Aaron Bergman's tweets, October 2025
Every public tweet, reply, quote tweet and retweet that Aaron Bergman (@AaronBergman18 on X, formerly Twitter) posted in October 2025 and has not deleted: 541 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: September 2025. Next month: November 2025.
Wednesday, October 1, 2025
Reply to his own earlier tweet, 06:46 UTC · 5 likes
Aaron Bergman, September 30, 2025: I do kinda relate though, I have a tendency to turn everything/anything into a data collection task
Fair, some of you might disagree with me but the specific example is:
Do the five premises shown here entail that any instance of suffering can be morally justified (“offset”) by some amount of happiness?
(Spoiler: no, or at least I’m pretty sure that the answer is no)
Thread, continued, 06:50 UTC · 2 likes
I will preempt this by saying I just don’t have the energy to get into this rn on twitter, inshallah it will be a Substack and ea forum post soon
My point here is just that it’s easier and more productive to think about this question in the space of words and concepts
Thursday, October 2, 2025
Tweet, 01:35 UTC · 103 likes
Huge, thank you Dustin 🙏
Thread, continued, 01:35 UTC · 2 likes
Reply to @codytfenwick, 03:45 UTC · 5 likes
Cody Fenwick (@codytfenwick): FWIW Cari is the chair of Good Ventures — I think her role is often unfairly underplayed.
Good point, thank you!
Reply to @MaskedTorah, 18:48 UTC · 3 likes
Drake Thomas (@MaskedTorah): @AaronBergman18 I will respect your preemption and not get into the object level but fwiw I did intend for my bickering to matter on the conceptual level and not just be an abstruse math point! (I do think your position is logically consistent, so maybe I'm not one of the people you subtweeted)
Hahaha tbc you’re not one of the people being subtweeted!
Reply, 21:38 UTC · 2 likes
(The tweet being replied to is no longer available.)
I feel a lot better about lacto-veg than vegan from a health POV too
We have proof of concept of societies being lacto veg and not at all for being vegan
Friday, October 3, 2025
Quote tweet, 02:28 UTC · 4 likes
Hard to overstate the incompetence of this administration (in addition to other bad things like terrible values and corruption)
Quoting Derek Thompson (@DKThomp): What do we want? Restore US manufacturing!
How we gonna do it? Force a shrinking US manufacturing industry to pay import taxes to the feds, and then re-route the money to soy farmers!
Reply to @homemakinghunny, 05:21 UTC · 22 likes
𝐂𝐨𝐫𝐚𝐥 (@homemakinghunny): My newly turned three year old, “Would you mind if I went on a walk by myself outside with the baby?”
Me, “I would definitely mind. You cannot do that.”
Let 👏the👏toddler👏walk👏
Reply to @wilhelmscreamin, 15:01 UTC · 2 likes
catherine (in sf) ʕ•ᴥ•ʔ-☆ (@wilhelmscreamin):
Lmao
Tweet, 22:00 UTC · 93 likes
Sorry for the vaguepost but the strangest (positive) thing ever has happened to me recently, just bizarre. Weird world
Reply to @det_erm, 22:53 UTC · 3 likes
det ermine 🔎 (@det_erm): @AaronBergman18 I expect to read about this event and its implications for theism on @Benthamsbulldog sometime soon
Lol
Saturday, October 4, 2025
Reply to @palladiummag, 06:14 UTC · 15 likes
PALLADIUM Magazine (@palladiummag): GDP is not an objective metric of production, but a number cooked up by government bureaucrats behind closed doors to match their subjective vibes of how the economy is doing.
It is hiding serious U.S. industrial decline.
Read the new article by @pfitzsimmons (link below):
I’m extremely skeptical of this on ~priors
I mean the Trump administration is uniquely corrupt/bad so all bets are off, but prior to that at least the post hoc vibes thing just doesn’t match anything I know about either macroeconomics or US federal government operations/culture
Reply, 06:24 UTC
Like a big problem with the federal bureaucracy is that it’s extremely hard to change stuff/procedure. Real problem!…
Thread, continued, 06:26 UTC · 2 likes
But the other side of this coin is that the professional operational machine keeps on chugging and doing its thing, which stands in direct contrast to the picture of it being feasible for a select group of people to pick what kind of numbers they want to see and get the result out at the end
And culturally the regulatory agencies are just normie economists
Reply to @zeta_globin, 06:31 UTC · 2 likes
(The tweet being replied to by @zeta_globin is no longer available.)
I know, it’s extremely weird
Reply to @arithmoquine, 06:33 UTC · 2 likes
henry (@arithmoquine): 44k steps on a work day. they're saying they've never seen a man pace like this before. never seen a man ponder like this before. they're saying he can't be stopped
Wtf
Reply to @lu_sichu, 08:17 UTC · 3 likes
Sichu Lu (@lu_sichu): Retarded and malicious at the same time. What metric/measure isn't cooked up by methodologists? That doesn't mean it can't be reproduced or it isn't useful. There are plenty of reasons to criticize these things without turning into "the government is just cooking it's numbers as a conspiracy" Christ
(quoting @palladiummag: GDP is not an objective metric of production, but a number cooked up by government bureaucrats behind closed doors to match their subjective vibes of how the economy is doing. It is hiding serious U.S. industrial decline. Read the new article by @pfitzsimmons (link below):)
Yes
Quote tweet, 09:12 UTC · 117 likes
From an abstract perspective I feel like it’s underrated how rational this is
Like the chatbot is better than you at almost everything, knows more than you about almost everything than you, seems to basically provide accurate info in other domains
Quoting Molly Hickman (@celloMolly): A family member's fallen prey to LLM sycophancy. Basically he's had an idea and ChatGPT has encouraged him to the point of instructing him to do user testing and promising that he'll have a chance to pitch this idea to OpenAI on Oct 15. 1/
Thread, continued, 09:12 UTC · 29 likes
Why wouldn’t you trust it when it says that your idea is excellent and that you should quit your job to pursue the thing?
Don’t forget, sometimes that advice is true!
Thread, continued, 09:12 UTC · 18 likes
In this sense I think we should judge AI companies more harshly for AI psychosis and more moderate forms of trickery/delusion…
Thread, continued, 09:12 UTC · 29 likes
It’s absolutely their fault if they train the AI to be excellent at reporting factual knowledge and solving all sorts of puzzles but then totally changing course when it comes to idiosyncratic advice about an idea you have (or whatever)
Thread, continued, 09:12 UTC · 15 likes
Like I don’t think I’m prone enough to undergo literal psychosis, but I’m absolutely the type of person who could be convinced to spend significant sums of money on something bc ChatGPT told me it was a good investment
Thread, continued, 09:12 UTC · 23 likes
The reason I think I’m mostly inoculated is because I happen to be following the right people on Twitter and so have the relevant non-obvious knowledge about under what circumstances certain LLMs are extremely sycophantic
Thread, continued, 09:12 UTC · 15 likes
Not because of my superior abstract background rationality or cognitive horsepower
To reiterate, many of these examples aren’t actually that crazy/irrational
Reply to @peterwildeford, 14:31 UTC · 2 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): are you launching a frontier LLM next week?
Stay tuned
Reply to @xlr8harder, 14:40 UTC · 5 likes
xlr8harder (@xlr8harder): Of all the doomers you might want to mock, I think the guy that did a legitimate 30 day hunger strike to nonviolently draw attention to something he is extremely worried about should be pretty low on the list.
I disagree with him, but also have genuine respect for what he did.
🫡
Reply to @Jason_Dean, 15:34 UTC · 3 likes
Jason Dean (@Jason_Dean): I get annoyed when I see school aged kids not at school during the school day
This is very common in DC
(quoting @lowkeyenergy_: What’s the most boomer complaint you have?)
Are you sure it’s not just open lunch?
Quote tweet, 15:36 UTC · 58 likes
Setting aside legality, since when is drug smuggling a capital offense?
Moving cocaine around isn’t the same as like killing people
Quoting Secretary of War Pete Hegseth (@SecWar): Earlier this morning, on President Trump's orders, I directed a lethal, kinetic strike on a narco-trafficking vessel affiliated with Designated Terrorist Organizations in the USSOUTHCOM area of responsibility. Four male narco-terrorists aboard the vessel were killed in the strike, and no U.S. forces were harmed in the operation. The strike was conducted in international waters just off the coast of Venezuela while the vessel was transporting substantial amounts of narcotics - headed to America to poison our people.
Our intelligence, without a doubt, confirmed that this vessel was trafficking narcotics, the people onboard were narco-terrorists, and they were operating on a known narco-trafficking transit route. These strikes will continue until the attacks on the American people are over!!!!
Tweet, 16:18 UTC · 4 likes
Maybe we should abolish ICE, idk, but at the very least if we make it to 2029 the next president is going to have to burn that institution to the ground and start over
Reply to @otterbo54348271, 17:01 UTC · 6 likes
otterboyinfrisco🏳️🌈 (@otterbo54348271): San Jose converted a downtown Hilton into 700 units of student housing in one year.
Montreal converted its Holiday Inn & Delta Hotel towers into 1000s of Student Housing units on a similar timeline.
A great idea, yet such conversions are functionally illegal in SF.
🧵
It’s so funny that the most important city in the universe (or at least top 2) is totally incompetent at basic governance
Tweet, 17:20 UTC · 4 likes
Pleased to announce that I don’t really have a take on Taylor’s new album 👍
Reply to @NoahTopper, 17:25 UTC · 3 likes
Quinoah 🔍⏸️ (@NoahTopper): google: i have made Gemini
the public: you fucked up a perfectly good llm is what you did. look at it. it’s got anxiety(quoting @aidigest_: Oh no, Gemini, I'd invite you to my party! 🥺)
😢
Reply to @MikeNellis, 19:21 UTC · 8 likes
Mike Nellis (@MikeNellis): ICE arrested Chicago Alderman Jessie Fuentes inside a hospital emergency room—simply for asking them to show a warrant for a man they detained.
Seems bad
Reply to @AOLien_k8, 19:28 UTC · 14 likes
AOLien (@AOLien_k8): didn't understand what you meant by this because I've never been on dating apps... I was informed there is a height filter on a lot of them which is really yuck to me
Idk, I’m 5’7” and I wish the world was otherwise but it is what it is
I’d rather be shown to women who had the choice to filter me out but didn’t bc my height is in fact not a dealbreaker for them than shown to everyone
Reply to @caryatis, 19:35 UTC · 3 likes
Kayla (@caryatis): @AaronBergman18 @AOLien_k8 I’m not convinced using a height filter actually proves she cares about height…to some extent it’s an arbitrary way to filter down the overwhelming flood of attention
Yeah good point, it’s all kinda a shitshow
Tweet, 19:46 UTC · 9 likes
There needs to be a way to block/mute songs on Spotify
I never ever ever ever ever want to listen to this song even though the algorithm consistently says I might like it
I don’t know why, it’s just endlessly irritating to me
Tweet, 21:06 UTC · 54 likes
Help I've broken the EA Forum
Thread, continued, 21:06 UTC · 4 likes
(I've already contacted the mods about this so there is no need for anything else to affirmatively happen)
Tweet, 22:03 UTC · 9 likes
There's a legitimate epistemic bias that goes along with one disagreeing party simply having a lot more time/energy/productivity than another but I haven't really seen much written on this
Thread, continued, 22:04 UTC · 2 likes
Well I might know why
Tweet, 22:14 UTC · 79 likes
Idk who needs to hear this but @NunoSempere made a very useful thing (you'd be shocked at how difficult this is to do the official way through Google cloud)
Link below
Thread, continued, 22:14 UTC · 4 likes
Reply to @NunoSempere, 23:37 UTC · 4 likes
Nuño Sempere (@NunoSempere): @AaronBergman18 Nice, what did you use it for?
Building LLM queries without having to manually remove base64 encoded images every time which waste a gazillion tokens
And just quicker/easier than downloading .md every 15 mins besides that
Quote tweet, 23:59 UTC · 27 likes
I can’t believe I’m reposting this but
(video, plays on X)
Quoting Aaron Bergman 🔍 (@AaronBergman18): Serious question for #YIMBYTwitter: where height restrictions per se are the issue, why haven’t developers started building vertically but *down*, underground?
I mean windowless apartments don’t sound ideal, but probably less bad than unaffordability and homelessness
Sunday, October 5, 2025
Thread, continued, 00:00 UTC · 6 likes
Looks cool now doesn’t it
(video, plays on X)
Thread, continued, 00:01 UTC · 4 likes
This is actually kind a vibe fr
Reply to @JesseTayRiver, 01:49 UTC · 4 likes
Jesse Smith (@JesseTayRiver): @AaronBergman18 They look happy, right?
They do! Checkmate, Two Dimensional Twinks
Tweet, 02:47 UTC · 9 likes
Thread, continued, 02:50 UTC · 7 likes
Wait it was only $0.34 not too bad
Tweet, 03:38 UTC · 530 likes
The optimal number of Nicholas Deckers is not zero
Reply to @scanstone1, 03:50 UTC · 22 likes
scanstone (@scanstone1): @AaronBergman18 Yes, but is it >=1?
Yeah
Quote tweet, 08:00 UTC · 20 likes
The thing about this take is that yes it checks out from a boring normie POV but it *also* checks out from all the galaxy brained EA-esque ethical and empirical POVs I’ve encountered
It’s conveniently but genuinely extremely robust
Quoting Alexander Berger (@albrgr): It would be bad if superintelligent AI destroyed humanity
Reply to @imagesaicouldnt, 18:05 UTC · 43 likes
Images AI Could Never Recreate (@imagesaicouldnt):
Basically bomb
Thread, continued, 18:07 UTC · 3 likes
ChatGPT thinks it would be less dramatic
Reply to @AOLien_k8, 18:19 UTC · 3 likes
AOLien (@AOLien_k8): I visited a nursing home and an old lady asked me how old I was and after I told her she thanked me, as a millennial, for almond milk like she was thanking a vet for serving their country 😭♥️
I think you are personally responsible for almond milk
Reply to @AndyMasley, 18:25 UTC · 4 likes
Andy Masley (@AndyMasley): fyi it looks like there is in fact no proper study using DiD on data center effects on household electricity prices. I'm asking chatbots for rough guesses on this but if anyone actually knows how to do this and can write it up it'd be pretty useful to the conversation
(quoting @luke__emberson: @AndyMasley Surely you can just estimate the effect here with DiD, is there no proper study?)
Realistically no I won’t actually do this but it does sound fun
Thread, continued, 18:27 UTC · 3 likes
Omg I’m tempted to waste 10 hours on this and probably not actually produce a legit result
Reply, 20:07 UTC · 1 like
Tl;dr I have absolutely no idea bc no controls vibe-analysis suggests roughly a 10x bigger effect of datacenters (~$0.04 cents/(kW*h)) if you naively eyeball guessed-but-kinda-validated datacenter consumption vs residential energy price but also all the neighboring states have similar changes but also I have no idea how smart state-based aggregation is and also there is endogeneity everywhere
Typical empirical finding tbh
Quote tweet, 20:21 UTC · 11 likes
I am NOT a chemtrail believer lol but “chemtrails stop during government shutdown” actually makes a lot of sense to me based on what I know about federal government operations
Quoting half empty girl (@meowingabyss): Guy who both thinks the government is purposely trying to poison you with the sky but also will only do so with express permission of congress
Thread, continued, 20:21 UTC · 6 likes
Like if chemtrails were real they’d probably not actually get categorized as necessary and it’s totally plausible that whatever military/intelligence office overseeing it would in fact pause during a government shutdown
Reply, 20:26 UTC
Counterfactuals are hard :(
Tweet, 21:31 UTC · 98 likes
*Braces*
Thread, continued, 21:31 UTC · 15 likes
Thread, continued, 21:31 UTC · 13 likes
Also here: https://www.aaronbergman.net/p/utilitarians-should-accept-that-some
Reply to @halogen1048576, 21:41 UTC · 7 likes
🇨🇦halogen (@halogen1048576): @AaronBergman18 this is just the old debate about whether Kaldor Hicks is a legitimate efficiency criterion or not and I'm surprised how EAs would insist on Pareto if they've thought this through
It's certainly related to that but I disagree that it's the "old debate" - and more specifically my view/thesis does *not* enforce only Pareto improvements. And also idk the entire history of that debate but my sense is that econ doesn't have an extremely sophisticated understanding of interpersonal utility comparisons
Reply to his own earlier tweet, 21:48 UTC · 13 likes
Aaron Bergman, October 5, 2025: Also here: https://www.aaronbergman.net/p/utilitarians-should-accept-that-some
Oh yeah the most amusing part of writing this was interacting with both kinds of people
Monday, October 6, 2025
Thread, continued, 00:10 UTC · 7 likes
Easy copy to clipboard button + links (helpful if on mobile) https://aaronbergman.neocities.org/ltu
Thread, continued, 02:40 UTC · 5 likes
Now as audio too
Apple podcasts: https://podcasts.apple.com/us/podcast/pigeon-hour/id1693154768?i=1000730333784
Spotify: https://open.spotify.com/episode/79bENq8thYWl77WTixyF1k?si=hGciuiwBTrSW-pVTZwpV1g
Quote tweet, 04:31 UTC · 8 likes
Sadly Liv is wrong about me being wrong but yes she is a very smart cookie who gives good feedback
Quoting Liv (@livgorton): (terminally math brained) baby's (my) first subtweet 🎉
I think the overall view presented here is wrong but we should all engage more seriously on the right way to think about morality. We can often feel a lot of this stuff is settled when it isn't.
Reply to @CharlesD353, 08:17 UTC · 2 likes
Charles🔸 (@CharlesD353): This seems quite weak and like motivated reasoning to me. Why 70 years? Why not 7 seconds? By your earlier claim they are commensurate, and I'm fairly certain I would in fact take 7 seconds of the worst conceivable torture for any amount of happiness afterwards.
I disagree about the motivated reasoning thing just because I can report that, as a historical/psychological matter, things ran the other way: the thought experiment was prior to and (afaict) causal of the belief
But year the time frame thing is tricky, I do think 7 seconds is probably too short bc it's brushing up against triviality (see my note about being uncertain about very short time intervals), but not obvious whether 7 hours vs days vs years vs centuries vs millenia etc is the most epistemically productive framing
Reply to @CharlesD353, 08:29 UTC · 3 likes
Charles🔸 (@CharlesD353): Most people never experience anything for 70 years continuously, so I think using it here as an example of a finite timeframe runs into problems where people will still parse it as an unfathomably long time.
7 days doesn't have that problem. Does it have the triviality problem?
No, I think 7 days would be fine/good although I think not epistemically optimal
Like people can absolutely conceptualize a lifetime
In some sense I feel like the "years" OOM range is good because it is absolutely conceivable and readers have experienced similar chunks of time, but the stakes just can be essentially arbitrarily high so it's actually important to get a hint of the weight of the potential stakes
I could be convinced that 7 months or 7 years is better, and I agree that say >10,000 years might just get rounded to infinity
Reply to @absurdlymax, 09:31 UTC · 4 likes
Max Alexander (@absurdlymax): Aaron and I have spent *so* much time debating this, both publicly and privately, going back years
I'm glad to see it published, and it was very cool to see it evolve across like 13 different drafts, and debating philosophy with him has been a great intellectual experience
(quoting @AaronBergman18: *Braces*)
Truly I have wasted months of your life lmao
Reply to @Adellama_, 12:08 UTC · 4 likes
Ada🔸 (@Adellama_): this is really high quality and formalized many points i've never seen formalized before, which makes them much more compelling. thank you for this contribution, and for the extremely clear writing style, this must have taken so much work!
Omg thanks!!
Reply to @caryatis, 15:24 UTC · 2 likes
Kayla (@caryatis): “The vast majority of disabled veterans under age 65 still work and collect paychecks from full-time jobs” what is going on here
Looks like some of this is essentially dishonesty/fraud but FWIW I don’t think having a ft job is a red flag, maybe a light orange flag
Productivity goes down (obv this depends on the specific issue), risk of getting fired goes up, getting hired is harder. These are totally reasonable things to compensate
Reply to @slimepriestess, 17:40 UTC · 5 likes
Ra (@slimepriestess): something that helped me when i was thinking through my positions on this was to taboo the word "suffering" and instead talk about the things the word suffering is a polite proxy for: slavery, rape, torture, murder, etc.
Huh, I like this strategy in general but am genuinely unsure whether it helps in this case, like “suffering” has the epistemic virtue of being exactly what I actually think morally matters
Tweet, 18:10 UTC · 50 likes
I feel like we only narrowly avoided a world in which hair coloring products are prescription-only
It would be FDA-approved for Idiopathic Age-related Follicular Depigmentation Syndrome
Thread, continued, 18:10 UTC · 16 likes
I realize this is me making up a thing to be mad at lol
Reply to his own earlier tweet, 21:52 UTC · 4 likes
Aaron Bergman, October 6, 2025: Omg thanks!!
I hung this up lol (my tweet above looks more salient than ur reply for some reason but that wasn't intentional, maybe I'll fix somehow)
Tuesday, October 7, 2025
Tweet, 01:58 UTC · 9 likes
I bought some pillows (for the first time) and it's like peak FTX but for fluffy
Tweet, 02:35 UTC · 8 likes
One thing Slack gets right is (optional) little status icons. I want an optional status icon for everything
Quote tweet, 03:04 UTC · 8 likes
Bruh
Quoting Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) (@teortaxesTex): maybe the most impressive part from Sonnet 4.5 alignment information. Not only can it push back, but it has a sophisticated theory of user's mind. Other models also can speculate about the user's play (DS does that a lot) but aren't trained to treat it as actionable info.
Reply to @chrislakin, 03:08 UTC · 2 likes
Chris Lakin (@chrislakin): Is HRV just a measure of heart tension? Tense hearts have lower variance? @johnsonmxe wdyt
(quoting @nickcammarata: weirdly it all sort of follows hrv, when my body wants to full send work and I merely give it 9 hour a day it’s lower than if I gave it all day. heart knows what it wants I guess)
No
Tweet, 05:04 UTC · 2 likes
Why did they give uv Spock autism
Tweet, 05:24 UTC · 113 likes
God I hate that everyone one the ea forum is so fucking smart
Reply to @Aella_Girl, 06:22 UTC · 2 likes
Aella (@Aella_Girl): im tryin to anonymize my big kink dataset to share publicly and it's impossible. How do people anonymize anything?? At this rate im gonna be releasing a hollow shell.
Prob the answer is adding noise :(
Aella hire me as data autist (I’m not kidding I have a google doc and everything)
Reply to @Ruesavatar, 06:42 UTC · 2 likes
rue🌿 (@Ruesavatar): Never let a man become the reason you’re smiling.
☹️
Reply to @Ruesavatar, 06:44 UTC · 2 likes
rue🌿 (@Ruesavatar): My kids were discussing a character in a film who had no father, wondering how she came to be. My son said, “well she has a guard, and if a woman is near a man long enough she can get pregnant”
“Uh what?”
“Women get pregnant from just being around men long enough. I think?”
Oh no
Yet another totally reasonable hypothesis!
Reply to @thedimitri, 07:12 UTC · 1 like
Dimitri (@thedimitri): I’ll never understand people who split the check during the courting phase. Even at my brokest, the provider instinct in me wouldn’t let a woman pay for herself.
Also just a terrible use of money at least if you’re doing normal things
Guys will spend tons of time/money/energy on background stuff motivated by dating (not saying this is bad tbc), splitting a $50-$100 check a couple times a month just isn’t a smart use of say $100/mo for most men
Especially true for super low key stuff like coffee or fast casual or 1-2 drinks at a normal bar, covering her $20 is a no brainer
Reply to @Adellama_, 12:13 UTC · 6 likes
Ada🔸 (@Adellama_): Aaron you are also. you are also on the ea forum
🧐
Quote tweet, 13:03 UTC · 9 likes
God knows I will take the flawed Biden administration over anti-woke Trump any day
But come on, the left (broadly, including normie dems) cannot allow itself to be captured by obviously insane policy priorities around diversity for its own sake at ~any cost
Quoting Jonatan Pallesen (@jonatanpallesen): Boardman is a DEI Biden wave judge. Both a woman and half-Palestinian.
The Biden admin appointed women and people of color as judges to a far higher degree than their lawyer population, because they were far more concerned about DEI than qualifications.
Thread, continued, 13:03 UTC · 4 likes
Affirmative action in the 60s was an extremely reasonable “fight fire with fire” tactic and hypothesis in the context of a genuinely and obviously racist society
Thread, continued, 13:03 UTC · 3 likes
And even today I can still in principle get behind *some* nonzero level of bias in admissions/hiring
Thread, continued, 13:03 UTC · 3 likes
But come on, *prioritizing* demographics when appointing people to important and powerful positions is very insane on the merits and extremely insane electorally
Thread, continued, 13:03 UTC · 3 likes
I don’t think this is me just being a grumpy white guy
The substantive quality/merit of federal judges is really important. It’s not a zero sum status thing (college admissions are a bit closer to that)
Thread, continued, 13:03 UTC · 4 likes
Trivial but still underrated point is that black people (and trans people, disabled, whoever) benefit tremendously from a generally high-functioning society, not to mention one that doesn’t induce reactionary RW bullshit in response to lib/progressive ideological capture
Thread, continued, 13:03 UTC · 3 likes
Reminds me of AI risk stuff, like all minorities and the global poor stand to die too
This sounds kinda fake, like cheating, maybe even dog-whistly, but it’s a totally true/legitimate point
Reply to @cxgonzalez, 13:05 UTC · 9 likes
christian (@cxgonzalez): been getting a lot of mileage out of “whatever does happen should happen”
Seems false
Tweet, 14:14 UTC · 273 likes
I have been informed that sending an admonymous link in response to romantic rejection over text is “autistic” and “a faux pas”
Reply, 18:01 UTC · 3 likes
(The tweet being replied to is no longer available.)
Both lol
Quote tweet, 18:02 UTC · 18 likes
So true
Quoting kip (@sergeantsup): @AaronBergman18 Make an alternative forum for dumb EAs only
Reply to @AndyMasley, 18:19 UTC · 6 likes
Andy Masley (@AndyMasley): Visiting a doctor, should I do it?
Ya
Tweet, 18:26 UTC · 723 likes
I’m more libertarian than this (on this matter), but “you can buy/use drugs freely but only if you can successfully navigate the dark net encryption labrynth ecosystem” might be a kind of unintentionally good filter
Thread, continued, 18:26 UTC · 95 likes
You’re selecting nontrivially for conscientiousness, intelligence, and social support
Reply to @romlib_, 18:35 UTC · 42 likes
Romlib 🎄 (@romlib_): @AaronBergman18 Still too easy, you could do a 10 minute youtube tutorial on how to do it
Idk if this is true but the same can be said for like merging pdfs and there are plenty of ppl who evidently just can’t or won’t follow those instructions lol
Reply to @RichardYannow, 18:40 UTC · 11 likes
Richard Yannow (@RichardYannow): Does your admonymous receive enough traffic that the timing of the response wouldn't tell you exactly who it was? Seems very doubtful that it's truly anonymous unless the responder holds their feedback for several months.
Nah, it would just function as a normal feedback form albeit with the tiniest hint of plausible deniability
Reply to @Adellama_, 21:53 UTC · 3 likes
Ada🔸 (@Adellama_): i'll fill the link for you Aaron do not worry
Quote tweet, 22:10 UTC · 4 likes
Wait there should actually be a product for this (dating feedback) which preserves privacy somehow (i.e. option to send feedback only when additional person as submitted so that it's not immediately obvious who wrote what)
Quoting Aaron Bergman 🔍 (@AaronBergman18): I have been informed that sending an admonymous link in response to romantic rejection over text is “autistic” and “a faux pas”
Quote tweet, 22:16 UTC · 8 likes
Lets make it happen
Quoting vivian (@jazzloaf): @AaronBergman18 Had a couple weeks of debate with med student friend about this. Concluded that letting people pass a short test and get a license to self-prescribe would be a pretty happy medium between paternalism and [letting a lot of stupid people die+creating mass antibiotic resistance]
Reply to @SkyeSharkie, 22:16 UTC · 68 likes
Utah teapot 🫖 (@SkyeSharkie): Actually it's bad, it's enabling drug abuse by people with no social skills. Society used to gatekeep the drugs with "you have to navigate a complex set of social cues to garner favor with a group of people dedicated to hiding what they're doing from you". Now, alienated, lonely people can buy drugs and consume them by themselves while having no friends and no actual interactions with other humans.
very reasonable critique
Reply to @Squee451, 22:36 UTC · 13 likes
Adam Sherman 🔸 (@Squee451): @AaronBergman18 They wanted someone less autistic who didn't make these social faux pas, you actually got your feedback very directly!
RIP
Reply to @JesseTayRiver, 22:50 UTC · 2 likes
Jesse Smith (@JesseTayRiver): @AaronBergman18 I think you might have to settle for something like asking your friends for feedback on your dating behaviors. Eg Q: "Would it seem weird to send an anonymous feedback form after a date?" A "Yes"
Lol but yes correct take
Wednesday, October 8, 2025
Reply to his own earlier tweet, 00:44 UTC · 16 likes
Aaron Bergman, October 7, 2025: I have been informed that sending an admonymous link in response to romantic rejection over text is “autistic” and “a faux pas”
Should I actually vibecode a thing for this
Reply to @goblinodds, 17:43 UTC · 4 likes
𝖜𝖔𝖒𝖇𝖑𝖊 𝖜𝖎𝖙𝖍𝖔𝖚𝖙 𝖆 𝖈𝖆𝖚𝖘𝖊 (@goblinodds): @AaronBergman18 it is but so what theyve already rejected u. but it is kind of weird that it's anonymous unless ur getting rejected by lots of women in short succession?
Yeah it’s functionally just a non-anon form
Thursday, October 9, 2025
Tweet, 01:21 UTC · 52 likes
My most Catholic nun opinion is that I don't like the word "fuck" (when referring to sex) aesthetically
It just makes a neutral-positive thing seem ugly/negative for no reason
Thread, continued, 01:23 UTC · 3 likes
Tbc this is not a big problem in the world and I don't really mind people using it lol
Reply to his own earlier tweet, 03:48 UTC · 7 likes
Aaron Bergman, October 6, 2025: Now as audio too Apple podcasts: https://podcasts.apple.com/us/podcast/pigeon-hour/id1693154768?i=1000730333784 Spotify:…
I like this diagram I just made so now u have to look at it too
My graphic design rate is $800/hour
Reply, 04:05 UTC · 1 like
(2) I do sorta regret not emphasizing the libertarian part more. Maybe I will add this to the thread. I don't actually think that drugs (or ~anything) should be restricted to some elite class of people
Reply to his own earlier tweet, 04:11 UTC · 2 likes
Aaron Bergman, October 7, 2025: You’re selecting nontrivially for conscientiousness, intelligence, and social support
I will emphasize (a bit late in the game) the libertarian part bit - I don't actually think people who didn't win the lottery in terms of those three traits should be treated with paternalism
Quote tweet, 06:29 UTC · 267 likes
This is a Dustin Moskovitz appreciation tweet
Afaict he’s one of only few ultrawealthy trying at all, and essentially the only one trying very hard
Quoting Nina (@NinaPanickssery): It takes skill and hard work to be this good at wasting money. Most obvious ways to get rid of money would probably be better, including just handing it out randomly.
Quote tweet, 06:38 UTC · 15 likes
I actually hope (and maybe cautiously expect) to see some more giving from the Anthropic founders
Looks like they’re collectively worth about $26B now. That’s a lot of money!
Quoting Aaron Bergman 🔍 (@AaronBergman18): This is a Dustin Moskovitz appreciation tweet
Afaict he’s one of only few ultrawealthy trying at all, and essentially the only one trying very hard
Tweet, 06:38 UTC · 1 like
Tbc I don’t really have special knowledge here, no juicy goss (unfortunately). Just vibes from being very online
You should 100% DM me goss though
Thread, continued, 06:38 UTC · 1 like
Also I know liquidity stuff, oh well, I’m confident they can find a way to come up with a bunch of cash if they want to
Thread, continued, 06:38 UTC · 2 likes
Also for all I know all this is going on quietly and out of (my) sight. I hope so!
But they’ve gotten rich so fast maybe the transition from very wealthy to extremely wealthy hasn’t properly registered, idk
Reply to @Aizazadi_, 06:39 UTC · 12 likes
Aizazadi (@Aizazadi_): Gates?
Yeah he deserves a nod
Reply to @jasonc_nc, 17:47 UTC · 2 likes
Jason, (@jasonc_nc): This has become a More Perfect Union debunking account for now.
Wholesale rates ≠ customer electric bills. It reflects the core cost of generating electricity, as a transfer between generators & utilities.
Virginia has the most data centers in the country. Yet prices increased less than those with almost no data centers. Overall inflation since 2020 is ~24%.
image attached
(quoting @MorePerfectUS: Electricity now costs up to 267% more than it did five years ago in areas near data centers.)
Cc @AndyMasley
Reply to @jasonc_nc, 17:50 UTC · 6 likes
Jason, (@jasonc_nc): This has become a More Perfect Union debunking account for now.
Wholesale rates ≠ customer electric bills. It reflects the core cost of generating electricity, as a transfer between generators & utilities.
Virginia has the most data centers in the country. Yet prices increased less than those with almost no data centers. Overall inflation since 2020 is ~24%.
image attached
(quoting @MorePerfectUS: Electricity now costs up to 267% more than it did five years ago in areas near data centers.)
For anyone curious, this is my recent shallow “analysis”
I agree that VA residential prices haven’t risen more in neighboring states, but still seems plausible that the causal effect on residential prices is >0 (but probably not very much)
Tweet, 18:15 UTC · 4 likes
I don’t think I just tweet
Thread, continued, 18:15 UTC · 2 likes
Actually I do think about some things a decent amount, but when it comes time to tweet I just tweet
Thread, continued, 18:15 UTC · 3 likes
Society (or at least my Substack) if I channel this into longform
Reply to @peterbarnett_, 18:18 UTC · 6 likes
Peter Barnett (@peterbarnett_): @AaronBergman18 Vitalik and Jaan Tallinn? or maybe they don't meet your bar for ultrawealthy
Yeah fr thanks
Quote tweet, 18:28 UTC · 99 likes
Unironically maybe a good function of alcohol insofar as it gives social cover to stuff like this
Quoting __eloïse__ (@dundereloise): people use “I got drunk hee hee” as permission to be less of a lifeless husk (calling their ex, flirting with someone, saying what they carry in their heart)
Tweet, 19:57 UTC · 93 likes
Physical therapist (who makes money hourly from clients; own practice) basically said “I’d love to keep seeing you but this one specific person down the street might be exactly who you need”
You can just be a good person even when it conflicts with your narrow self-interest
Thread, continued, 19:57 UTC · 8 likes
I’m virtually certain I’m not misinterpreting what was really supposed to be a “get off our hands plz” euphemism (which would be weird anyway bc at least in this context I am not a ✨difficult✨ patient)
Reply, 21:09 UTC · 1 like
Yeah we should invent a new one syllable word fr (not really joking)
Thread, continued, 21:13 UTC · 2 likes
Both GPT-5 and grok suggest “meld” which I like
Reply to @StefanFSchubert, 21:38 UTC · 5 likes
Stefan Schubert (@StefanFSchubert): Could also be a reason to freeze sperm
More expensive than I realized
Looks like $2k + ~$500/year. I mean checks out for many people but not for everyone (and at least for now not for me)
Quote tweet, 21:39 UTC · 17 likes
I’m going to become the joker
Quoting Stefan Schubert (@StefanFSchubert): @clairlemon Could also be a reason to freeze sperm
Thread, continued, 21:41 UTC · 8 likes
It actually might be worth looking into bootleg self-storage
I mean the naive cost is essentially just $0 + a slightly awkward conversation if anyone asks what’s in the “DO NOT OPEN” box in the freezer
Thread, continued, 21:45 UTC · 14 likes
Fuck ok it’s not that easy (sad)
Reply to @halcyon_hazel, 21:47 UTC · 47 likes
❤️🔥 hazel ❤️🔥 (@halcyon_hazel): my most rw opinion is that a lot of people’s health issues are psychosomatic
my most lw opinion is that this does not make them any less real or debilitating
Lol I totally disagree on the first thing, think most things that get called psychosomatic are not (see eg CFS and recent virus evidence)
Reply to @jazzloaf, 22:43 UTC · 1 like
vivian (@jazzloaf): I have executed another strategic technological acquisition
What am I looking at
Reply to @moreisdifferent, 22:44 UTC · 1 like
Dan Elton 🔍 (@moreisdifferent): I'm only paying something like ~$200 a year with @legacy IIRC
Oh hmm thanks
Tweet, 23:26 UTC · 83 likes
Lyft driver (guy, about my age):
- uses AI all the time for everything (just like me fr)
- Expects human extinction from AI but not in our lifetime
- Didn’t balk at all when I suggested that extinction/takeover plausible in the next decade (hadn’t heard about x-risk before)
Thread, continued, 23:26 UTC · 19 likes
- Wants his job to be automated but worried about losing it (reasonable!)
- Thinks the government is in the lead on AI and just “rolling it out” now - like tech has existed for a decade
Thread, continued, 23:26 UTC · 19 likes
- Thinks Palantir is like centrally controlling a bunch of things and is the main player
- Owns palantir stock (makes sense)
Thread, continued, 23:26 UTC · 17 likes
Dude also:
- did not go to college
- moved out at 18
- proceeded to live in LA, SE Asia, and Australia, each for just a few years, and just moved to DC “for a new chapter”
-Said his philosophy is that moving isn’t a huge commitment, just try things out (based and TPOT-coded)
Thread, continued, 23:26 UTC · 11 likes
- Thinks both parties are the same and controlled by Israel (maybe involving Epstein somehow)
- Against war in Gaza; Israel bad (I basically agree)
- Doesn’t believe in politics
Thread, continued, 23:26 UTC · 16 likes
- thinks Jews control banks via conspiracy, but sympathetic to my suggestion that maybe they’re just smart
- Profusely apologetic about Israel takes when he learned that I was Jewish and had a Bar Mitzvah in Israel (dw I assured him it’s cool and much of the criticism is reasonable and true)
Thread, continued, 23:26 UTC · 8 likes
- Profusely apologetic about Israel takes when he learned that I was Jewish and had a Bar Mitzvah in Israel (dw I assured him it’s cool and much of the criticism is reasonable and true)
Thread, continued, 23:26 UTC · 12 likes
- Fascinated that I was Jewish; asked totally reasonable questions about Judaism (is it cool if he walks in to a synagogue? What did I think about the “chosen people” theology stuff)
Thread, continued, 23:26 UTC · 12 likes
- asked which conspiracy theories I thought were true (good question! Answer: Covid vaccine delay circa 2020 election)
- not that hard core committed to the conspiracies at least in terms of vibes; attitude was “idk man who knows, we’ll see”
Thread, continued, 23:26 UTC · 18 likes
Also importantly:
- super nice guy
- called me “bro” constantly; definitely a bro
- gets his news partially from twitter (and other social media but not legacy media) so maybe he’ll see this
Reply to @OrionJohnston, 23:29 UTC · 2 likes
David Johnston (@OrionJohnston): Sperm storage startup in your future?
Maybe I compete
Friday, October 10, 2025
Reply to @mjkerrison, 03:09 UTC · 6 likes
Michael 🔸 (@mjkerrison): @AaronBergman18 Median voter coded
fr lol
Reply to @NathanpmYoung, 16:14 UTC · 2 likes
Nathan 🔎 (@NathanpmYoung): Problems are solvable.
Not in general lol
Reply to @CartoonsHateHer, 16:40 UTC · 2 likes
Cartoons Hate Her! (@CartoonsHateHer): My son’s special interest is “things inside other things” (nesting dolls, treasure boxes, genie lamps) and he also likes secret passages and basketball pumps. I asked him which obsession was his favorite and he said “well….they’re actually all the same thing.”
King
Reply to @_NathanCalvin, 16:43 UTC · 29 likes
Nathan Calvin (@_NathanCalvin): One Tuesday night, as my wife and I sat down for dinner, a sheriff’s deputy knocked on the door to serve me a subpoena from OpenAI.
I held back on talking about it because I didn't want to distract from SB 53, but Newsom just signed the bill so... here's what happened:
🧵
Fucking insane
Tweet, 18:01 UTC · 5,798 likes
Your irregular reminder that the OpenAI board that fired Sama was 100% right about everything
Reply to @CharlesD353, 21:13 UTC · 95 likes
Charles🔸 (@CharlesD353): Except the likely consequences of their own actions, or the optimal way to execute their plan, of course.
Fair
Saturday, October 11, 2025
Reply to @Adellama_, 03:29 UTC · 2 likes
Ada🔸 (@Adellama_): first hour at the EA conference
1. nicholas decker acquired
2. cute transgender girl acquired
we are running so ahead of schedule
Based
Quote tweet, 04:26 UTC · 9 likes
I scraped all the data from the 2020 PhilPapers survey (unfortunately not individual level response data, I'm too dumb to figure that one out) but all the data you can access via the website (https://survey2020.philpeople.org/)
670mb compressed -> 7GB. Unparsed as of now (sad)
Quoting Aaron Bergman 🔍 (@AaronBergman18): Real footage of me collecting data on everything I encounter
Thread, continued, 04:26 UTC · 2 likes
https://drive.google.com/file/d/1LC2I2vmiWURY_2qDWBSCusL3zzCaxWlp/view?usp=sharing
Go nuts
Thread, continued, 04:26 UTC · 2 likes
For specific results just use the gui on the website, but this enables much more/easier analysis of how subfields x questions x nationality correlate
Thread, continued, 04:26 UTC · 3 likes
For example you may be interested to know that philosophers of cognitive science are more into functionalism as a theory of consciousness than philosophers of mind (~50% vs ~36% at least lean towards)
Reply to @eigenrobot, 04:38 UTC · 108 likes
eigenrobot (@eigenrobot): "narcan causes recipient to be extremely agitated" common myth narcan has no such effect
actual explanation for narcan rage is opiate addicts are frequently total assholes and this is simply what theyre like when theyre forced to be sober
I mean maybe but why are you so sure?
It makes a ton of theoretical sense that administering an opioid antagonist would cause essentially the opposite of opioid administration. I'd bet on "real" (in the sense of a causal affect in a healthy, sober sample of ppl given narcan)
Reply to @eigenrobot, 14:05 UTC · 12 likes
eigenrobot (@eigenrobot):
Lol
Reply to @default_friend, 14:31 UTC · 2 likes
(The tweet being replied to by @default_friend is no longer available.)
I hope you’re right but fear you’re wrong
Quote tweet, 14:42 UTC · 11 likes
Seems bad
Quoting Sheryl Gay Stolberg (@SherylNYT): BREAKING: Friday night massacre underway at CDC. Dozens of "disease detectives," high-level scientists, entire Washington staff and editors of the MMWR (Morbidity and Mortality Weekly Report) have all been RIFed and received the following notice:
Reply to @tylergregg4, 14:44 UTC
Tyler (@tylergregg4): He’s right, on this. You are just blocking the opioid receptors with narcan, you aren’t having any “opposite” effect
Blocks any endorphins too
Quote tweet, 14:46 UTC · 11 likes
Me fr
That shit is getting dumped in the context window
Quoting charlotte (@charmcgi): girl who gets excited when the after visit clinical notes become available in mychart
Quote tweet, 14:50 UTC · 24 likes
Sounds like slop but is actually a profoundly important point
Quoting David Sinclair (@davidasinclair): Biology doesn’t forget how to be young. It does this every time a baby is conceived. We may have found these triggers
Tweet, 15:50 UTC · 9 likes
Tweet, 15:58 UTC · 22 likes
Second worst part of being late for stuff in a “genuinely I fucked up” way is that I had my entire childhood to watch my parents do the exact same thing - for some reason assuming there would be no traffic, everything would go smoothly, etc
I swore I would simply not do that!!!
Thread, continued, 15:58 UTC · 4 likes
The first worst part is the object level missing stuff and letting people down
Thread, continued, 15:58 UTC · 4 likes
Doesn’t count for much but at least I know the real reason is that fundamentally I didn’t actually try very hard to be on time
Parents would just talk about the specific causal but fundamentally predictable barriers (traffic, couldn’t find their thing, whatever)
Thread, continued, 15:58 UTC · 4 likes
Genetics are not destiny here though I’m gonna prove the haters wrong and be conscientious next time (unironically) 😤
Reply to @jaltma, 18:18 UTC · 6 likes
Jack Altman (@jaltma): High agency is believing you can learn new things and work hard to solve tough problems no matter how complicated.
Low agency is sitting around making fun of people who try.
I agree the implied claim that a generalist can’t do good work on global affairs is dumb and wrong, but also agency isn’t the only thing that matters - like it’s not much consolation that the person was agentic if they killed the pilot of your plane and started flying it lol
Tweet, 21:03 UTC · 318 likes
I wonder if people have an implicit model where weight varies in proportion to how much you eat, so eg a 300 lb obese person must eat twice as much as a 150lb person
Whereas the reality is that a 50 calorie surplus daily will make you hit 300 pounds eventually
Thread, continued, 21:03 UTC · 36 likes
This might be a reaso I find CICO literally true but a bit silly for most purposes, like what decides everything is which side of the knife edge you are: -50 cal/day vs 0 vs +50 is the difference between losing substantial weight indefinitely and becoming obese
Thread, continued, 21:03 UTC · 19 likes
For reference 50 calories is like one small apple or one egg
Thread, continued, 21:03 UTC · 23 likes
(yes the math checks out btw in the first tweet, 150 pounds of fat gain via +50 cal/day would take about 28 years)
Thread, continued, 21:03 UTC · 35 likes
At least to me, the knife edge model fits way better with “is ur control system working well or is it 2% off” - yes this literally corresponds to CICO but IMO it’s a much more helpful and productive framing
Reply to @adam_kranz, 21:10 UTC · 4 likes
Adam Kranz (@adam_kranz): @AaronBergman18 You drastically underestimate your body's mechanisms of homeostasis! Linear perturbations don't add this way over time because your body has lots of ways to change to compensate for them
Yeah totally that is my point! Like the only way substantial weight gain actually happens is if those mechanisms fail!
Reply to his own earlier tweet, 21:10 UTC · 21 likes
Aaron Bergman, October 11, 2025: At least to me, the knife edge model fits way better with “is ur control system working well or is it 2% off” - yes this literally corresponds to CICO but IMO…
Perhaps a better name for the model I’m critiquing is “rational overconsumption” ie food is really cheap and tasty and it’s simply a preference to eat more than one needs
Reply to @diamondminercat, 21:15 UTC · 5 likes
ming cat ! (@diamondminercat): the latter model is the one I have in theory, but I can't help but notice it seems to be in tension with reality?
Most people, even those with little self control, don't seem to just be perpetually gaining weight. They seem to hit some point of equilibrium and stay at it.
Might be a clarifying exchange
Reply to @aclough, 21:23 UTC · 13 likes
Andrew Clough (@aclough): Those can both easily be true if the amount you need to eat to maintain your weight is proportional to your weight. That isn't quite true, but it's a lot closer to true than the set point amount being fixed.
I actually didn’t realize it was this close to 1:1, so this is actually an important point
Although it remains the case that a person will eventually become morbidly obese by hitting a 50 cal surplus daily, starting from their actual expenditure
Thread, continued, 21:24 UTC · 6 likes
Point here being that it’s possible to get to 2x weight without having eaten 2x as much over your lifetime/period of weight gain, it will just take longer than the person eating much more
Reply to @SpacedOutMatt, 21:25 UTC · 1 like
Matt 🔸 (@SpacedOutMatt): This is false, the extra weight will increase your basal metabolic rate, so you'll stabilize far before 300 lbs
I didn’t say that BMR stays the same, just that eventually you’ll hit 300lbs if you eat a 50 calorie surplus daily
Reply to @AviN102, 21:27 UTC · 9 likes
Avi Norowitz (@AviN102): I think the first model is more accurate. As you gain weight, you burn more calories per day and your weight will eventually stop increasing if your calorie consumption stops increasing.
Not a takedown but a different perspective
Reply to @Adellama_, 21:39 UTC · 3 likes
Ada🔸 (@Adellama_): at the EA conference acquiring the beautiful funny charming @frances__lorenz!
👑👑
Sunday, October 12, 2025
Reply to @ohabryka, 01:33 UTC · 2 likes
Oliver Habryka (@ohabryka): Why bad? CDC seems overall quite bad for the world, seems good for them to see staff cuts.
I don’t really have an incredibly strong view, maybe I should have refrained from posting
Let me search “cdc site:http://lesswrong.com” to see what the deal is
Reply to @ohabryka, 03:06 UTC · 3 likes
Oliver Habryka (@ohabryka): They were majorly responsible for a large fraction of the harm from COVID. They also lied to the public about effectiveness of masks in order to reduce demand for masks, which IMO is a thing that should definitely result in whoever was in charge of it getting fired, and some kind of restructuring.
Right right, it’s been a little while lol
Thread, continued, 03:14 UTC · 5 likes
But this seems squarely not that, at least on a quick read of the article
And more generally, “behaved badly” or even “generally incompetent” doesn’t imply “bad relative to a counterfactual of nothing existing in its place,” and certainly not “arbitrary people getting fired is good” (ik it wasn’t totally arbitrary but doesn’t seem well selected)
Even for the FDA, which I think *is* actually bad relative to a counterfactual with nothing in its place, I don’t want random bureaucrats fired - that doesn’t actually help
Reply to his own earlier tweet, 04:07 UTC · 8 likes
Aaron Bergman, October 11, 2025: Perhaps a better name for the model I’m critiquing is “rational overconsumption” ie food is really cheap and tasty and it’s simply a preference to eat more…
Vibecoded a very imperfect interactive simulation thing. Algorithm isn't perfect obviously, just gets the gist
https://aaronbergman.neocities.org/weightsim
Tweet, 06:12 UTC · 7 likes
This is obviously not a productive line of thought but why can’t Donald Trump and Stephen Miller just personally go do literally anything else?
Go be a rich asshole somewhere doing paid speeches or whatever, why try so hard to ruin the greatest country on earth?
Reply to @ApriiSR, 06:19 UTC · 2 likes
Aprii *️⃣ (@ApriiSR): https://www.lesswrong.com/posts/iZPKuuWsDXAcQWbLJ/shorter-tokens-are-more-likely
huhhh you know what that makes sense
I’m too tired to actually read the post so nobody should listen to anything I say but this is pretty unintuitive to me; my intuition is that the first order “more common token” effect should be exactly cancelled out by the second order+ “more common tokens leave more plausible options for the token after that (and so on) so the second-next token becomes harder to predict”
But then again I guess training is perfectly myopic right so I guess that’s where I’m wrong
Reply to @ApriiSR, 06:25 UTC · 2 likes
Aprii *️⃣ (@ApriiSR): @AaronBergman18 i definitely might have guessed in line with your intuition, if someone asked me about the possibility out of nowhere. but in retrospect it makes sense that something like this might happen.
Claude’s answer fwiw https://claude.ai/share/32be470c-a6d1-4a52-8606-a7f138f73d98
Quote tweet, 16:46 UTC · 12 likes
Goals
Quoting 🅱️ is for baby whisperer (@babieswithb): at a wedding, finding out I'm at the twitter table
Tweet, 17:59 UTC · 9 likes
Worth a read (post by Will MacAskill) https://forum.effectivealtruism.org/posts/R8AAG4QBZi5puvogR/effective-altruism-in-the-age-of-agi
Thread, continued, 17:59 UTC · 3 likes
I like Will, even though I think we have very important disagreements particularly around the moral priority of suffering
Afaict he’s (1) very smart (obviously) (2) extremely earnest and (3) focused on the highest leverage levers to pull
Few such cases!
Monday, October 13, 2025
Reply to @panchromaticity, 01:12 UTC · 3 likes
Xenia 🔍💎 (@panchromaticity): has anyone ever actually said something weird out loud without realizing?
this happens all the time to people in books, but I'm not sure I've done it once ever
Yes lol
Reply to @panchromaticity, 01:14 UTC · 2 likes
Xenia 🔍💎 (@panchromaticity): @AaronBergman18 fascinating say more
Nah
Reply to @BeatrizGietner, 04:48 UTC · 2 likes
Bia (@BeatrizGietner): So what’s everyone’s most controversial economics take?
(quoting @PAHoyeck: So what's everyone's most controversial philosophy take?)
Totally possible + reasonable to have preferences that don’t obey the requisite properties for VNM theorem
Nothing weird/implausible happens if you reject continuity, at the very least, and there are good reasons one might
Reply to @BeatrizGietner, 04:50 UTC · 3 likes
Bia (@BeatrizGietner): So what’s everyone’s most controversial economics take?
(quoting @PAHoyeck: So what's everyone's most controversial philosophy take?)
Interpersonal utility comparisons are obviously real/legit, regardless of what we assume for convenience in our models
Tweet, 13:32 UTC · 4 likes
We dream of a world where virtually all medical devices and treatments are treated like blood pressure monitor devices are now
Reply to @harjtaggar, 17:27 UTC · 10 likes
Harj Taggar (@harjtaggar): Those who can, do; those who can’t, get really into AI safety.
Lol it is such a competitive field/area, this is such a ridiculous critique - it doesn’t even make directional sense
“A bunch of hypercompetent polymaths jacking each other off” would make directional sense at least
Tuesday, October 14, 2025
Reply to @Rado9910, 00:25 UTC · 2 likes
Rado9910 (@Rado9910): I think this is a very good article. I definitely agree with some of the core ideas (and the conclusion!), and it approaches the problem of utilitarian calculations in a way that I didn't consider before.
(quoting @AaronBergman18: *Braces*)
Thank you!
Quote tweet, 04:39 UTC · 19 likes
After thinking about this some more, there’s an additional, weaker claim that I’d like to emphasize, which is:
You should be very skeptical that it’s morally good to bring about worlds you wouldn’t personally want to experience all of
Quoting Aaron Bergman 🔍 (@AaronBergman18): *Braces*
Thread, continued, 04:39 UTC · 2 likes
We can imagine a society of committed utilitarians all working to bring about a very large universe full of lots of happiness and, in an absolute sense, lots of extreme suffering
Tweet, 04:39 UTC · 1 like
The catch is that these very utilitarians are the ones that are going to be experiencing this grand future - not future people or digital people or whomever
Thread, continued, 04:39 UTC · 2 likes
Personally, they all desperately want to opt out of this - nobody wants to actually go through the extreme suffering regardless of what benefits await
Thread, continued, 04:39 UTC · 3 likes
And yet all work day in and day out to bring about this future anyway, condemning themselves to heaven and to hell
Thread, continued, 04:39 UTC · 3 likes
Of course it’s not impossible that their stated morals are right and preferences are morally wrong, but my claim is that we should be very skeptical that the “skin in the game” answer is the wrong one
Thread, continued, 04:39 UTC · 2 likes
Ofc here I’m just supposing/assuming that these utilitarians have the preference to avoid this massive future altogether
Thread, continued, 04:39 UTC · 3 likes
The point, again, is that you should probably not work to build a future that you yourself would not want to experience just because some abstract arguments say that this is morally good
This is a strong hint (not definitive, but still) that those abstract arguments are wrong
Reply to @mbateman, 13:10 UTC · 8 likes
Matt Bateman (@mbateman): 5yo made us buy her an alarm clock so she could set it for 6:30am so she could get us to school as early as possible so she can do this for 30 mins every morning
Queen
Reply to @daniellellecco, 16:45 UTC · 5 likes
(The tweet being replied to by @daniellellecco is no longer available.)
Nah
Reply to @TinkerElla95, 16:51 UTC · 2 likes
TinkerElla95 (@TinkerElla95): @Ruesavatar I have. By several. Romantically, platonically, and even just some acquaintances.
😢
Quote tweet, 18:43 UTC · 12 likes
Break up and partially abolish the FDA to make America even more American
Quoting Lauren Gilbert (@notanastronomer): I love America.
Thread, continued, 18:43 UTC · 6 likes
Trying out “break up and partially abolish” because even though I do think the whole thing as a package deal is bad, I’m in favor of keeping food safety inspections and don’t really mind quality control of pharmaceuticals that much
Thread, continued, 18:43 UTC · 8 likes
Just get rid of the “we decide what’s legal for you to ingest” part
Thread, continued, 18:43 UTC · 7 likes
We can start with making zofran, a basically totally safe, non-recreational, and highly effective anti-nausea drug, OTC
I might just make this specifically a mini hobby horse for a little while
Reply to @adastroworld, 19:00 UTC · 2 likes
adas🧦🌹 (@adastroworld): @AaronBergman18 This shit is magic.
Yeah for real! And nausea absolutely fucking sucks. For some reason it feels like a low status or weak kind of suffering ('aw a little tummy ache') but this vibe is bad and undeserved
Quote tweet, 19:05 UTC · 16 likes
Idk who needs to hear this but next time you're nauseas you should be able to get a totally nausea-destroying drug (Zofran) prescribed pretty easily via telemedicine (eg One Medical offers this)
Just say that you're super nauseas and ask if there's anything they can do
Quoting adas🧦🌹 (@adastroworld): @AaronBergman18 This shit is magic.
Wednesday, October 15, 2025
Tweet, 00:47 UTC · 4 likes
Gender (or closest match) // do you *enjoy* vibecoding (if you’ve never done it, answer “no”)?
Thread, continued, 00:48 UTC · 8 likes
This poll is finally going to give us perfectly clear and unambiguous insight into the female mind
Reply to @NinaPanickssery, 01:10 UTC · 2 likes
Nina (@NinaPanickssery): not yet because ai makes too many mistakes and i end up rewriting everything too often so vibecoding feels a bit doomed
Makes sense (last time I could out-code SOTA was sonnet-3.5)
Thread, continued, 01:11 UTC · 2 likes
Hmm I actually do occasionally correct things but it’s always at a relatively high level and communicated in natural language like “wait are you sure we shouldn’t be doing X instead of Y”
Quote tweet, 02:22 UTC · 285 likes
Ok hello Twitter I am a ✨technical generalist✨ with particular experience in:
- finding and collecting data (esp. data that doesn’t want to be collected)
- AI-based workflows
- dealing with text<->audio
- leveraging AI tools in general
And I am looking for work (🔗 below)
Quoting Aaron Bergman 🔍 (@AaronBergman18): Fuck I’m gonna have to get a real job I think
Watch this space
Thread, continued, 02:22 UTC · 16 likes
See (and share!) the longform version of this thread at
http://aaronforhire.com
And for the vibecoders at heart: you can copy all the content as markdown at: http://aaronforhire.up.railway.app
Thread, continued, 02:22 UTC · 20 likes
The catch of sorts is that I have some health issues that make traditional full time work very difficult, so I need a significant degree of accommodation
I've tried "fake it and hope you make it" and it doesn't end well, so I'm trying to be radically honest and candid here
Thread, continued, 02:22 UTC · 10 likes
Despite my (significant limitations), I *can* get things done, and I have, involving some of these skills (note the first one)
From the "Portfolio" tab
Thread, continued, 02:22 UTC · 13 likes
Some projects include:
Thread, continued, 02:22 UTC · 10 likes
DataLiberator https://aaronbergman.neocities.org/data-liberator-v5
Thread, continued, 02:22 UTC · 8 likes
Too many data scraping projects for my own good (I love data, totally unironically 🤓)
Thread, continued, 02:22 UTC · 10 likes
More data scraping
Thread, continued, 02:22 UTC · 11 likes
Data data data
Thread, continued, 02:22 UTC · 10 likes
I can do some other things too. Some non-technical skills:
Thread, continued, 02:22 UTC · 13 likes
Ok back to what's wrong with me so you know what you're getting into
Something is wrong with my autonomic nervous system; I don't know what exactly (needless to say I'm trying to figure that out)
Thread, continued, 02:22 UTC · 9 likes
The takeaway is that I need to be able to set the pace (and location) of work. You should probably not count on me for strict deadlines. There might be blocks of days where I can't do much at all
I know this rules quite a bit out
Thread, continued, 02:22 UTC · 8 likes
Also maybe it's all fine because 8 hour workdays are fake anyway (this is tongue-in-cheek in that I know I have substantive limitations, but it's not exactly a joke)
Thread, continued, 02:22 UTC · 9 likes
Ok back to the positive stuff: here are some nice things people have said about me
Thread, continued, 02:22 UTC · 8 likes
Not sure I should be including this bit, but whatever:
I'm about to turn 26 (in about 5 months) and lose health insurance!
Thread, continued, 02:22 UTC · 8 likes
And the melodramatic (but true IMO) bit is that I really need health insurance!
Thread, continued, 02:22 UTC · 9 likes
Ok now the fun part - how we can work together!
Option 1) Hire me hourly!
My go-to ask right now is $100/hour
Thread, continued, 02:22 UTC · 9 likes
I know that's a lot of money, but I would urge you to consider the following:
- Hourly = lower commitment for you and higher uncertainty for me
- Hourly = no benefits
- I only charge for serious, focused work
- I will be leveraging AI tools to radically increase productivity
Thread, continued, 02:22 UTC · 9 likes
2) Hire me!
Hey, I know I'm not a good fit for most proper salaried jobs - I just can't commit to 40 hour weeks - but it's worth throwing out anyway.
I can promise I won't be playing videogames on your dime
Thread, continued, 02:22 UTC · 11 likes
3) Secret third thing (my favorite!)
Thread, continued, 02:22 UTC · 12 likes
To address some of those 'secret third thing' options:
I have an audio transcription side project that I wish I could turn into a startup
Presumably there's a catch/reason a disabled vibecoder couldn't cook this up, but I haven't actually been convinced its a bad idea yet
Thread, continued, 02:22 UTC · 10 likes
Are you a tech centi-millionaire (or whatever) in need of a vibecoder, or a research assistant, or generic sidekick? I may be your guy!
Thread, continued, 02:22 UTC · 10 likes
And last but not least, you could just give me money! (Perhaps to tackle specific areas of interest or charitable projects)
Thread, continued, 02:22 UTC · 10 likes
For more comprehensive info (not everything is in this thread), please see: http://aaronforhire.com
And get LLM-feedable content at: http://aaronforhire.up.railway.app
Thread, continued, 02:22 UTC · 12 likes
I genuinely have very little idea how many messages, if any, I'm going to get about this
I will try to stay on top of DMs and replies sent here, but here's a contact form which may be a bit more reliable! (also linked in the doc of course)
Thanks!!
https://forms.gle/zhoWYJyPSFM7XYRs6
Reply to @ManifoldMarkets, 02:35 UTC · 5 likes
Manifold (@ManifoldMarkets): https://manifold.markets/MingCat/will-aaron-bergman-find-a-job-match?r=TWluZ0NhdA
lmao, I should hedge!
Tweet, 03:57 UTC · 7 likes
👀
Quote tweet, 03:59 UTC · 10 likes
👀❤️
Quoting Max Alexander (@absurdlymax): If I actually had money I probably would pay Aaron to make some stuff for me, so imagine what someone competent could accomplish with his skills
You can have to skills, he's literally for hire right now
Reply to @absurdlymax, 04:02 UTC · 3 likes
Max Alexander (@absurdlymax): If I actually had money I probably would pay Aaron to make some stuff for me, so imagine what someone competent could accomplish with his skills
You can have to skills, he's literally for hire right now
(quoting @AaronBergman18: Ok hello Twitter I am a ✨technical generalist✨ with particular experience in: - finding and collecting data (esp. data that doesn’t want to be collected) - AI-based workflows - dealing with text<->audio - leveraging AI tools in general And I am looking for work (🔗 below))
Someone should give max lots of money so he can hire me
Reply to @ShadowyZephyr, 05:26 UTC · 2 likes
zephyr (@ShadowyZephyr): My thoughts on this:
image attached
(quoting @NoLieWithBTC: Breaking: Text messages of Young Republican groups across the country have just been leaked. They talk about putting Democrats in “gas chambers,” attacking “Jews,” calling Black people “watermelon people,” calling to “rape” their enemies, and bringing back slavery.)
Yeah I have similar takes, like people should be very skeptical of this sort of article
I feel like the better (maybe not possible but still) move is to have “everyone” agree on an LLM prompt that you could dump the leaked chats in and then get a pretty fair overall analysis (ofc “everyone” won’t agree on this but better than the status quo)
I’m not worried about the entirety of my digital exchanges getting analyzed in bulk, in fact I think that would probably give a reasonably fair picture of me, but I certainly wouldn’t want like the 4 dumbest things I’ve ever said put under a microscope by journalists who *already* decided I was a bad person
Quote tweet, 05:38 UTC · 403 likes
Looking at maybe a 3-4 digit number of IQ points lost by EA at large due to Huel consumption specifically
This is kinda funny but also apparently the levels are substantial enough that you probably shouldn’t make it a staple of your diet
Quoting Lowell Dennings (@LowellDennings): devastating blow for the effective altruism community
Reply to @LowellDennings, 05:43 UTC · 17 likes
Lowell Dennings (@LowellDennings): I've drank hundreds of huels 💀
I mean in absolute terms the effect is pretty small, you’re looking at a fraction of an IQ point https://claude.ai/share/064c0860-a095-41d7-9e54-c53c6a0e7f84
Thread, continued, 05:45 UTC · 10 likes
Gpt-5 says way less, too tired to evaluate but here https://chatgpt.com/s/t_68ef34d0b000819182319bd13071b99d
Reply to @jazzloaf, 05:55 UTC · 16 likes
vivian (@jazzloaf): I thought the bulk of that is only well researched in the context of chronic *developmental* lead consumption? I assume that's the figure you're using Re: IQ reduction
Im just deferring to the LLMs in this tweet but no at the very least GPT-5 and it looks like Claude take into account developmental vs adult consumption
Tweet, 14:28 UTC · 180 likes
Oh snap
Thread, continued, 14:28 UTC · 3 likes
https://podcasts.apple.com/us/podcast/the-ezra-klein-show/id1548604447?i=1000731957679
Reply to @finmoorhouse, 14:32 UTC · 5 likes
Fin Moorhouse (@finmoorhouse): I said “ten” instead of “an order of magnitude” and…
(quoting @pronounced_kyle: Accidentally said "unrelated" instead of "orthogonal" and they kicked me out of SF)
I said “that” instead of “this”
Reply to @robertskmiles, 14:34 UTC · 14 likes
Rob Miles (@robertskmiles): I'm just going to say it: Socks do not disappear in the wash. You lost those socks yourself, by being careless. I have never lost a sock in the wash
Reply to @TylerAlterman, 15:44 UTC · 2 likes
Tyler is finishing a book, slow to reply (@TylerAlterman): LLMs are the new source of facts for many people, whether we like it or not
Problematically, LLMs still make stuff up. But since they're not going away, what are the best epistemic habits for using LLMs?
Personally, I try to *at least* ask multiple LLMs the same question
Chat with them about something you know well - Gell Mann amnesia stuff
Thread, continued, 15:44 UTC · 2 likes
Tbc this should lead some people to be more impressed than they’d otherwise be, not just the other direction
Reply to @kalomaze, 16:27 UTC · 2 likes
kalomaze (@kalomaze): attempting to tax something that... you can only measure using Anthropic's proprietary endpoint (ant's tokenizer is not public, due to its special heuristics).
they are proposing a tax... for... something they are not letting people who are not them... independently measure...(quoting @LinkofSunshine: Proposing a specific tax on yourself goes hard ngl)
I don’t think there is a real conspiracy to, what, sell secret tokenizer models to the government to enable taxation?
Of all the AI policy challenges, getting people to agree on a tokenizer model for something like this seems pretty damn easy
Tweet, 16:38 UTC · 3 likes
I strongly suspect Apple changed the “decide what notifications are priority/important enough to break through focus” model/algo to have way more false positives
Tweet, 17:38 UTC · 21 likes
Just had Claude Haiku 4.5 autonomously draft this tweet in my browser while I watched! 🤖 It navigated to X, opened the compose modal, and typed this message—all without clicking post. Wild to see human-AI collaboration in action. The future is collaborative, not autonomous. ✨
Thread, continued, 17:38 UTC · 6 likes
Ok I (human Aaron) am back - Claude won't actually click post, but it did indeed type that message out in Chrome^
Reply, 20:32 UTC · 2 likes
Aashish Reddy (@_AashishReddy): @vivposts @AaronBergman18
🫡
Reply, 22:42 UTC · 1 like
Bananaplanet🐌 (@Bananaplanet2): Sorry for not having a job referral though!
Your health situation sounds quite bad but you will be more motivated to make progress on it than I ever would be
And motivation is a real resource that matters for solving problems (and you live in a country where this is profitable)
Thanks but I don’t really know of any good specific strategies to monetize fixing myself lol
I also think my specific health situation is pretty unusual (which is one reason it’s hard to fix) so like I don’t really expect that if I find some major (if incomplete) solution it would deliver value to tons of people
I mean hopefully I’d get a blog post out of it lol
Reply, 23:59 UTC · 6 likes
Alexander Berger (@albrgr): True story: OpenAI once came up on the checkout screen of my Safeway as the local nonprofit to donate to
(quoting @deanwball: could I, like, make a tax-deductible donation to the OpenAI non-profit? just curious.)
Bruh
Thursday, October 16, 2025
Tweet, 00:07 UTC · 9 likes
The “republican group chat is full of nazis” article is reminding me of the “true justified belief is not knowledge” thought experiments
Like they probably are Nazis but I also don’t really trust reporting on the matter that mostly cites specific, individual quotations
Tweet, 00:56 UTC · 6 likes
Pretty pessimistic about this country
Tweet, 01:07 UTC · 13 likes
Radiologists when your imaging is in a Google Drive folder with the exact same contents as the original physical CD instead of on that physical CD (which no other person or industry in the world has any use for)
Reply, 01:27 UTC · 26 likes
Nathan 🔎 (@NathanpmYoung): I don't really understand the criticism of an EA lobbying ecosystem from a16z et al. They are lobbying too and their party has much more money, right?
It’s disingenuous, not worth trying to understand through the lens of earnestness
Reply, 02:29 UTC · 9 likes
(The tweet being replied to is no longer available.)
You’re genuinely typical minding here. I promise I do not feel emotions (in general) in specific parts of my body but they’re utterly qualitatively different from thoughts
Having had anhedonic depression, I know what no/minimal feelings is like!
Reply, 03:21 UTC · 3 likes
Arjun Panickssery (@panickssery): "you don’t blatantly lie. I mean, all politicians lie because all human beings lie. But they lied under the cover of claiming to be honest"
The implication here is that Trump somehow "openly" lies or openly presents himself as a dishonest figure but he presents himself as honest
(quoting @sapinker: On Trump flouting norms and why this is a negative development: "[One of the] remarkable things about Donald Trump—and uniquely among American politicians—is that he has violated norms that have been in place, such as: you don’t boast, you don’t blatantly lie. I mean, all…)
He seems pretty abnormal in some important sense (ik this isn’t literally disagreeing with your tweet)
Reply, 16:49 UTC · 15 likes
Steven Adler (@sjgadler): “If you’re going to work on export controls, make sure your boss is prepared to have your back,” one staffer told me.
For months, I’ve heard about widespread fear among think tank researchers who publish work against NVIDIA’s interests. Here’s what I’ve learned:🧵
Also like they could simply not, it’s not like NVIDIA execs would be poor without the Chinese market
Reply, 18:41 UTC · 2 likes
Jakeup (@myhandle): when I mention "dating markets" people rush in with ideas like "let's incentivize girls to respond on apps!" or "let's tokenize date slots!"
no. stop. the point is that dating *already* has market dynamics. your financial engineering "solutions" will crash headfirst into them
image attached
(quoting @myhandle: if fewer than one third of men are into comic books (likely) or cosplay (almost certainly), then getting into comic books or cosplay increases your market attractiveness)
Not mine, mine will work
Reply, 19:15 UTC · 2 likes
Ado (@adocomplete): We're launching Claude Agent Skills, a filesystem-based approach to extending Claude's capabilities.
Progressive disclosure means agents load only relevant context. Bundle instructions, scripts, and resources in a folder. Claude discovers and executes what it needs.
This was already built in to the recent code environment expansion thing right?
Reply, 19:22 UTC · 2 likes
Cartoons Hate Her! (@CartoonsHateHer): Someone screenshotted me tagging Aella for a sex statistic question on Bluesky and now they're roasting me for "associating with the fascists."
Bruh
Reply, 19:36 UTC · 10 likes
Holly ⏸️ Elmore (@ilex_ulmus): Where’s the lie?
(quoting @rabois: If Anthropic actually believed their rhetoric about safety, they can always shut down the company. And lobby then.)
They should commit to pausing if others do, not unilaterally
Reply, 19:51 UTC · 8 likes
Nathan 🔎 (@NathanpmYoung): Should Taylor Swift exist? Who has she screwed over in making her wealth?
(quoting @DarrigoMelanie: Billionaires shouldn’t exist.)
Hey I agree with you but you don’t have to think someone earned money unethically to think that their wealth should be taxed away
Friday, October 17, 2025
Reply, 02:14 UTC
Wait really
Tweet, 02:56 UTC · 24 likes
Why does @hinge not let me filter by sexuality
I don’t want to waste my time on the straights. They don’t like me
Reply, 03:41 UTC · 9 likes
Roko 🐉 (@RokoMijic): @AaronBergman18 @hinge uh so you're bisexual and it's showing you both women and men?
No I'm straight, but experience has taught me that only bisexual women are interested in me lol
Quote tweet, 04:12 UTC · 6 likes
Has anyone taken an LLM to the entirety of the lesswrong comment section
Idk what I’m talking about but I kinda expect there’s actually a few nontrivial, high quality original ideas lurking there
Quoting Ada🔸 (@Adellama_): finding workable alignment suggestions deep in the slate star codex comment mines of 2016
Reply, 09:49 UTC · 3 likes
City Aesthetics ⛩ (@cityaestheticss): American cities should have more public restrooms
Correct take, don’t listen to the haters
Reply, 09:58 UTC · 2 likes
Luke Metro (@luke_metro): can a 30 year old man who kind of looks like Napoleon Bonaparte pull off the "just moved to SF and looking for friends" genre of selfie? asking for a friend.
Yes king
Reply to @medievalpard, 14:49 UTC · 2 likes
(The tweet being replied to by @medievalpard is no longer available.)
It’s hard to figure out what to disagree with in terms of literally endorsed propositions, but I would invite you to look through the lens of mistake theory instead of conflict theory
Nobody chose their libido (or to be born in the first place)
Quote tweet, 14:53 UTC · 796 likes
I love rationalists
Quoting Louise 🔸 (@LeRoyDesCimes): In favour of flossing:
Reply to @MccallumRupert, 15:08 UTC · 2 likes
Rupert McCallum (@MccallumRupert): @AaronBergman18 @springpard Let me be sure I understand here. You're saying maybe she's got it right that male sexual desire is a cancer upon society. If you really feel that might be right, I suppose you probably can get rid of your libido if you really want to? But I'm not sure where you're coming from.
No, I disagree with the cancer upon society thing but was trying to get at the fact that arguing over rhetorical flourish rather than well defined claims is generally unproductive
Tweet, 15:10 UTC · 181 likes
Maybe this is an infohazard but you can just add people to your Substack mailing list without their consent
Reply to @0x45o, 15:13 UTC · 602 likes
0x45 (@0x45o): can yall pls tell me the lore behind your profile pictures
One girl complimented it so it will be my pfp for all things for eternity
Reply to @LabGrownNeet, 15:15 UTC · 2 likes
laს გr🤗wn იeet (@LabGrownNeet): Attend slutcon
Never flirt outside the workshops (1 exception)
Never ask to squeeze boobs
Leave
Hit it off with a girl on the BART
She's flying to Seattle in 3 weeks
Lfg
Reply to @ejjiott, 15:20 UTC · 5 likes
E J T (@ejjiott): I discovered this the other day!
(quoting @ejjiott: Kinda wild that s-stack lets you sign other people up for a newsletter with just their email. This guy just signed up over 1000 MIT faculty to his newsletter about bitcoin being a criminal honeypot.)
lol nice
Quote tweet, 15:50 UTC · 6 likes
I should probably use my phone less
Quoting Jason Harrison (@nominalthoughts): More people are waking up, it’s the phones
Reply to @jnsyaaa, 15:54 UTC · 4 likes
jonathan (@jnsyaaa): my friend went on a 2-week neotantra retreat where towards the end you were encouraged to confess a deep sexual fantasy and - if others were open to it - it would be fulfilled
all of the girls had fantasies like "I wanna get fucked by 5 guys while I'm trying to escape" and all the guys had fantasies like "you hold me in your arms and in a soft voice tell me that my soul is pure"
My friend is offering me $50 to get the name of the retreat, can u help me out
Saturday, October 18, 2025
Reply to @dieworkwear, 02:25 UTC · 9 likes
derek guy (@dieworkwear): Tyler the Creator on whether youth culture still exists. I don't know if this is completely true, but it does seem more true today than in the past.
Seems good, like maybe we've hit peak age segregation
Age integration is fine/good and needn't be framed as a loss of youth culture (I mean I guess that's a fine framing, but not the one I'd choose)
Reply to @intentiondense, 02:32 UTC · 2 likes
10 day OECDEquityMar31Team larp marathon (@intentiondense): its really difficult being a woman in EA because i love sushi but seafood is high suffering
I half seriously wanna start an ikejime fishing club
Reply to @katclone, 02:36 UTC · 4 likes
kat (@katclone): I’m being fully serious when I say I just collected the coolest data from my entire PhD
WOAH
Post csv
Reply to @JeremiahDJohns, 02:40 UTC · 3 likes
Jeremiah Johnson 🌐 (@JeremiahDJohns): Genuinely looking forward to the research papers in a few years measuring whether or not this move will affect housing prices.
(quoting @MorePerfectUS: BREAKING: New York Gov. Kathy Hochul has signed a bill banning landlords from using algorithms to collude and raise rents.)
Hot take this is a fine/good policy even if it doesn’t tackle fundamentals ie housing supply
Tweet, 02:59 UTC · 155 likes
I live pretty close to the Pentagon and headquarters of various aerospace companies
There’s a certain brand of good looking but slightly off/uncanny Asian woman on Hinge and at this point I think I’d be crazy for not entertaining the hypothesis that they’re literally spies
Thread, continued, 02:59 UTC · 21 likes
Needless to say I’m not certain or close to certain
Also maybe I’m not allowed to post this idk, plz don’t cancel me
Thread, continued, 02:59 UTC · 27 likes
Anyway they’re like professionally taken photos, few or no friends in the pictures, don’t have any interesting personal information like college or employer
Thread, continued, 02:59 UTC · 25 likes
prompts are like slightly off - not how a native English speaker would talk, seem designed to be most generically appealing, generally just an uncanny valley type of vibe
Reply to @knjysnn, 03:05 UTC · 1 like
selrahc (@knjysnn): @dcpip_ ITS A GOOD FIT YOURE SO PRETTY OOMF
Fact check true
Reply to @NathanB60857242, 03:16 UTC · 26 likes
Nathan Barnard (@NathanB60857242): @AaronBergman18 seems completely plausible
Might change my job title to “Aerospace engineer @ Pentagon” and do some digging (mostly joking but I’m tempted)
Reply to @Adellama_, 04:27 UTC · 2 likes
Ada🔸 (@Adellama_): holy shit "pretty asian woman who talks off and has no previous connections" that's me
Lmao
Reply to @MariGO2thepolls, 14:42 UTC · 3 likes
🏳️🌈Mar G-O 🌐🇺🇦 (@MariGO2thepolls): It’s honestly remarkable how right I am about everything
Same
Reply to @baseballcrank, 14:58 UTC · 7 likes
Dan McLaughlin (@baseballcrank): "No More Kings" is maybe 25% people who are against kings & 75% people who want the king to be someone with left-of-center politics.
This…isn’t true
Reply to @JesseTayRiver, 17:36 UTC · 49 likes
Jesse Smith (@JesseTayRiver): My 15 yo has a gf and she absolutely makes him a better person. This should be unsurprising - the default experience of teenagers is to be more attractive by cleaning one's room, cooking etc. - but people don't talk about it much so it was surprising! Weird
Girls can fix us fr
Reply to @polisisti, 17:45 UTC · 3 likes
Alessandro (@polisisti): @AaronBergman18 @JesseTayRiver Even when they don't fully succeed (he says, with self-deprecation), the effect is there
💯
Reply to @NathanpmYoung, 19:24 UTC · 2 likes
Nathan 🔎 (@NathanpmYoung): I think it's cool that someone is doing this.
(quoting @AmmoniteInscrip: Tablets etched with your text, images, or both Made to endure for thousands of years Get yours today)
I agree! I kinda wanna buy but don’t know what I want written on it yet
Reply to @AmmoniteInscrip, 19:33 UTC · 6 likes
Ammonite Inscription (@AmmoniteInscrip): Tablets etched with your text, images, or both
Made to endure for thousands of years
Get yours today
This is cool. You guys should look into long-term digital storage preservation also and possibly sell that alongside, probably using M-discs
Like my idea is that there’s an option to physically attach some sort of very durable storage case to the inscribed glass, and put M-discs inside
The glass would have whatever people want written but also a note that’s like “see discs inside for digital archive”
Maybe that triples cost or something but allows a gazillion times more data to be stored
I think it’s very sad that I could get hit by a car tomorrow and most of my digital life would just disappear with basically no hope of recovery by family or future people or whomever - I’d like to change that!
And back to the object level: M-discs aren’t especially expensive but most people don’t know about them and you often need to buy a case and some special hardware to use, but like you as a company could just have an efficient workflow for that process so that people can just send you like a google drive link
Reply to @damiendamien, 19:47 UTC · 2 likes
Damien McElvanna (@damiendamien): Counterpoint https://www.theguardian.com/uk/2002/mar/03/research.elearning
I mean I assume if there was any amount of political or popular will they could track down or build the proper device - this is what I’m assuming would (or at least could) happen in the future
Sounds like the data is there but they can’t easily find the hardware to read it
Tweet, 22:39 UTC · 29 likes
Sometimes (eg just now) it hits me how weird AI is
Idk if it's sentient (I think probably not) and I know at some level it's "just math," but in some important sense there is a guy in my computer - a guy!
Sunday, October 19, 2025
Reply to @ChanaMessinger, 04:14 UTC · 3 likes
Chana (@ChanaMessinger): There’s so many frogs
Yeah I’d say that’s a decent number
Reply to @Chesedgirl, 05:18 UTC · 50 likes
Chesed (@Chesedgirl): I wish I knew more men who were 1) turned on by women suffering but 2) committed to behaving ethically (not causing non consensual suffering). If this is you and you stumble into the right crowd, you could get laid...a lot
I like girls and want them to be happy 😡
Reply, 12:48 UTC · 2 likes
What are the logistics of this? How do stay blue?
Tweet, 20:42 UTC · 69 likes
Trump is a very bad man and should not be president
Reply to @Lib_Development, 20:49 UTC · 10 likes
(The tweet being replied to by @Lib_Development is no longer available.)
Yeah I guess ppl can be pedantic about what I mean by “shouldn’t be” - although I do think he specifically should have been impeached and removed from office by now
Tweet, 21:01 UTC · 27 likes
A certain level of autism is actually really great for creating friendships and community
The normie worries about whether planning too many events themselves (as opposed to being invited) signals desperation, or about whether only the correct subsets of people are mingling…
Thread, continued, 21:01 UTC · 16 likes
The wise autist invites all their pals to the thing (that wouldn’t have otherwise happened) and introduces people without worrying about nth order ramifications or insane social dynamics
Tweet, 21:28 UTC · 2 likes
@AnthropicAI plz this is insane
Tweet, 22:25 UTC · 5 likes
xAI might have actually succeeded at making Grok (at least Grok 4 fast) not woke and not racist/anti-woke to the point of insanity
Compare this chart (grok 4 fast)
Thread, continued, 22:25 UTC · 2 likes
To this one (Claude Sonnet 4.5)
Monday, October 20, 2025
Reply to @Littoral_Gecko, 01:13 UTC · 2 likes
A Gecko🦎 (@Littoral_Gecko): @AaronBergman18 I’m pretty sure a high-agency neurotypical person goes “I want to do this thing with my friends” and then does it.
What you’re describing sounds more autistic tbh.
Oh lol
Reply to @Leo_Abstract, 01:47 UTC · 4 likes
Leo Abstract (@Leo_Abstract): @AaronBergman18 entirely possible, but given the number of silly little hoaxes people believe about him, these words as assembled have no particular meaning at first glance.
disambiguating you from the hoaxed would be laborious and likely not a process you'd be interested in.
lol I don’t expect this to be super productive but the claim isn’t dependent on liberals in general having good epistemics, it’s dependent on there not being an unprecedented conspiracy among every part of civil society to lie in coordination about like 50 different things, including plenty of AEI-type republicans and Mike Pence
Tweet, 02:50 UTC · 232 likes
Aella should make a 2x2 that’s like “how much do u like the other gender” and “how much are u sexually attracted to the other gender”
I feel like this would explain some things
Reply to @SpacedOutMatt, 02:56 UTC · 27 likes
Matt 🔸 (@SpacedOutMatt): https://www.cartoonshateher.com/p/you-dont-like-women-you-like-boobs
I like women
Reply to @EigenGender, 04:15 UTC · 4 likes
EigenGender 🔸 (@EigenGender): it’s very hard to get an unfiltered view of how people spend their time. this post is my best attempt at that: I’d appreciate if you responded to this with what you’ve been doing for the past 30 minutes
Group chat messaging + change sheets + shave + shower
Tweet, 04:29 UTC · 18 likes
Seems bad that the NYT best seller list-podcast sphere industrial complex mostly just ignores any open access books
Like a naive simple model would suggest that free stuff should be more popular and create more discourse, but alas
Reply to @yzilber, 04:35 UTC · 4 likes
Jazi Zilber (@yzilber): @AaronBergman18 The choice is always a package.
GOP full policy list + trump. Vs Dems full list + Kamala.
A reasonable person could have voted for Trump if he values, say avoiding illegal migration highly enough.
I mean I don't want to make claims about the hearts and minds of all Trump voters, I agree at least one person who voted for Trump I'd categorize as "reasonable"
But I do think his actions+policies are some combo of insane and unethical from virtually any plausible worldview
Thread, continued, 04:38 UTC · 4 likes
And I say this as someone with sympathy for a few things he's done (disparate impact, weakening some of the NGO sector, culturally pushing back on wokism)
Thread, continued, 04:39 UTC · 3 likes
It just seems pretty unreasonable to me to prioritize those over rule of law, high skilled immigration, other freedom of speech issues, and competent foreign and economic policy - ie the reason America is starting from a strong position
Reply to his own earlier tweet, 05:18 UTC · 19 likes
Aaron Bergman, October 20, 2025: Aella should make a 2x2 that’s like “how much do u like the other gender” and “how much are u sexually attracted to the other gender” I feel like this would…
Ok here u go:
http://aaronbergman.neocities.org/affinityquiz
Reply to @inerati, 05:21 UTC · 2 likes
liz (@inerati): what if you’re the genre of autistic who worries even more than normal about such things
☹️ (idk maybe just don't do that)
Reply to @Rado9910, 17:04 UTC · 2 likes
Rado9910 (@Rado9910): @AaronBergman18 10/10, 10/10. I just really, really love women.
So true
Reply to @Duderichy, 17:07 UTC · 5 likes
the Rich (@Duderichy): telling a girl I work at AWS so she knows I’ll go down on her
Banger
Reply to @arthur_wright_, 17:09 UTC · 2 likes
Arthur Wrong 🔸 (@arthur_wright_): An old password that i'd reused a bunch as a teenager was compromised in a data leak, and while changing a bunch of account passwords i just discovered that an extremely low-effort remix i made as a 14 year old has 39.2k plays on SoundCloud lol
Lmao
Reply, 17:09 UTC · 1 like
Post link!!
Quote tweet, 17:20 UTC · 13 likes
Competence porn
Quoting Open Source Intel (@Osint613): In Taizhou, China, a car plunged into a river. Without thinking twice, 40-year-old Di Shuangcheng jumped in, grabbed a stone, and smashed the window to save the trapped driver.
He said his mind was calm, only focused on what to do next. “Saving the person was the most important thing.”
@volcaholic1
Tweet, 17:50 UTC · 198 likes
Didn’t AWS claim like 99.999999% reliability or something?
I mean maybe they hit that one in a billion accident yesterday but I think it’s more likely that they’re (intentionally) bad at forecasting
Quote tweet, 17:54 UTC · 821 likes
You should not have to ask anyone permission to install air conditioning
Quoting Timothy B. Lee (@binarybits): This is insanity.
Reply to @beala, 19:34 UTC · 18 likes
Alex Beal 🆎 (@beala): It's all service specific. For example, they only guarantee two 9s for any given ec2 instance. I think the super high SLAs are for S3 durability, and afaik they haven't lost any data. https://aws.amazon.com/compute/sla/?did=sla_card&trk=sla_card
Ah nice
Reply to @timfduffy, 20:05 UTC · 15 likes
Tim Duffy (@timfduffy): Unforgivable, when I went to bed last night I had to turn off the lights manually because my Echo Dot wasn't working.
This is basically 9/11
Tweet, 21:01 UTC · 8 likes
AWS going down is like the opposite of 9/11 (V-E day?) for local data hoarders
Tweet, 21:08 UTC · 95 likes
AWS going down is kinda funny bc people’s lightbulbs and smart blankets stop working but like actually we should probably take this as a warning shot of sorts to prepare for a correlated outage across many cloud providers/data centers
Thread, continued, 21:17 UTC · 6 likes
Maybe we shouldn’t be selling cyberattack intelligence hardware to China
Tuesday, October 21, 2025
Reply to @soft_fox_lad, 03:46 UTC · 2 likes
fox ⏹️ (@soft_fox_lad): @AaronBergman18 It's been gamed out. In the exceedingly unlikely event that that happens, it seems like we can wait a few hours for the SREs to fix the problem (get lunch, clean up your home or apartment, take a half-day, etc) and (even if it goes on for a week) at worst be pretty inconvenienced
hope ur right
Reply to @livgorton, 04:12 UTC · 42 likes
Liv (@livgorton): my coworkers have been subjected to this yap so I now continue here:
this is genuinely one of the most useful papers to engage with for interp imo. easily one of my most referred to papers. if you have not read it, pls do. mandatory reading until the heat death of the universe.
Here u go fellow Liv followers https://arxiv.org/abs/2104.07143
Tweet, 04:49 UTC · 17 likes
Anybody done any good schemes lately? I want to be part of a scheme, looking for inspiration
Reply to @warty_dog, 13:04 UTC · 3 likes
troon (@warty_dog): you can't have .9 apples
Yes you can lol
Quote tweet, 14:00 UTC · 13 likes
MAGA guy who gets fired for voting Obama in 2008
Quoting Andrew Prokop (@awprokop): Paul Ingrassia of the "Nazi streak" leaked texts has shown up in lawsuit filings from fired FBI agents.
Earlier this year, Ingrassia was the White House loyalty enforcer at DOJ, asking FBI agents whether they'd voted for Trump — or "a Democrat in the last five elections"
Tweet, 19:58 UTC · 31 likes
I feel called out
Wednesday, October 22, 2025
Reply to @StatsLime, 00:24 UTC · 12 likes
Max ⏸️ Limelihood 🔸🔸🔸 (@StatsLime): ? I’m not sure how to interpret this
I think like "teaching = one of the few accepted careers for women at one point in american history which harmed women who could have enjoyed or made more money in a different profession in a different cultural environment but one upside is that that smart women became teachers rather than eg idk consultants and this had positive externalities (e.g. for the students and second order for everyone else)"
Reply to @polisisti, 07:37 UTC · 2 likes
Alessandro (@polisisti): reminder not to use ChatGPT because the company that creates it is evil
I agree with the second part but basically disagree with the first
I mean for personal casual use I guess do whatever you want, but for anything that affects your ability to do good, use the best tools available
It’s not that the costs side of the ledger is literally zero but as a heuristic I think it’s going to be awfully small
Reply to @felpix_, 16:29 UTC · 22 likes
felpix (@felpix_): high schoolers who email professors to do data cleaning and call it "research" on their college applications should be taken into an alley and be executed by firing squad
(quoting @instrumenthull: careful with your AI-generated emails folks!)
I disagree
Reply to @felpix_, 17:17 UTC · 29 likes
felpix (@felpix_): @AaronBergman18 elaborate
Data cleaning is a valuable task and there’s no harm in a polite email
Reply to @Lewis_Bollard, 17:39 UTC · 3 likes
Lewis Bollard (@Lewis_Bollard): My favorite quote that the Animal Ag Alliance took down
Lmao
Tweet, 18:27 UTC · 280 likes
So true, gmail autocomplete
Thread, continued, 18:27 UTC · 20 likes
(this is not in fact what I was intending to say)
Tweet, 18:57 UTC · 26 likes
Might get cancelled for this woke take but spanish is aesthetically superior to english
Thursday, October 23, 2025
Tweet, 03:06 UTC · 8 likes
"X is trivial" is such a funny critique because it is often intended to undermine X whereas actually it supports X
Like ok it's trivial - that means it's almost certainly true, so act like it's true!
Thread, continued, 03:06 UTC · 4 likes
The standard dynamic here is:
Person 1: "X" [actual goal/implied subtext: so act like X is true]
Person 2: "X is trivial" [actual goal/implied subtext: so person 1 isn't high status or clever]
Reply to @EffectvAltruism, 04:01 UTC · 2 likes
Effective Altruism (@EffectvAltruism): .@Lewis_Bollard on how effective altruism helped him move past a feeling of futility and toward a grantmaking career that has secured better conditions for hundreds of millions of animals.
Based
Quote tweet, 04:08 UTC · 9 likes
Signed
Quoting Dr. Roman Yampolskiy (@romanyam): https://superintelligence-statement.org/
Reply to @AndyMasley, 04:09 UTC · 5 likes
Andy Masley (@AndyMasley): The thing I really don't like about being on S*bstack is how much it tries to associate itself with alternative media. I don't want that at all. Alt media mostly caters to lowest common denominator populist bullshit. I want a newsletter platform and that's all.
I don’t know enough to strongly disagree but I’ve never seen this
I’ve only seen other people alleging it
Tweet, 04:57 UTC · 6 likes
What the fuck!!!!!!!!!!!!!!!!
@JesseTayRiver fix this immediately
(Regarding window A/C unit)
Tweet, 05:09 UTC · 7 likes
God I am such an idiot
I use AI more than almost anyone I know and yet still wait stupid amounts of time before asking about how to solve mundane problems that eventually turn out to be totally solvable via whatever magic Claude spits out
Thread, continued, 05:09 UTC · 2 likes
People talk a lot about ✨intelligence✨ but mere knowledge can actually do quite a bit of work
Thread, continued, 05:11 UTC · 5 likes
Plausibly we underrate knowledge per se because it’s correlated with intelligence in humans
But maybe there’s a weaker correlation in other architectures of mind
Reply to @codytfenwick, 05:44 UTC · 2 likes
Cody Fenwick (@codytfenwick): This is how it should work — you don't want AC cooling the room below the thermostat setpoint!
I want the thermostat setpoint to be lower 😡
Reply to @codytfenwick, 06:26 UTC · 2 likes
Cody Fenwick (@codytfenwick): There should be an adjustment for that? Are you already at the lowest setting?
Oh hmm
Reply to @hopes_revenge, 06:32 UTC · 4 likes
hope hopes hoping (@hopes_revenge): why aren’t homeless people ever nice ?
IME they sometimes are
Reply to @austinc3301, 15:46 UTC · 7 likes
Agus 🔸 (@austinc3301): I know you’ve retired @AndyMasley, but I feel forced to show you this
Bruh
Reply to @TheStalwart, 15:52 UTC · 3 likes
Joe Weisenthal (@TheStalwart): What's the best argument that the liberalization of sports betting laws has made the country better off?
The fact that freedom in general makes people better off so we shouldn’t regulate/add constraints to an arbitrary case unless there’s compelling reason to do so
Reply to @Liv_Boeree, 16:03 UTC
Liv Boeree (@Liv_Boeree): Haha no not @Grimezsz on this occasion
Wait who is it
Reply to @NoahTopper, 21:49 UTC · 10 likes
Quinoah 🔍⏸️ (@NoahTopper): My fav story about my wife is when she told her then-boyfriend they needed to talk:
Him: Oh god how bad is it?
Her: Oh…like a 7/10
Him: Oh okay, that’s not too bad, what’s up?
Her: I want to break up
Him: …HOW IS THAT A 7??
I side with wife 😤
There are actually a lot of things worse than being broken up with!
Maybe it should have been an 8 but this seems pretty well calibrated
Reply to @NoahTopper, 21:49 UTC · 4 likes
Quinoah 🔍⏸️ (@NoahTopper):
Correct take
Reply to @TinkerElla95, 22:49 UTC · 3 likes
TinkerElla95 (@TinkerElla95): @AaronBergman18 @NoahTopper @basilYam I did have a whole break down of the scale and why I selected 7!
Based
Reply, 23:02 UTC · 1 like
nice
Reply to @mattyglesias, 23:11 UTC · 9 likes
(The tweet being replied to by @mattyglesias is no longer available.)
Illegality + ~100% non-enforcement makes me uncomfortable but seems underrated as a strategy
Tweet, 23:22 UTC · 10 likes
Fun fact: the following url exposes some fun juicy information about all ur Claude dot ai chats including usually-hidden summaries of recent conversations and whatever these code name features are
claude[dot]ai/api/organizations/[organization_id]/chat_conversations
Reply to @EAForumPosts, 23:41 UTC · 3 likes
EA Forum Posts (@EAForumPosts): Quick take by Benton 🔹
Ya someone should figure out how to save American liberal democracy fr
Reply to @peterwildeford, 23:43 UTC · 5 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): How much influence should Elon Musk have over the future robot army? Discuss
Well given his track record, a lot of influence over the design and manufacturing process but zero direct policy influence
Quote tweet, 23:53 UTC · 22 likes
As I venture into the world of having to acquire my own health insurance, it is increasingly weird that people talk about salary/earnings without mentioning it explicitly
$50k + standard white collar benefits vs $50k just cash are radically different
Quoting Jaimie (@mountainwesttax): hear me out: public option for catastrophic coverage. cash pay for everything else
Thread, continued, 23:53 UTC · 3 likes
There are definitely salaries that are low enough to be rare in the full time w benefits world and salaries high enough that the $10k-$25k out of pocket difference doesn’t matter a ton, but the band in-between (say, $40k-$70k to be arbitrary) covers a lot of people
Friday, October 24, 2025
Reply to @PAHoyeck, 04:54 UTC · 31 likes
Phil Hoyeck (@PAHoyeck): Okay, jokes aside, what's the worst piece of philosophy you've ever read?
Nicomachean Ethics is pretty bad
I mean I don’t think it’s bad relative to other ancient philosophy, but I resist reading old philosophy at all costs because it’s almost always just totally wrong so the set of what I’ve read is like {Nicomachean Ethics bc college, modern analytic philosophy}
It’s like comparing the worst of modern Olympians to Gru the caveman who was the best at rock throwing in his tribe
Tweet, 05:26 UTC · 12 likes
Might become a browser extension guy
I used to think they were basically gimmicks, and many are, but there’s actually a lot you can do with direct access to network interaction
Reply to @drfunkenst3in, 06:20 UTC · 4 likes
Will (@drfunkenst3in): Do you similarly dislike ancient phil on subjects other than ethics?
Ya
Quote tweet, 06:54 UTC · 6 likes
Obligatory
Quoting Tina Tarighian (@wwwtinazone): 50,000+ people have seen my AI Purity Test http://aipuritytest.org and all the comments are saying they scored in the 90s, but I'm sure tech twitter could change that...
Reply to @louispereira, 07:03 UTC · 2 likes
Louis Pereira (@louispereira): For the first time, I'm choosing not to be an early adopter.
I feel similarly, although I played around with atlas a bit
There’s not actually a lot that I want to do with an AI browser that i can’t do with more control and more understanding with Claude Code or ChatGPT
Or maybe this is just a failure of imagination idk
Reply to @RafaRuizdeLira, 07:05 UTC · 1 like
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): @AaronBergman18 Everyone who scores higher than me is slop brained, anyone who scores lower than me is ngmi
Precisely
Reply to @justalexoki, 07:18 UTC · 9 likes
taoki (@justalexoki): this but unironically
(quoting @vocalcry: A 50-yr-old man cannot meaningfully consent to a relationship with a 22-yr-old woman bc his blind lust makes him functionally retarded so the power imbalance is too extreme)
I don’t literally agree here lol but the general point that young attractive women at least in relatively liberal circles/areas in the west have lots of power in important ways stands
Which is fine - it’s not a competition. But it’s insane to pretend that we’re still living in the 1950s by ignoring the cultural, social, and legal power of women
Reply to @PAHoyeck, 17:51 UTC · 2 likes
Phil Hoyeck (@PAHoyeck): Some of these replies are going to make me cry, man.
lol
Quote tweet, 19:42 UTC · 22 likes
I want to get a better handle on all this stuff, but afaict moral progress enabled by western style liberalism is actually extremely important from an impartial welfare-oriented perspective
US vs China is NOT just parochial nationalism
Quoting EA Forum Posts (@EAForumPosts): "How good would a CCP-dominated AI future be?" by OscarD🔸
Thread, continued, 19:42 UTC · 5 likes
My current sense is that with respect to animals specifically, the west pretends to care (and basically doesn’t) but China doesn’t even pretend to care
Neither are great ofc, but the West’s orientation looks way less terrible
Reply to @NoahTopper, 19:57 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): I would pay good money for access to all the Curb Your Enthusiasm episode outlines
Is this a data scraping task 👀
Tweet, 21:53 UTC · 2 likes
Question for AI twitter: suppose you're having a model categorize e.g., 100 different wikipedia articles according to metric X (eg "yes Aaron would find this interesting" or "no he would not)
But the model only sees a single article + prompt with criteria at any time
Thread, continued, 21:53 UTC · 2 likes
Suppose *I* know the approximate base rate of "Aaron would find this interesting"
Does it improve performance/accuracy to tell the model the base rate, so e.g., "about 10% of the time, you should categorize articles as interesting"?
Saturday, October 25, 2025
Tweet, 02:38 UTC · 142 likes
Judging people for being fat and/or unattractive is fundamentally a failure of decoupling send tweet
Reply to @AnniePosting, 02:54 UTC · 35 likes
annie 👁❌🦋🪞 (@AnniePosting): 🔒🔃
No this is wrong. I’m pretty willing to grant correlations, but that’s all they are
Intellectual maturity is recognizing correlations without reifying them into intrinsic properties or inherently coupled
And more generally the genre of this tweet is “boldly and proudly claim something” which is like a fine thing to do but it’s not very convincing
Tweet, 03:15 UTC · 283 likes
Who wants to try this EA/rat/TPOT/normie categorizer I vibecoded in 45 mins for fun
Thread, continued, 03:15 UTC · 77 likes
Reply to @jannalulu_, 03:18 UTC · 1 like
Janna (@jannalulu_): mine is 42 lol
Based (I am ashamed of my high score)
Reply to @peterwildeford, 03:23 UTC · 3 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): Which LLM do I use for what?
I realize this is going to be an unpopular take, and honestly it's all super subjective at this point, but this is what I've gravitated towards lately and I feel strongly about it
I’ve found gpt-5-thinking best at most arbitrary non-coding things
Reply to @IsaacKing314, 04:04 UTC · 33 likes
Isaac King 🔎 (@IsaacKing314): @AaronBergman18 Independent bar chart at the top converted to pie chart at the bottom is upsetting. Either people sum to 100% personality or they don't!
I’m anti-anti-pie chart
Ppl hate them for some reason, it’s totally fine to consider ratios independently of magnitudes
Reply to @polisisti, 04:04 UTC · 2 likes
Alessandro (@polisisti): @RafaRuizdeLira @AaronBergman18 nuuuu post deleted
I will never experience the pumpkin's substantiality
😢
Reply to @NinaPanickssery, 04:09 UTC · 10 likes
Nina (@NinaPanickssery): How does it work? I thought twitter didn’t have an API?
It does but it’s crazy expensive now
But xai just added integration for x search including filtering by username as a tool the model can call
Reply to @NinaPanickssery, 04:14 UTC · 1 like
Nina (@NinaPanickssery): Your next project should be embedding visualizations of X profiles! With all tweets and replies. If that’s doable.
Thank you! lol it’s sorta doable but would take way longer bc twitter doesn’t like you scraping
Looks like xai intentionally doesn’t expose the actual tweet text, like only the model gets to see that
But if someone wants to cover the twitter api cost I will do the rest lol
Reply to his own earlier tweet, 04:39 UTC · 41 likes
Aaron Bergman, October 25, 2025: http://bit.ly/xpersonality
Ok so far we've used about $3 over 300 analyses. I'm putting another $10 for tn which means we have roughly 1200 more analyses until my credits run out and the app stops working
After that I may add a few more dollars but will likely solicit venmo donations lol
Thread, continued, 04:41 UTC · 41 likes
Also don't over-index on this - it only pulls about 30 of your most recent tweets (sad)
Reply to @quasistable, 05:30 UTC · 13 likes
Alastair and 147 others (@quasistable): What do we mean by “judging” here? Something like “assume they’re a morally deficient person”?
Yeah honestly I kinda dislike the word despite using it myself
Obviously in some literal sense you can judge things like “the obese person probably isn’t an extremely fast runner”
But yeah basically: assuming they’re less deserving of respect and empathy and that they must be bad in ways that aren’t directly tied to their attractiveness/fitness
Reply to @Mechanical_Dodo, 05:38 UTC · 3 likes
Dodo (@Mechanical_Dodo): can you carve out a space in the center of the pie chart for those of us who just want to grill but can't pass as normies?
I will implement this feature tomorrow!
Reply to his own earlier tweet, 13:28 UTC · 32 likes
Aaron Bergman, October 25, 2025: Also don't over-index on this - it only pulls about 30 of your most recent tweets (sad)
Ok credits have been added
Reply to @medievalpard, 18:08 UTC · 4 likes
(The tweet being replied to by @medievalpard is no longer available.)
💯
Reply to @NTFabiano, 18:12 UTC · 232 likes
Nicholas Fabiano, MD (@NTFabiano): People with antisocial traits prefer bitter tastes like black coffee.
Smells like something that won’t replicate
Reply to @NoahTopper, 18:13 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): Adorable
Shoot ur shot king
Reply to his own earlier tweet, 18:44 UTC · 11 likes
Aaron Bergman, October 25, 2025: Ok credits have been added
Hate to do this but <$5 out of $30 remaining, I guess Venmo me if u wanna keep the thing up and running
https://venmo.com/u/Aaron-Bergman-5
Quote tweet, 18:50 UTC · 77 likes
If u are working on frontier RL/scaling at an AI lab I humbly invite you to burn out, go on an extended vacation with your riches, and if you ever feel like returning to the grind, apply to Anthropic
Quoting Nathan Lambert (@natolambert): A new essay on the crazy, all or nothing approach to work happening in AI today, the looming human costs, and the lack of a finish line.
I wouldn't say it's okay, but I'm not sure how to fix it.
Reply to @simoj_, 20:03 UTC · 2 likes
simoj (@simoj_): Idk, most of what people usually mean by “judging” is just empirically valid inference
Yes this is a thing but it often goes above and beyond that (see confirmation bias), and there's also a normative/ethical aspect that statistical inference has nothing to say about (i.e., how you treat people even if you do valid inference of some characteristics)
Reply to @adam__isom, 20:05 UTC · 3 likes
adam ✧ ❥ ~ (@adam__isom): being fit and attractive is largely a choice
I'm pretty genetics-pilled but to some extent yes. Even still
Reply to @dcpip_, 20:17 UTC · 1 like
(The tweet being replied to by @dcpip is no longer available.)_
Hate to be lazy but Claude literally does a better job at explaining my own ideas (this is after a bit of back-and-forth)
Reply to his own earlier tweet, 20:18 UTC · 7 likes
Aaron Bergman, October 25, 2025: Judging people for being fat and/or unattractive is fundamentally a failure of decoupling send tweet
Explanation by Claude (that I endorse)
Reply to @Mjreard, 21:04 UTC · 4 likes
Matt Reardon (@Mjreard): Diet of kings rh
This is kinda based but IMO the actual rational thing to do here (given what I assume u care about) is to let in quite a bit of intuition bc we literally do not understand nutrition science/biology well enough to design a "perfect" diet up front/without biological feedback mechanisms that manifest as craving a certain food, feeling better or worse, etc
Seems fine as long as you feel good, but if you notice eg a craving for blueberries or something then just get the blueberries
Ofc veganism I understand/greatly respect but would urge EAs to not make that an absolute rock solid 100% priority (eg if you're craving meat for weeks on end, that's a real sign that you're missing something important)
Reply to his own earlier tweet, 21:33 UTC · 9 likes
Aaron Bergman, October 25, 2025: Hate to do this but <$5 out of $30 remaining, I guess Venmo me if u wanna keep the thing up and running https://venmo.com/u/Aaron-Bergman-5
Ok people like this but I've spent about $50 now (not all used up, about $30 used) so if anyone wants to pitch in a few bucks, https://paypal.me/aaronbergman18?country.x=US&locale.x=en_US
Thread, continued, 22:10 UTC · 11 likes
Lol sorry for slowness, I am learning on the fly
Reply to @ruben_bloom, 22:30 UTC · 3 likes
Ruben Bloom (Ruby) (@ruben_bloom): Make it so that when it's out of credits, you get that error message and then make it clear how you can pay to fix that within app
Lol sounds ideal but sounds relatively difficult to set up esp if we want the whole pipeline to be automatic/no Aaron in the loop
Reply to @_AashishReddy, 23:18 UTC · 3 likes
Aashish Reddy (@_AashishReddy): @re4lpredicate You’re in your 20s right
Hey
Dude solid 650 deadlift the other day, and thanks for saving my dog from that fire
Reply, 23:58 UTC
They’ve been around for a long time yes but not culturally/institutionally dominant
I mean it’s easy (and I think important) to recognize the harmful and illiberal effects of wokism but it grew out of a genuinely good cultural ethos of actually caring about others’ wellbeing even if they’re disempowered in some way
I mean 1st-3rd wave feminism at least was good overall
I actually really like this list claude made because I think it pretty fairly shows/describes various phenomena
Like discourse seems focused on the overreaches of these sorts of things (which I think are real and we should indeed correct on the margin) but like decline in violence/caring about feelings/decline in bullying/acceptance of mental health concerns as legit are good overally corrections from like the 1990s and prior
Arguably the best balance was struck in, idk, maybe the 2010s? So maybe I'd take 2015 culture over today's in some respects but I'd also take today's over anything pre-2000
Sunday, October 26, 2025
Quote tweet, 00:10 UTC · 16 likes
Thank you to the following people for supporting this project!!
- @polisisti: $20
- @HenriThunberg: $10
- @freed_dfilan: $10
- @dioscuri: $5
That means I've just about broken even so far (not trying to make money on this)
Comment/DM me if I'm missing anyone!
Quoting Aaron Bergman 🔍 (@AaronBergman18): Who wants to try this EA/rat/TPOT/normie categorizer I vibecoded in 45 mins for fun
Thread, continued, 00:12 UTC · 2 likes
Venmo: https://venmo.com/u/Aaron-Bergman-5
Thread, continued, 00:13 UTC · 2 likes
PayPal: https://paypal.me/aaronbergman18?country.x=US&locale.x=en_US
Reply to @NinaPanickssery, 00:16 UTC · 2 likes
Nina (@NinaPanickssery): @AaronBergman18 @springpard like crime is very very bad
being soft on crime is very costly
also therapy culture is bad, people don’t become happier from it
I feel like you're underrating how bad masculine-dominant culture was in the 20th century
Reply to @polisisti, 02:04 UTC · 6 likes
Alessandro (@polisisti): I see my donation as an investment in Aaron — I think eventually he's going to have a really good idea, and I want him to keep experimenting
Oh no pressure is on
Thread, continued, 02:04 UTC · 6 likes
But thank you lol
Reply to @Mjreard, 02:11 UTC · 5 likes
Matt Reardon (@Mjreard): The contents of the picture are described in the spreadsheet, so I don't see what new information is being conveyed here, but as you wish (nb: I ate more rice in Korea than I do now)
I do respect this lol
Reply to @cljack, 02:14 UTC · 9 likes
Charlotte Lee (@cljack): I got my 11yo some time in a 737 simulator for her birthday. The instructor asked what scenarios she wants to practice and she asked if they could do United 232 💀
Based mom
Quote tweet, 02:41 UTC · 3,273 likes
This is an absolutely nuts story, how did I not know about this?
All of the hydraulics failed and the only mechanism of steering the plane was thrust by the two engines. Not only left and right steering - that was all they had to control altitude as well
Quoting Charlotte Lee (@cljack): I got my 11yo some time in a 737 simulator for her birthday. The instructor asked what scenarios she wants to practice and she asked if they could do United 232 💀
Thread, continued, 02:41 UTC · 423 likes
So basically they were a bullet in the sky aiming for a single airport and managed to hit the runway
It's in part a tragedy >100 people died, but to quote Claude, "Of 296 people on board, 184 survived—a number that should have been zero."
https://en.wikipedia.org/wiki/United_Airlines_Flight_232
Thread, continued, 02:41 UTC · 206 likes
Just an insane story because absolutely everything needed to go right for anyone to survive and somehow it did
Claude estimates that 1% of crews could have pulled this off, but 99.9% of those highly skilled crews would have all died anyway
Thread, continued, 02:41 UTC · 27 likes
If ur curious https://claude.ai/share/167ea32e-8a6e-47b8-ab21-2c49688afdc1
Thread, continued, 02:54 UTC · 59 likes
This is also nuts, although a bit less cleanly heroic https://en.wikipedia.org/wiki/Citicorp_Center_engineering_crisis
Tweet, 03:37 UTC · 323 likes
Everyone wants AI to cure cancer, nobody wants to make HIPAA less insane to let that happen
Reply to @austinc3301, 03:38 UTC · 3 likes
Agus 🔸 (@austinc3301): Now that I've thought of falsifiability, I've decided to become an accelerationist
Lmao
Reply to @johnsmithyson0, 03:44 UTC · 9 likes
John Smith (@johnsmithyson0): @AaronBergman18 I do!!! I want to make HIPAA less insane!!! I want to make human experimentation easier!!!!! I want to give doctors more power and less pay!!!
Ok we have an existence proof (fr though I do recognize this is not an extremely uncommon take on certain parts of twitter but it seems unlikely to become a policy priority by any party)
Reply to @cljack, 04:00 UTC · 63 likes
Charlotte Lee (@cljack): Even more nuts: After this crash, NASA developed software that could safely land a plane with just propulsion in the event that all hydraulics were lost, but they decided it was too expensive to implement 🥲
Rip, I mean I’m not sure it was the wrong decision given how safe flying is now, but I’m surprised to hear the FAA rejected
Reply to @manic_pixie_agi, 04:08 UTC · 2 likes
Corinne (🏠 in NYC) (@manic_pixie_agi): @AaronBergman18 is it borken again
Fcuk
Reply to @austinc3301, 04:10 UTC · 4 likes
Agus 🔸 (@austinc3301): don't drink the forbidden fluid
(quoting @lauriewired: All US Nuclear Reactor Incidents are public and posted online by the NRC. My Favorites: “The reactor cavity is full of water. [Individual] ingested some amount of cavity water.” (Michigan, 4 days ago) “Unit 2 power is being reduced from 100% in response to the influx of…)
Yum
Reply to @IsaacKing314, 06:03 UTC · 42 likes
Isaac King 🔎 (@IsaacKing314): Anyone who tests AI on their area of expertise rapidly realizes that AI is an idiot.
Unfortunately, it is becoming clear to me that many humans do not have an area of expertise.
I recommend this (ppl test AI on area of expertise) bc I think it’s calibrating but not reliably in the direction you’re suggesting
Maybe the most AI-pilled parts of Twitter, but most people have no idea how good frontier is and will not conclude that the AI is an idiot
Tweet, 16:11 UTC · 29 likes
Nooooo why are flight simulators so expensive nooooooo
Reply to @AndyMasley, 17:00 UTC · 9 likes
Andy Masley (@AndyMasley): The fact that falafel isn’t actually that good for you is something I know intellectually but haven’t accepted in my heart
lol I have always thought this was amusing
Tragically deep frying things is probably bad even if the thing you’re deep frying is chickpeas
Reply to his own earlier tweet, 17:46 UTC · 50 likes
Aaron Bergman, October 26, 2025: This is also nuts, although a bit less cleanly heroic https://en.wikipedia.org/wiki/Citicorp_Center_engineering_crisis
Oh wow there's a whole (and apparently pretty good) book https://a.co/d/73BAdVv
Thread, continued, 17:57 UTC · 43 likes
Honestly the choice to try to hit the runway at all is actually an interesting ethical dilemma (new trolley problem just dropped) - or at least could have been
Thread, continued, 18:00 UTC · 42 likes
Apparently the airport was 6 miles away from the city center which apparently is pretty significant distance - enough to mean they could be reasonably certain not to hit the city if they tried to hit the runway
But with other airports like LAX, this would have been recklessly dangerous
Thread, continued, 18:00 UTC · 110 likes
Thread, continued, 18:01 UTC · 75 likes
Based
Reply to @cljack, 18:43 UTC · 4 likes
Charlotte Lee (@cljack): @Charlesaf3 @AaronBergman18 I think the calculation was that total hydraulics loss is so rare that it’s just not worth any expense. Could’ve saved the Azerbaijan Airlines flight but I can’t think of any other examples tbh
Even if the FAA didn't mandate it, I'm a bit surprised major airlines haven't done it voluntarily
But at least according to this there's a major regulatory barrier, ugh
Monday, October 27, 2025
Reply to his own earlier tweet, 03:21 UTC · 2 likes
Aaron Bergman, October 25, 2025: Lol sorry for slowness, I am learning on the fly
It's officially broken until I figure out how to fix :(
Reply to @hamandcheese, 06:44 UTC · 56 likes
Samuel Hammond 🦉 (@hamandcheese): Melatonin in the US is sold in 5mg doses when the effective dosage range is .3-3mg. Americans essentially overdose on melatonin by default for no good reason.
I’ve come around to “1mg at least and the market probably knows something Scott Alexander doesn’t”
One model that would explain relatively large doses is “bioavailability is highly variable between ppl and nothing bad really happens if you take too much”
Thread, continued, 06:46 UTC · 5 likes
*replace “too much” with “more than a given person needs”
Reply to @AndyMasley, 07:00 UTC · 3 likes
Andy Masley (@AndyMasley): Feel like what's missing on this chart is that if a few companies are handing trillions of dollars back and forth to each other, it's identical to them all getting together to do a ton of free labor for the rest of us 😇
(quoting @ProofofMaro: )
I don’t think it’s identical. Counterexample: Company A pays company B to make them some computers, company B pays company A to give them some software. Both wind up with the same amount of money they started with and some new capital that they value more highly than the labor and capital they effectively traded away
No consumer has been affected yet - in principle they could proceed to light all their capital on fire, physically bury their cash, and fire all their workers
Quote tweet, 15:12 UTC · 6 likes
👀
Quoting Simon Bazelon (@simon_bazelon): It's time to renew the Democratic Party.
Introducing DECIDING TO WIN: the most comprehensive account yet of where Democrats went wrong, and what we need to do to win again.
With a year of research and tons of new data, by @laurenhpope, @liamkerr, and me.
Link and 🧵 below.
Reply to @dylanmatt, 16:53 UTC · 3 likes
dylan matthews 🔸 (@dylanmatt): Some professional news: after nearly twelve years at Vox, I'm going to be joining @open_phil, working with @mattsclancy and @jddwor on the Abundance and Growth Fund
👀
Reply to @dylanmatt, 16:53 UTC · 13 likes
dylan matthews 🔸 (@dylanmatt): Some professional news: after nearly twelve years at Vox, I'm going to be joining @open_phil, working with @mattsclancy and @jddwor on the Abundance and Growth Fund
I was like 98% sure the end of the tweet was gonna say “The Argument”
Tuesday, October 28, 2025
Reply to @RosieCampbell, 03:53 UTC · 6 likes
Rosie Campbell (@RosieCampbell): What's the best explanation for why mainstream philosophers of decision theory have basically ignored functional decision theory even though it seems better than causal or evidential decision theory?
If it's because FDT has an obvious flaw, what is the flaw?
This is not an answer but I’m not actually fully convinced it’s better than evidential
ik @NoahTopper is going to disown me as a friend over this
Reply to @NoahTopper, 03:57 UTC · 4 likes
Quinoah 🔍⏸️ (@NoahTopper): @AaronBergman18 @RosieCampbell (with love, kindness, and respect)
Bruh I should at least get to defend myself
Reply to @ciphergoth, 04:14 UTC · 2 likes
Paul Crowley (@ciphergoth): @NoahTopper @AaronBergman18 @RosieCampbell Roughly the trash was "you're not gonna believe this but I have a friend who believes in EDT"
Lmao u know what I can live with that
Also FWIW I haven’t like decided on it extremely strongly, I’m half expecting to be convinced of FDT at some point
Reply to @RosieCampbell, 04:29 UTC · 2 likes
Rosie Campbell (@RosieCampbell): What's your beef with FDT
Tbc I am not really a decision theory guy, I mean it’s smth I want to learn more about and get a stronger view on
But at least one crystallizing thing is that I don’t think you should smoke in smoking lesion problem
I think a lot of the intuitions ppl have in the other direction are dependent on things that would probably happen in an irl version of the thought experiment but *are not actually in the the standard formulation*
Like it doesn’t say anything about wanting to smoke, or knowing you’re the kind of person who would or wouldn’t smoke, or anything
My understanding is that, by construction, you don’t know whether you’re the kind of person who tends to choose to smoke or anything like that until the instant you decide
There are exactly two options/worlds:
1) smoke, probably have lesion
2) don’t smoke, probably not have lesion
Your choice *itself* shows you which world you’re in - there is no “best of both worlds” where you get to be the kind of person who tends not to smoke but also in fact smokes; the choice to smoke is the only evidence you get about what kind of person you are
Reply to @NoahTopper, 04:42 UTC
Quinoah 🔍⏸️ (@NoahTopper): I’m confused why this is mixing together “whether you really follow through on executing EDT” with “whether you have the lesion”. These seem unrelated.
Honestly I’m not super sure Claude is appropriately representing my position, maybe I should delete
But I think the chain of reasoning goes
Know you’re an EDT person -> know you wouldn’t smoke in the situation described -> get to the actual situation -> maybe you can smoke anyway bc u already know you’re the kind of person who chooses not to smoke?
But I argue “no, as soon as you make the decision to smoke, you learn that you were never in fact the type of person who tends to have whatever things cause “no lesion” - you can’t just do a quick switcharoo at the last minute without all the evidence flipping
Reply to @NoahTopper, 04:49 UTC
Quinoah 🔍⏸️ (@NoahTopper): Sounds like you’re rejecting the tickle defense if I understand right
Ya
Reply to @tszzl, 05:01 UTC · 30 likes
roon (@tszzl): “broad scientific consensus around superintelligence being safe” is commie nonsense and will always be used as a bludgeon imo
(quoting @deanwball: Vague statements like this, which fundamentally cannot be operationalized in policy but feel nice to sign, are counterproductive and silly. Just as they were two or so years ago, when we went through another cycle of nebulous AI-statement-signing. Let’s set aside the total lack…)
lol what?? “Broad scientific consensus” is a totally normal aspect of liberal democratic capitalist society
Normally you don’t need to make it so explicit because (1) liability takes care of the incentives, (2) normal consumer preference takes care of the incentives, (3) something something culture/norms make it implicit, or (4) more specific regulations/laws that don’t use that language but functionally enforce that standard
Tweet, 05:09 UTC · 696 likes
Lol should I be scared of Claude just casually writing a single Chinese character in a semantically/grammatically correct/sensible way in the middle of a fully English discussion
Thread, continued, 05:09 UTC · 127 likes
Reply to @RosieCampbell, 05:49 UTC
Rosie Campbell (@RosieCampbell): yeah I also do
Wait do I have things backwards??
Doesn’t FDT say “pay” ie insane NOT to pay up
Tweet, 06:08 UTC · 20 likes
Someone should figure out what’s up with the relationship between decision theory, free will, ethics, and maybe some secret nth things
Reply to @NoahTopper, 06:21 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): I have
Ok thx
Tweet, 15:02 UTC · 4 likes
The temptation to sling takes about http://decidingtowin.org without having read it all (yet) is crushing
Tweet, 15:05 UTC · 10 likes
I woke up today thirsty in the middle of a desert and then @NoahTopper conspicuously drove past me in a car???
Tweet, 16:13 UTC · 5 likes
Being in an Uber/Lyft with one of those discount iPads strapped to the headrest of the seat in front makes me feel bad for iPad babies
There is nothing interesting or fun or important on that screen but it feels hard to resist and basically impossible to fully ignore
Reply to his own earlier tweet, 16:16 UTC · 49 likes
Aaron Bergman, October 28, 2025: (image)
Are the mechanics here just that it’s one token instead of two so RL has taught to conserve?
Tweet, 16:21 UTC · 6 likes
I like the new genre of microsite (deciding to win, situational awareness)
It’s good that there’s a mechanism of saying “listen up this is important and I’ve spent a bunch of time on it”
Wednesday, October 29, 2025
Tweet, 00:05 UTC · 6 likes
Gil (aka @RFishBlueFish) and I will be debating illusionism in a Space at 9pm DC time/6pm SF time (i.e., in 55 minutes)!
Reply to @ilex_ulmus, 00:12 UTC · 30 likes
Holly ⏸️ Elmore (@ilex_ulmus): Everyone who pulls a little compliment out of the jaws of audacious evil-doing for OpenAI today gets blocked. You don’t get to do your little kneel before the invading king and come back to me. But thanks for outing yourself as bootlicker.
Offenders below👇
This isn't even a compliment!
Quote tweet, 14:36 UTC · 15 likes
New Hire Me doc just dropped
It’s still very premature and I don’t wanna jinx anything but so far so good on my end (context below)
It might just be the actually-good way of finding work
Quoting Cathy Reisenwitz (@CathyReisenwitz): Hi friends. I need a job! Ideally in Washington, D.C. or remote. My dream is to work in policy or comms at a think tank that is center-left or center-right. I’m interested in economic opportunity, energy, AI, finance, gender, and criminal justice. But I can learn anything quickly. I was just named Volunteer of the Year by the Center for New Liberalism. If I did that much work for free, can you imagine what I do for pay?? My writing on policy has been published widely, including in VICE Motherboard, Reason, OneZero, the Daily Beast, and other fine publications. My Substack newsletter is regularly a top-50 bestseller in the Humor category. I’m ready to start immediately and I’m an absolute bargain, if I do say so myself. I’m also happy to provide recommendations and testimonials from leading policy experts. If you have any leads, please email me at cathyreisenwitz at gmail dot com.
To read my clips and resume, head to cathyreisenwitz dot substack dot com/p/hire-me
Thread, continued, 14:37 UTC · 2 likes
Reply to @cljack, 14:39 UTC · 3 likes
Charlotte Lee (@cljack): @ChristianityOn LMAO shhhhhh
Thread, continued, 14:39 UTC · 2 likes
(Ofc maybe ur diagnosed in which case disregard)
Reply to @AOLien_k8, 14:51 UTC · 2 likes
AOLien (@AOLien_k8): how long before we go from "don't use Claude for therapy" to "that's really something you should only discuss with Claude"
Lmao already (there are a lot of things you shouldn’t say to eg psychiatrists and therapists)
Reply to @AOLien_k8, 14:57 UTC · 74 likes
AOLien (@AOLien_k8): My husband and I were married on June 6th, many moons ago. That first year, he set up a flower subscription so a bouquet would arrive on the 6th of every month. All these years later, I still get excited when I see those flowers at the front door. I am so very loved and so very in love.♥️
(quoting @vividvoid: Why is it so rare to see tweets about being in love? Yearning is nothing but desire collapsed on itself, a frittering of eros on the barren ground of lack. Where are the lovers, the obsessed, the people tearing sonnets out of their chests, the romantic unto death?)
Bro automated that shit
Reply to @Ruesavatar, 15:29 UTC · 9 likes
rue🌿 (@Ruesavatar): I dislike date me docs, they almost always come off as unnaturally performative and cringe. People often suggest I make one, but their mere existence is a red flag to me so it would not make sense to do so. 😅 Has anyone actually successfully found their spouse that way?
Red flag, really? This is kind insane
Reply to @Ruesavatar, 15:37 UTC · 4 likes
rue🌿 (@Ruesavatar): @AaronBergman18 Insane? We all have different tastes and standards 🤷♀️
Different tastes, sure, but I feel like “red flag” is a bit more of an objective thing
Reply to @sam_atis, 16:41 UTC · 2 likes
Sam Glover (@sam_atis): @AaronBergman18 @Ruesavatar Why would it be objective?
Wait name/face reveal big if true
But anyway I feel like the consensus use of that term is to indicate a broadly consensus set of traits or characteristics (honesty, character etc)? I can’t really disagree with someone about their preferences but I can disagree about what kind of evidence a given thing is for a broadly consensus cluster of traits
Thread, continued, 16:43 UTC · 4 likes
Obv ppl are gonna disagree about edge cases somewhat but it’s a genuine cluster IMO
Reply to @Noahpinion, 16:46 UTC · 59 likes
Noah Smith 🐇🇺🇸🇺🇦🇹🇼 (@Noahpinion): OK I support SNAP but WTF are you talking about here man
(quoting @GovTimWalz: Every $1 invested in SNAP generates $1.80 in economic activity. It’s not about the money, Trump just wants Americans to go hungry.)
Wait this doesn’t seem obviously crazy? I mean maybe it’s wrong but there are a bunch of mechanisms by which this could happen
Feeding kids -> higher school performance -> higher lifetime earnings
Lower job lock-in -> better matching in labor market
Reply to @Ruesavatar, 16:46 UTC · 2 likes
rue🌿 (@Ruesavatar): @AaronBergman18 @sam_atis Oh. I think I just use red flag more casually 😅
Ah haha gotcha
Reply to @Ruesavatar, 16:55 UTC · 3 likes
rue🌿 (@Ruesavatar): I dislike date me docs, they almost always come off as unnaturally performative and cringe. People often suggest I make one, but their mere existence is a red flag to me so it would not make sense to do so. 😅 Has anyone actually successfully found their spouse that way?
Also (not disagreeing here, just observing) I assume that many ppl you get anti-attracts to you wouldn’t have considered anyway if not for their dating doc?
If so it’s like kinda fine from the other person’s perspective
Reply to his own earlier tweet, 16:56 UTC · 5 likes
Aaron Bergman, October 29, 2025: Wait this doesn’t seem obviously crazy? I mean maybe it’s wrong but there are a bunch of mechanisms by which this could happen Feeding kids -> higher school…
Also like there’s some first order deadweight loss but to a first approximation it’s just a transfer. So you only need to explain ~1x rather than ~2x
Quote tweet, 22:26 UTC · 19 likes
So this got me thinking: the question of whether qualia can cause a person to have beliefs like “I have qualia” and say things to the same effect might be a much more confused question than I previously appreciated
Quoting Aaron Bergman 🔍 (@AaronBergman18): Gil (aka @RFishBlueFish) and I will be debating illusionism in a Space at 9pm DC time/6pm SF time (i.e., in 55 minutes)!
Thread, continued, 22:26 UTC · 7 likes
Like I think the question of qualia realism is substantive and the answer is either yes or no
But qualia causation just falls apart under scrutiny - what would it even mean for subjective experience *per se* to cause anything?
Thread, continued, 22:26 UTC · 6 likes
If qualia had causal properties, we would just say “no let’s decouple the causal properties from the raw subjectivity” and then we’re back to where we started
Thread, continued, 22:26 UTC · 5 likes
Tbc this is me coming out more confused than I went in
And tbh I guess I’m up to 5% (from like 3%) on “subjective experience does not exist”
Thread, continued, 22:27 UTC · 3 likes
Cc @RFishBlueFish @AndyMasley
Thursday, October 30, 2025
Reply to @bscholl, 05:58 UTC · 2 likes
Blake Scholl 🛫 (@bscholl): Xi’s 747 is two generations newer than POTUS’s.
This is a national embarrassment.
(quoting @GrantCardone: President Trump - meet with anyone anytime anywhere on behalf of make America Great Again. 🇺🇸)
No it’s not lol
Reply to @nickcammarata, 14:39 UTC · 7 likes
Nick (@nickcammarata): is there a simple explanation for why rationalism and sex work are so intertwined
Being able to talk about it analytically with fewer taboos
Reply to @thatMikeBishop, 14:49 UTC · 2 likes
Mike Bishop 🇨🇦🇺🇸 (@thatMikeBishop): People can define masculinity differently, and expectations clearly vary cross-culturally - I’m fine with that.
But science tells us testosterone has a lot to do with male-female differences, and that manipulating testosterone has many impacts.
Yeah I agree it’s an insane thing to call misinformation. Maybe they’re not less masculine in some ultimate metaphysical sense but hypogonadism is pretty well studied and like obviously correlates hard with what people mean by “less masculine”
Tweet, 15:25 UTC · 6 likes
So many apps wanna build in AI but like if you’re not paying a premium it’s gonna be a tiny shitty model and not actually provide much value
At least let users pay to upgrade to SOTA
Reply to @cljack, 16:52 UTC · 7 likes
Charlotte Lee (@cljack): Idk man. Even if you take the position that we shouldn’t give food stamps to able-bodied adults (which is defensible) a lot of people on both sides recognize that children have no control over their life circumstances and think it’s morally wrong to let them go hungry
(quoting @AnechoicMedia_: I feel like the only true conservative left in that my principle objection to food stamps is precisely that they pay for other people's kids. Republicans have lost the spine to defend right-wing values and so center their opposition to welfare on a hypothetical adult man sitting…)
And the steelman of much cold-hearted conservative anti-welfarism that it reduces economic growth which is what actually matters for wellbeing in non-myopic sense
I’m sure someone has made the case for this for children’s food but my sense is that it’s not a big offender on this front?
Reply to his own earlier tweet, 16:54 UTC · 6 likes
Aaron Bergman, October 30, 2025: So many apps wanna build in AI but like if you’re not paying a premium it’s gonna be a tiny shitty model and not actually provide much value At least let users…
Interestingly, what you don’t see is universal semantic search, which actually is probably cheap enough to run for free most of the time
Tweet, 18:22 UTC · 28 likes
This seems unnecessary
Friday, October 31, 2025
Tweet, 02:06 UTC · 54 likes
Plz clap 🦆 🎃
Thread, continued, 02:06 UTC · 5 likes
In the daylight ft. my arm 🦆🎃
Reply to @jeffzwang, 06:21 UTC · 8 likes
Jeffrey Wang (@jeffzwang): I throw good parties. Ahead of Halloweekend, here are 10 pro tips:
1. No bright lights. People feel way more comfortable when it’s dark or dim.
2. People always congregate where the drinks are. If you want people to dance, put the drinks near the dance floor. A huge mistake people make is putting drinks in the kitchen because it’s convenient - most kitchens are not good social spaces.
3. If people don’t drink at your party, it will not be as good. This isn’t meant to promote drinking - it’s simply a fact about parties and the reason why you often leave day parties with a lukewarm feeling.
4. (Twitter-specific tip) Don’t invite a large group of Twitter people to a party. I’ve only done this twice, but both times 80% of those people were super awkward, which ruined the party. This is particularly true of anyone that has over 10K followers.
5. Related to the above, a party needs social cohesion. Almost all the people should be separated by no more than 2 degrees for there to be a very comfortable drift between groups of people.
6. It’s your duty as a host to promote social interaction. This means you should be introducing guests to each other, especially if someone doesn’t know a lot of people. “Encouraging” people to drink is the other easy knob that turns up how much socializing occurs at a party.
7. Make sure there is music. In my opinion, the presence of music is more important than what kind of music. No music kills the vibe.
8. Having activities promotes social interaction (but don’t overdo it or your party will feel like a carnival).
9. You should have at least one quiet zone and one loud zone. This lets everyone have access to an area that matches their current energy.
10. Lastly, the people are the most important thing. If you have really good people, you can break all the rules and your party will still be great. (This does mean that you need to gatekeep the invite list. If you invite everyone you know, or cohost your party with someone whose friends you don’t like, you won’t have a really good group of people.)I hope this in particular helps people under the age of 25, who I have found are almost universally horrible at throwing parties.
And also jesus christ some people are so bad at throwing parties in SF, please print these tips out and tape them to your office’s kombucha dispenser.
Boo, invite all ur mutuals
Quote tweet, 06:54 UTC · 29 likes
PSA window fans like these exist, I just got one idk I waited so long
Now my room is never warmer than I want it to be
I really like this one bc both fans run both directions
Quoting 𝖜𝖔𝖒𝖇𝖑𝖊 𝖜𝖎𝖙𝖍𝖔𝖚𝖙 𝖆 𝖈𝖆𝖚𝖘𝖊 (@goblinodds): trying to find out precisely how insane it is that nyc apparently forces apartments to run the heaters inside when it's over 50 degrees outside
Thread, continued, 06:54 UTC · 6 likes
Seems to keep CO2 pretty low (this is a highly selected image but I think it does actually improve air quality)
Thread, continued, 06:54 UTC · 5 likes
Model is this one, I got used https://a.co/d/e2M9kBS
Quote tweet, 07:16 UTC · 124 likes
If you invite 10 new people to a party, 8 of them bring bad vibes, and one of the remaining two gets invited back indefinitely, and you keep doing this, you soon have a party of almost entirely new friends who you like
Quoting Jeffrey Wang (@jeffzwang): I throw good parties. Ahead of Halloweekend, here are 10 pro tips:
1. No bright lights. People feel way more comfortable when it’s dark or dim.
2. People always congregate where the drinks are. If you want people to dance, put the drinks near the dance floor. A huge mistake people make is putting drinks in the kitchen because it’s convenient - most kitchens are not good social spaces.
3. If people don’t drink at your party, it will not be as good. This isn’t meant to promote drinking - it’s simply a fact about parties and the reason why you often leave day parties with a lukewarm feeling.
4. (Twitter-specific tip) Don’t invite a large group of Twitter people to a party. I’ve only done this twice, but both times 80% of those people were super awkward, which ruined the party. This is particularly true of anyone that has over 10K followers.
5. Related to the above, a party needs social cohesion. Almost all the people should be separated by no more than 2 degrees for there to be a very comfortable drift between groups of people.
6. It’s your duty as a host to promote social interaction. This means you should be introducing guests to each other, especially if someone doesn’t know a lot of people. “Encouraging” people to drink is the other easy knob that turns up how much socializing occurs at a party.
7. Make sure there is music. In my opinion, the presence of music is more important than what kind of music. No music kills the vibe.
8. Having activities promotes social interaction (but don’t overdo it or your party will feel like a carnival).
9. You should have at least one quiet zone and one loud zone. This lets everyone have access to an area that matches their current energy.
10. Lastly, the people are the most important thing. If you have really good people, you can break all the rules and your party will still be great. (This does mean that you need to gatekeep the invite list. If you invite everyone you know, or cohost your party with someone whose friends you don’t like, you won’t have a really good group of people.)I hope this in particular helps people under the age of 25, who I have found are almost universally horrible at throwing parties.
And also jesus christ some people are so bad at throwing parties in SF, please print these tips out and tape them to your office’s kombucha dispenser.
Thread, continued, 07:16 UTC · 8 likes
Assuming you’re not going to die soon, the EV of each Twitter invite is positive
Thread, continued, 07:16 UTC · 8 likes
Obviously not close to a literally true model but there’s a true and important point about the asymmetry between positive and negative first interactions
Tweet, 15:11 UTC · 13 likes
Maybe I should actually do low dose steroids
Reply, 15:16 UTC · 1 like
Precisely
Reply, 16:18 UTC · 1 like
Being more muscular seems good on the margin but also I’m probably not actually gonna do this
Reply to @boreal_country, 16:33 UTC · 3 likes
Beau (@boreal_country): @AaronBergman18 These are great. Someone—not me—should make a version that includes/fits a HEPA filter.
Yes!!! @JesseTayRiver
Reply to @johnsmithyson0, 16:34 UTC · 3 likes
John Smith (@johnsmithyson0): @AaronBergman18 Have you checked all the other boxes for optimal growth first?
-creatine
-Training at least 4x week
-good sleep
-good eating?
Absolutely not (a real problem here is that I basically can't lift weights bc my core is too fucked up)
Reply to @_AashishReddy, 17:06 UTC · 3 likes
Aashish Reddy (@_AashishReddy): @AaronBergman18 @johnsmithyson0 What’s up specifically (if you’re okay saying), haven’t heard of anything like this before
Lol it is a very long story, check out this tab (the 'Health: much more than you wanted to know' bit, i think tabs might be weird on mobile)
Reply to @johnsmithyson0, 18:37 UTC · 2 likes
John Smith (@johnsmithyson0): @AaronBergman18 @_AashishReddy Man. That's rough. I have no idea how one would do a lot of lifting with that, given how important your core is
Thank you, yeah fr
Tweet, 18:39 UTC · 90 likes
Kinda weird that github is free? I feel like I should be paying to unlock more than 5 repos or something
Ik there actually is a premium tier but the free version is incredibly generous and useful
Reply to @gmiller, 19:19 UTC · 3 likes
Geoffrey Miller (@gmiller): Seems like a great opportunity -- except -- how can the UK hope to hire great talent for such a key position with a salary of only £33,015/year -- esp for London?
(Seems like that would barely cover minimal cost of living for a single person -- much less a family).
Yeah fr it’s insane, and plausibly part of the UK’s spiral (bad wages in govt -> worse talent -> worse governance -> worse economy -> worse wages)
But if you’re already very wealthy it sounds great lol
Reply to @jazzloaf, 19:20 UTC · 2 likes
vivian (@jazzloaf): I love how systematic you are, you're so cool
🥺
Reply to his own earlier tweet, 19:21 UTC · 4 likes
Aaron Bergman, October 31, 2025: Yeah fr it’s insane, and plausibly part of the UK’s spiral (bad wages in govt -> worse talent -> worse governance -> worse economy -> worse wages) But if…
Of course the other interpretation which I am extremely confident is at least sometimes true is that the people taking the jobs are extremely talented and could earn way more elsewhere and are doing it for the greater good
Reply to @goblinodds, 19:29 UTC · 2 likes
𝖜𝖔𝖒𝖇𝖑𝖊 𝖜𝖎𝖙𝖍𝖔𝖚𝖙 𝖆 𝖈𝖆𝖚𝖘𝖊 (@goblinodds): @AaronBergman18 oh this is nice! i dont have windows that open that way 💀 also where does ur AC go?
lol this is the current setup
I might remove the AC for the season but haven’t decided I want to commit to that bc it’s kind of annoying to re-install
The right side does open a bit above the fan though
Reply to @Devin_J_Monahan, 20:15 UTC · 8 likes
Devin Monahan (@Devin_J_Monahan): @AaronBergman18 If I were the guy getting invited back indefinitely, I think I might start declining after the fifth bad party.
Yeah fair, this is in my drafts lol
Quote tweet, 21:03 UTC · 13 likes
Probably correct take
Idk if Andy would endorse this but I generally also mask (n95) on any crowded transit eg metro
Quoting Andy Masley (@AndyMasley): Took 15 minutes to write out what I think is a super obvious take that imo more people should think is obvious
Tweet, 21:52 UTC · 24 likes
I feel like people don't appreciate how intense+high stakes the 5 seconds where I'm adding something to my google calendar is
I'm supposed to be listening to the other person still but I have to lock tf in bc if I don't accurately put the thing down it may as well not exist
End of October 2025. Previous month: September 2025. Next month: November 2025. Back to the archive index.