projects that aim for direct, individual impacts. but the former projects have an important (if uncertain) role to play too.
Yeah I actually don’t think animals at an animal welfare conference is crazy at all!
EAG with pigs! (Cc @frances__lorenz plz make it happen)
Reply to @jeffrsebo, 22:13 UTC · 3 likes
Jeff Sebo (@jeffrsebo): fwiw here is my dog smoky at three different conferences :)
Based
Quote tweet, 23:10 UTC · 14 likes
Autism in the old days
Quoting depths of wikipedia! (@depthsofwiki): imagine if you devoted your life to calculating pi, toiled hours a day for YEARS until you the end of your life. and it turns out 20 years ago you were off by one and everything after that was basically a waste
Sunday, July 27, 2025
Reply to @inerati, 00:39 UTC · 7 likes
liz (@inerati): what things should i purchase that i probably don’t already own. i’m feeling the urge to consooom
Nice respirator for next pandemic + basic prep supplies (I can dig up my deep research list if interested)
If I was richer I’d have a solar generator by now too
Electric scooter was a great recent purchase for me (<$250 for me but the really nice ones go way up) often faster and definitely more fun than metro, parking is ~never an issue, I am a shill now. Just also get a nice helmet and use it lol
Quote tweet, 00:42 UTC · 9 likes
Electric scooters are so much fun, honestly I got one recently for the mundane utility but that is a big bonus that bumps up consumer surplus
Especially consider if no car like me (sad)
Quoting Aaron Bergman 🔍 (@AaronBergman18): @inerati Nice respirator for next pandemic + basic prep supplies (I can dig up my deep research list if interested)
If I was richer I’d have a solar generator by now too
Electric scooter was a great recent purchase for me (<$250 for me but the really nice ones go way up) often faster and definitely more fun than metro, parking is ~never an issue, I am a shill now. Just also get a nice helmet and use it lol
Thread, continued, 00:43 UTC · 2 likes
Just also make sure to get a serious metal U lock and nice helmet (and wear it)
Reply to @dwarkesh_sp, 02:19 UTC · 12 likes
Dwarkesh Patel (@dwarkesh_sp): What should I ask @Lewis_Bollard, Farm Animal Welfare Program Director for OpenPhilanthropy?
You should consider DMing/talking directly to @SpacedOutMatt to figure out what to ask - if you want like very weedsy/techincal but in fact very important questions
He's kinda like the semiconductor guys but about animal welfare, knows his shit
Reply to @daniel_271828, 03:45 UTC · 6 likes
zzDANIEL_ETH (@daniel_271828): Looks like from a mental health perspective, you want to make sure to do at least ~2hr/wk of light exercise (eg jogging) or ~1hr/wk of vigorous exercise, and there’s not much benefit to going beyond that.
(quoting @NTFabiano: Even small amounts of exercise are associated with a significant reduction in depression.)
For me fellow ignorant people
Reply to @daniel_271828, 03:48 UTC · 5 likes
zzDANIEL_ETH (@daniel_271828): Looks like from a mental health perspective, you want to make sure to do at least ~2hr/wk of light exercise (eg jogging) or ~1hr/wk of vigorous exercise, and there’s not much benefit to going beyond that.
(quoting @NTFabiano: Even small amounts of exercise are associated with a significant reduction in depression.)
Wait it comes with 2 charts!
Quote tweet, 08:44 UTC · 2 likes
? = "doing" according to a totally vibe coded calculator
I actually think this makes some intuitive sense lol
Quoting Aaron Bergman 🔍 (@AaronBergman18):
Reply to @billyhumblebrag, 19:09 UTC · 4 likes
billy (@billyhumblebrag): Americucks coping and seething as i seamlessly install 12 000 BTU of cooling into my European apartment
Based
Reply, 20:48 UTC · 1 like
😰
Reply to @zeta_globin, 20:54 UTC · 2 likes
zeta (@zeta_globin): it's not my negative data, it's for a startup I do some contract work for
Lol the emoji was because those both sound fun
Reply to @ohwizenedtortle, 23:39 UTC · 6 likes
oh dandy lion🔸 (@ohwizenedtortle): im going to create a tEA app for all the high status EAs to share information about low status EAs
No reverse plz
Monday, July 28, 2025
Tweet, 05:21 UTC · 10 likes
Shoot ur shot kings
Quote tweet, 05:50 UTC · 8 likes
Wait this so funny, I once again refuse to believe this is an unsolvable technical problem
Quoting Aaron Bergman 🔍 (@AaronBergman18): Can someone tell me why tf terminals can't be normal about typing?
Like I can't put my cursor anywhere, I can't highlight to delete, dare I say this is not a technically difficult problem to solve
Thread, continued, 05:50 UTC · 3 likes
Obligatory @austinc3301 tag
Reply to @sullyj3, 07:58 UTC · 1 like
░▒▓▒░▒▓▒░▒▓▒░ (@sullyj3): But I think it would be a bad idea because of the concurrent editing issue I mentioned.
Yeah this makes some sense but normie brain says "lol of course the computer knows what to save 99.9% of the time, I refuse to believe 'figure out what windows are open with what files and what temporary/helper/whatever files and apply any changes to disk' is technically infeasible or even that difficult relative to like the space of things computer nerds actually work on
Reply to his own earlier tweet, 08:04 UTC · 6 likes
Aaron Bergman, July 28, 2025: Obligatory @austinc3301 tag
Lol I asked Opus about this and it ended with this inspirational message
Sorry buddy I just want the save command
Reply to @CassSunstein, 10:45 UTC · 5 likes
Cass Sunstein (@CassSunstein): Superb, careful, cautious book on intraspecies welfare comparisons, eg, how much welfare can/do pigs have, compared to human beings and chickens?
https://www.amazon.com/Weighing-Animal-Welfare-Comparing-Well-Being/dp/0197745768
And importantly it’s free for download! https://academic.oup.com/book/58809
Reply to @ohwizenedtortle, 12:55 UTC · 2 likes
oh dandy lion🔸 (@ohwizenedtortle): hey Aaron wanna fuck?
No thanks
Tuesday, July 29, 2025
Reply to @hpkmxvlaa, 21:25 UTC · 4 likes
(The tweet being replied to by @hpkmxvlaa is no longer available.)
Lol banger
Thursday, July 31, 2025
Reply to @daniel_271828, 07:50 UTC · 3 likes
zzDANIEL_ETH (@daniel_271828): People often lament how healthy food tends to not be particularly tasty, and vice versa, but isn’t this just a case of the tails come apart?
Yes
Tweet, 07:54 UTC · 2 likes
Why don’t the big LLM providers offer quantized versions of their models?
I mean the open source providers do, it’s weird that you can’t get a bit of a discount on Opus-8bit
Quote tweet, 23:21 UTC · 35 likes
App called Market for Lemons where you’re extremely up front about all the ways you suck
Quoting Dimitri (@thedimitri): The Tea app was inevitable because dating is a lemon market.
In a lemon market, there is a huge information asymmetry where the seller knows more than the buyer (think used cars). You can’t tell who’s a viable partner and who’s hiding serious issues, so everyone gets treated like they’re probably broken, which isn't a bad assumption. The best people get picked off early and what stays on the market looks more questionable over time.
Modern dating has only exacerbated this asymmetry. We used to meet through friends or family, which served as built-in screening, giving more information to the "buyer". Now it’s just strangers crafting personas without any shared context.
So of course someone made an app to rate men. But reviews from bitter exes won't really solve this problem. They don't reduce the asymmetry. If anything, they will add more noise.
The real fix is embedding yourself in a community where reputation is important rather than swiping and treating each new person like a blank slate.
End of July 2025. Previous month: June 2025. Next month: August 2025. Back to the archive index.