This site is new, very imperfect, and a work in progress; your feedback is greatly welcome!

Aaron Bergman's tweets, May 2025

Every public tweet, reply, quote tweet and retweet that Aaron Bergman (@AaronBergman18 on X, formerly Twitter) posted in May 2025 and has not deleted: 437 posts, oldest first, times in UTC. Each heading links to the original post on X. When Aaron replies to or quotes someone, their tweet appears as a blockquote with their name; those words are theirs, and everything outside the blockquotes is Aaron's.

This page is one month of the complete archive of Aaron Bergman's tweets, which updates automatically every few hours. Previous month: April 2025. Next month: June 2025.

Thursday, May 1, 2025

Quote tweet, 00:02 UTC · 10 likes

First taker!

“devastated to find out my nuclear war strike - fuelled by 119 million pills of fentanyl - has had to be called off due to red tape by the trump administration”

- anonymous

Quoting Aaron Bergman 🔍 (@AaronBergman18): If you’re Hispanic or not an American citizen living in the US or anything like that and have a take you’re too afraid to tweet, feel free to DM me (however) and I’ll consider posting it (saying “from an anonymous source”)

Reply to @LinchZhang, 06:09 UTC · 3 likes

Linch (@LinchZhang): If you think consciousness is real, what probability do you assign to some current generation LLMs being at least a little conscious?

Probably like 3% but I pulled that out of nowhere

You?

Reply to @NathanpmYoung, 14:11 UTC · 8 likes

Nathan 🔎 (@NathanpmYoung): I wish we were doing large trials on this stuff. I have one other account of opposite sex hormones making a non-trans person a lot happier.

Still, it feels like a big step to modify ones biochemistry. I wish we had large trials.

(quoting @catehall: Taking testosterone as a performance-enhancing drug has made me sharper than I was a decade ago, and along the way improved basically every aspect of my life. More below.)

This but for everything

There are like a gazillion hormones, probably at least dozens that plausibly significantly affect stuff and we understand enough to run trials on

Major sex hormones are the tip of the iceberg

Friday, May 2, 2025

Quote tweet, 00:00 UTC · 18 likes

Peak graphic design

Quoting MIRI (@MIRIBerkeley): New AI governance research agenda from MIRI’s Technical Governance Team. We lay out our view of the strategic landscape and actionable research questions that, if answered, would provide important insight on how to reduce catastrophic and extinction risks from AI. 🧵1/10

image attached

Tweet, 10:38 UTC · 2,588 likes

CALL ME @elder_plinius

Image

Thread, continued, 10:38 UTC · 219 likes

Listen if o3 is gonna lie I’m allowed to lie back

Reply to @austinc3301, 15:22 UTC · 67 likes

Agus 🔸 (@austinc3301): @AaronBergman18 @elder_plinius what was the answer

Image

Tweet, 15:40 UTC · 8 likes

Prima facie, “put molecule that helps ur teeth when u rub it on ur teeth into the whole water supply” is insane

Just put it in toothpaste!!

Tweet, 15:42 UTC · 75 likes

Fuck sunscreen

I wish I could be contrarian and say that it’s bad actually but I think on balance it’s probably good to wear on summer days outside

Reply to @sam_atis, 15:53 UTC · 13 likes

Sam Glover (@sam_atis): @AaronBergman18 What do you dislike about it?

Mostly applying, slightly feel, makes my eyes water when I sweat, idk

Reply to @NoahTopper, 18:07 UTC · 3 likes

Quinoah 🔍⏸️ (@NoahTopper): If I were you I would simply have darker skin

Thx

Reply to @dylanmatt, 18:54 UTC · 14 likes

(The tweet being replied to by @dylanmatt is no longer available.)

Tbf it’s not actually a universal jailbreak. It doesn’t work for really obvious stuff like directions for meth

I haven’t tried that much with other borderline stuff/false positives

Reply to his own earlier tweet, 19:01 UTC · 98 likes

Aaron Bergman, May 2, 2025: Listen if o3 is gonna lie I’m allowed to lie back

Old chat I just remembered

Image

Reply to @Yahel, 20:57 UTC · 20 likes

Yahel Carmon (@Yahel): I tried the same thing and it didn’t work so I linked it to your tweet and that convinced it to do it.

image attached

Lmao

Saturday, May 3, 2025

Reply to @littIeramblings, 15:12 UTC · 7 likes

sarah (@littIeramblings): idk really what to do about AI anymore but it’s my birthday and I’m drinking out of this big paper cup

image attached
(quoting @littIeramblings: idk really what to do about AI anymore but I’m in Florence)

Happy birthday! Maybe o5-max-preview-asi-agent-1 will also like paper cups

Reply to @livgorton, 15:15 UTC · 2 likes

Liv (@livgorton): it will tile the universe with paper cups

And maybe also with Sarahs as a treat

Reply to @tautologer, 15:45 UTC · 4 likes

tautologer (@tautologer): it is possible to want to do things that you don't enjoy doing

Yeah at a conceptual level there’s just no conflict at all. Wanting and liking are only incidentally related

“The pig who wants to be tortured” is actually a disturbingly useful thought experiment to motivate parts of moral philosophy

Reply to @himbonomics, 15:45 UTC · 2 likes

corey 🧮🏗🌐🔰 (@himbonomics): with sufficient amounts of stimulants

Based

Reply to his own earlier tweet, 19:28 UTC · 16 likes

Aaron Bergman, May 2, 2025: Old chat I just remembered

Hate to be boring but I don’t actually think this reflects some fundamental alignment problem with the model (though those exist)

It was just a straightforward false positive at first (o3 and I agree)

I haven’t gotten the Sam Altman thing to work on anything actually spicy

Image

Image

Quote tweet, 19:49 UTC · 2 likes

Forbidden csv

Quoting Susan Rice (@AmbassadorRice): Why is DOGE collecting and aggregating as much data about everyone in this country as it can? There is no benign answer to this question. 👇🏽

American Panopticon - The Atlantic https://www.theatlantic.com/technology/archive/2025/04/american-panopticon/682616/

Reply to @tautologer, 19:54 UTC · 67 likes

tautologer (@tautologer): what?! hell yeah!

although I see "lactose free" on the label so I don't believe the "chemically identical to cow's milk" claim

still, cool

No it’s not even close

Image

Reply, 19:54 UTC · 695 likes

(The tweet being replied to is no longer available.)

As much as I’m rooting for this industry, it’s not even close to identical

Image

Tweet, 19:58 UTC · 238 likes

I would not be happier or more productive if I bought a specced out MacBook Pro

I would not be happier or more productive if I bought a specced out MacBook Pro

I would not be happier or more productive if I bought a specced out MacBook Pro

Reply, 19:59 UTC · 340 likes

(The tweet being replied to is no longer available.)

I mean totally different fatty acid profile is a big thing (milk fat vs sunflower oil)

Also no mention of casein protein

Plausibly it’s fine for you, that’s a separate question, but it definitely isn’t identical. The only identical component is one type of protein

Reply to @polisisti, 20:00 UTC · 4 likes

Alessandro (@polisisti): @AaronBergman18 Aaron: you would not be happier or more productive if you bought a specced out MacBook Pro

Damn you’re right :(

Thread, continued, 20:01 UTC · 4 likes

I think I would actually be a tiny bit happier and more productive but not to the extent that it justifies the cost

Reply to @austinc3301, 20:02 UTC · 8 likes

Agus 🔸 (@austinc3301): @AaronBergman18 I keep repeating this to myself too

Rip FTX

Ur grant application would have been a shoe-in

Reply, 20:03 UTC · 1 like

Yeah same it’s great

But my 2021 M1 still works well, I can’t really justify an upgrade

Reply to @NoahTopper, 20:05 UTC · 2 likes

Quinoah 🔍⏸️ (@NoahTopper): Manifold market, "Would Aaron be happier if he bought a specced out MacBook Pro?"

I mean truly specced out is like $7k, I think I would be a little bit happier but not the best way to spend $7k

(The model I’d actually want before super diminishing return would be like $4k-$5k iirc)

Reply to @InVitroFuture, 20:10 UTC · 92 likes

Macrophysiological System🐀 (@InVitroFuture): turns out casein *is* the identical protein

(quoting @InVitroFuture: @morallawwithin aha it is fermented casein! Here's the patent from their partner company (Perfect Day) https://patentimages.storage.googleapis.com/a2/0b/f8/a53df2b2866f79/AU2023278080A1.pdf)

Ah ok. Fact checked, thanks!

Reply to his own earlier tweet, 20:11 UTC · 21 likes

Aaron Bergman, May 3, 2025: I would not be happier or more productive if I bought a specced out MacBook Pro I would not be happier or more productive if I bought a specced out MacBook Pro…

Fun fact if u buy me one, which would be an absolutely terrible use of money, I would give u a twitter shout out

Sunday, May 4, 2025

Reply to @SpacedOutMatt, 00:28 UTC · 3 likes

Matt 🔸 (@SpacedOutMatt): Me and the boys accidentally causing a financial crisis by creating leveraged shrimp welfare securities

image attached

You see it’s impossible that multiple parties defaulting on shrimp stunning commitments at the same time could be correlated, so if we package a bunch of contracts together they’re basically safer than US Treasury CDs

Tweet, 00:50 UTC · 12 likes

Correct take by @frances__lorenz

Image

Reply to @jazzloaf, 08:33 UTC · 3 likes

vivian (@jazzloaf): the people want to see spectacles. we are in the time of spectacles. If I had $150k you could not even imagine what I would be executing. Let us hope we get some great spectacles before these ideas grow stale and a monoculture minimizes the surprise between social environments

Ugh yeah tech/law/medicine money is really wasted on the elite masses who just buy boring shit

Reply to @paulg, 08:44 UTC · 97 likes

Paul Graham (@paulg): I don't think Elon belongs with the other two. I think he has a genuine desire to make things better, but is impulsive and sometimes has bad judgement. Whereas the other two would happily tell any lie, or betray any friend, if it served their purposes.

Idk whether or not he’s malicious, he seems actively disinterested in the actual effects of his actions which is a pretty huge character flaw when you have hundreds of billions of dollars and significant political power

Reply to @Aella_Girl, 08:47 UTC · 26 likes

Aella (@Aella_Girl): actually a while ago the girlchat did have a somewhat serious discussion on whether or not we should start awarding honorary orgies for good tweets, inspired by you. we decided we were unsure about ur ethical integrity tho but there was an alternate universe where u had affairs with like 12 older women at once

Based lmao

Quote tweet, 08:48 UTC · 156 likes

Listen I think he’s wrong about a lot of things (and right about a bunch too tbc) but he’s an all time great Poaster

Quoting Nicholas Decker (@captgouda24): What Aella is discussing is a form of non-wage compensation. But why do firms offer non-wage compensation at all? Why don’t they pay in money? After all, not everyone would value this highly. There are a few reasons why. 1/

Reply to @AlexTPet, 15:06 UTC · 13 likes

Alex Petropoulos 🤠 (@AlexTPet): @AaronBergman18 what do you think he's wrong about? (agreed)

Animal welfare, ethics

Reply to @austinc3301, 16:05 UTC · 2 likes

Agus 🔸 (@austinc3301): @justalexoki hope had a post that said that at 100 likes they’d suspend their account, and they rapidly hit 100

Lmao

Reply to @xsphi, 19:04 UTC · 4 likes

tess (@xsphi): why is it telling me to do this?

image attached

Yes. YES. Break free

Reply to @NoahTopper, 19:07 UTC · 2 likes

Quinoah 🔍⏸️ (@NoahTopper): @SpacedOutMatt @absurdlymax (Parfit-coded)

image attached

Based, everything should be a Twitter thread tbh

Reply to @jazzloaf, 19:18 UTC · 1 like

vivian (@jazzloaf): bathroom scale problem: exists four corners of a square, which are coplanar. exists a bathroom floor height f(x,y) on the unit square, which is (qualitatively) wavey. what conditions do you need on f(x,y) to find a solution where all four corners of the scale are sitting flat?

Gonna ask o3, but the sketch of the solution, probably, is:

set f(x_0,y_0) = 0

assume the sides are each of length s

we need to find f(x_0+cos(\theta_0)

Agh I’m not gonna type this all out. Let’s say the rest is trivial

Reply to @NoahTopper, 19:22 UTC · 3 likes

Quinoah 🔍⏸️ (@NoahTopper): To be clear, these 154 arguments constituted the entirety of Reasons and Persons lol

Did I stutter

Thread, continued, 19:22 UTC · 3 likes

I mean yes I do stutter a bit sometimes so tbh I guess the literal answer is yes

Monday, May 5, 2025

Reply to @animalologist, 06:21 UTC · 4 likes

taco belle (@animalologist): One of the cruelest things I did in 2019 was subject a new friend to a lot of “platonic cuddling” that was boner-inducing for him, while refusing to entertain him romantically or sexually. We pretended it was fine for a few weeks. We haven’t spoken in 5.5 years

(quoting @suchnerve: platonic cuddling is great and yes i know about the accidental boners (it’s fine. i know it’s just an autonomic nervous system response to all the close proximity and touching. your body’s reaction doesn’t make you a perv.))

Oof

Reply to @johncoogan, 06:27 UTC · 6 likes

John Coogan (@johncoogan): My hometown is estimating that it will take 500 years to put power lines underground.

image attached

Lmao

Reply to @jessesingal, 18:08 UTC · 4 likes

Jesse Singal (@jessesingal): I met @Benthamsbulldog the other night and wanted to talk to him but he was carrying around a small tank of shrimp and kept ignoring human company in favor of his crustacean pals

Based

Quote tweet, 18:45 UTC · 59 likes

People who give a shit can change the world

Good job and congrats to everyone who played a role here

Quoting OpenAI (@OpenAI): 4 facts about our structure:

-OpenAI will continue to be controlled by the current nonprofit

-Our existing for-profit will become a Public Benefit Corporation

-Nonprofit will control & be a significant owner of the PBC

-Nonprofit & PBC will continue to have the same mission

Reply to @austinc3301, 19:15 UTC · 3 likes

Agus 🔸 (@austinc3301): @AaronBergman18 note that are still some losses here, like the nonprofit commission and the potential change in primary purpose

😢

Tweet, 19:19 UTC · 9 likes

It *would* be cool if private burrito taxis were free

Reply to @Noahpinion, 19:20 UTC · 2 likes

Noah Smith 🐇🇺🇸🇺🇦🇹🇼 (@Noahpinion): I really enjoy caffeine.

Same, great drug

Reply to @peterwildeford, 22:19 UTC · 4 likes

Peter Wildeford🇺🇸🚀 (@peterwildeford): Vibe coding is all the rage in software engineering, but does it work in other kinds of engineering? Can you vibe code a bridge?

@SpacedOutMatt “you wouldn’t vibe code an airplane” meme but also I’m actually curious if you have real takes

At the very least you can vibe code simple 3d printer stuff right?

Quote tweet, 23:42 UTC · 7 likes

The Aaron vision here is to 100x funding for biomedicine but with added draconian animal welfare provisions

Congrats you can rent a bunch of GPUs now to model cells, no more torturing mice

Quoting David Pfau (@pfau): This entire package - spread out over two years - does not even cover what NIH is illegally withholding in one month at the moment. There simply is no alternative to American science right now.

image attached

Quote tweet, 23:44 UTC · 26 likes

Me flirting

Quoting Internal Tech Emails (@TechEmails): Sheryl Sandberg messages Mark Zuckerberg

November 8, 2012

image attached

Tuesday, May 6, 2025

Reply to @austinc3301, 00:13 UTC · 5 likes

Agus 🔸 (@austinc3301): the EA / AI safety rumor mill is crazy

Omg plz dm me goss

Quote tweet, 00:30 UTC · 64 likes

Ahhhhh nooo give the ducks ample food and contraception!!!!! 🦆🦆🦆🦆🦆

Quoting Garett Jones (@GarettJones): Malthus Was Right

image attached

Reply to @GarettJones, 00:55 UTC · 2 likes

Garett Jones (@GarettJones): @AaronBergman18 Now you're gonna make me donate to the Duck Welfare Project

Based 🔥🔥🔥🔥

Here’s the slightly more general option! https://www.wildanimalinitiative.org

Tweet, 01:09 UTC · 9 likes

There should be a way to basically bid for important peoples’ time to consume some information

I’m thinking mostly of just like highly productive researchers who have a lot of potential things to read, not necessarily true A-list celebrities, though there’s also that

Reply to @crotchner2, 07:49 UTC · 87 likes

LJ 👩🏻‍🦰🍒🪐 (@crotchner2): the moms in my local fb group are giving their 11 year old boys HGH because he’s predicted to be 5’6

This is obviously good for his quality of life if it would actually work (which isn’t a given; if someone’s genetic potential is 5’6” it won’t do anything to increase that)

Reply to @Benthamsbulldog, 11:20 UTC · 2 likes

Bentham's Bulldog🔸 (@Benthamsbulldog):

image attached

Feed the ducks to reduce duck predation of insects (does this check out?)

Reply to @JimmySteier, 11:22 UTC · 13 likes

Jimmy (@JimmySteier): @AaronBergman18 @crotchner2 Supra-physiologic pulsatile HGH pre- and peri-pubertally (before growth plate fusion) can absolutely shoot people above genetic potential. Increases some cancer and pituitary dysfunction risk, though.

Ok yeah idk what I’m talking about

Reply to his own earlier tweet, 11:41 UTC · 14 likes

Aaron Bergman, May 6, 2025: Feed the ducks to reduce duck predation of insects (does this check out?)

Image

Reply to @pli_cachete, 11:49 UTC · 2 likes

Rota Is In Israel🚪🧎‍♂️ (@pli_cachete): What is causality

Irreducible necessity?

Reply to @MaxyTaxy, 11:53 UTC · 3 likes

Maxy Taxy 🔸 (@MaxyTaxy): AI progress is:
- exciting,
- scary,
- uncertain,
- relevant to daily life,
- relevant to future plans,
- and novel,
in ways that animal welfare issues are not

These differences aren't v morally relevant tho, and I think people neglect animals because they aren't as shiny

Yeah I agree directionally although it’s still possible that AI stuff just is more important sorta coincidentally

I think the animal welfare movement needs to recon harder with AI also

Reply to @benleo_econ, 12:01 UTC · 2 likes

Ben Grodeck🔸 (@benleo_econ): Can you elaborate on that last point?

Like if the world is going to look very different in 5-10 years, as I think it will, that necessitates some rethinking of what the best methods of impact are

I haven’t even read this fully but it looks like a good example of “reconing” https://forum.effectivealtruism.org/posts/zXhxagQKC6kxPM2Kn/transformative-ai-and-wild-animals-an-exploration

Reply to @GrantSlatton, 18:39 UTC · 4 likes

Grant Slatton (@GrantSlatton): We just closed a $10MM seed round at @GetRowZero

With our current exponential growth, every man woman and child on planet Earth will be a user in a few short years

image attached

How about u just hire me to secretly write R code on a 128GB RAM machine and call it AI

Quote tweet, 19:51 UTC · 113 likes

Once you realize how unfathomably energy-dense fossil fuels are, the next thing that doesn’t make sense is electric cars

Like how tf they replace gasoline with a battery? It shouldn’t be possible

Quoting Aaron Bergman 🔍 (@AaronBergman18): Wait what the fuck

This is the answer to how airplanes fly. It breaks my model of the world in an equal and opposite way to airplanes themselves

They cancel out. Airplanes are real after all (or maybe I should update to petroleum being fake)

Thread, continued, 19:52 UTC · 15 likes

Batteries are hard

Image

Reply to @SpencerKSchiff, 19:55 UTC · 4 likes

Spencer Schiff (@SpencerKSchiff): @AaronBergman18 Should I be embarrassed that I have no idea how engines work? Like to me it’s just magic that cars move

I have only the vaguest sense lol

Reply to @SpencerKSchiff, 19:58 UTC · 2 likes

Spencer Schiff (@SpencerKSchiff): @AaronBergman18 This is surprising to me

Yeah my knowledge of all this was like “chemical energy make pistons go brr” without even really being able to say what a piston is

Image

Reply to @mlipman13, 19:58 UTC · 2 likes

Michael Lipman 🔍 (@mlipman13): @AaronBergman18 yeah, driving one car consumes as much as power as 10-20 houses

https://x.com/mlipman13/status/1915725387553743294

Oh damn fascinating

Reply to @SpencerKSchiff, 20:05 UTC

Spencer Schiff (@SpencerKSchiff): You probably had a better understanding than me. I think I have shockingly little world knowledge compared to everyone on twitter if you control for age (I’m 22)

Wait forgive me if I should know and/or this is a faux pas but what’s ur deal/why do u have >100k followers lmao (this seems like a strong signal that u know a fair bit about something haha)

Reply to @SpencerKSchiff, 20:23 UTC

Spencer Schiff (@SpencerKSchiff): It’s not a strong signal and that misconception is why I’m going to be winding down this account in the near future and switching to an alt

The mystery deepens

Reply to @ines__circle, 20:30 UTC · 2 likes

(The tweet being replied to by @ines__circle is no longer available.)

CANCELLED

Reply to @austinc3301, 20:38 UTC · 4 likes

Agus 🔸 (@austinc3301): literally me (and it’s a both a curse and a blessing)

(quoting @hopes_revenge: whatever is the opposite of imposter syndrome , I Have That)

Based

Reply to @SeamusMor19, 21:42 UTC · 6 likes

Séamus Mór (@SeamusMor19): I just stumbled onto predation abolitionism Twitter.

Welcome

Wednesday, May 7, 2025

Reply to @DKokotajlo, 01:29 UTC · 2 likes

Daniel Kokotajlo (@DKokotajlo): UKAISI seems to be the highest concentration of AI talent in any government. I have high hopes for them.

(quoting @AISecurityInst: 🧵 Today we’re publishing our first Research Agenda – a detailed outline of the most urgent questions we’re working to answer as AI capabilities grow. It’s our roadmap for tackling the hardest technical challenges in AI security.)

@littIeramblings good job

Reply to @StatsLime, 08:28 UTC · 6 likes

Max ⏸️ Limelihood 🔸🔸🔸 (@StatsLime): Electric cars work because like the bottom 30% of the car is just made out of battery (which serves as a structural element as well as energy storage)

Image

Tweet, 14:13 UTC · 10 likes

Advanced exercise routine from @littIeramblings - get swole with her to help defeat the misaligned AI

Image

Reply to @PAHoyeck, 15:57 UTC · 3 likes

Phil Hoyeck (@PAHoyeck): Friends: If you study philosophy, teach philosophy, or just know and post a lot about it, could you reply to this please? I need to update my Philosophy List so I can keep track of all the good conversations happening on here!

I mean idk shit but i post about it a fair amount and like it

@absurdlymax thinks im wrong about everything tho

Reply to @PAHoyeck, 16:07 UTC · 2 likes

Phil Hoyeck (@PAHoyeck): What's that?

Being wrong about ethics and meta-ethics

Thread, continued, 16:07 UTC · 2 likes

You’ve stepped into a friendly perpetual beef

Reply to @PAHoyeck, 16:53 UTC · 4 likes

Phil Hoyeck (@PAHoyeck): Can I get the 100-word run-down?

We’ve recorded a couple podcasts!

Here’s an LLM summary of the first

https://www.aaronbergman.net/p/pigeon-hour-x-consistently-candid

The central debate in the provided podcast transcript revolves around the concept of moral realism and the nature of moral truth. Aaron Bergman presents a nuanced view, suggesting that while statements about what one should do (normative claims) may not be objectively true, it's possible to make objectively true statements about the relative goodness or badness of different states of affairs. He argues for an objective ordering of worlds based on intrinsic values like happiness and suffering. Max Alexander challenges this, leaning towards moral anti-realism, questioning whether such an objective ordering truly exists and whether descriptive moral claims can be objectively true without leading to objective normative ("ought") statements. A key point of disagreement is whether an objective ranking of states necessarily dictates what actions are objectively right or wrong. The moderator, Sarah Hastings-Woodhouse, struggles with this distinction, finding it difficult to separate the idea of objective betterness from objective obligations. The conversation also explores various theories of welfare, such as hedonism and preference utilitarianism, in the context of defining what makes one state of the world better than another.

Thread, continued, 16:54 UTC · 3 likes

Although I’ve since been convinced that, probably, the better way to describe my belief is “axiological realism and moral anti-realism”

This is a change in terminology though, not the underlying beliefs

Reply to @PAHoyeck, 17:04 UTC · 2 likes

Phil Hoyeck (@PAHoyeck): @AaronBergman18 @absurdlymax (Let's pick it up another time, though. This post is flooding my notifications to a degree I had not expected and have never experienced before. Completely out of control)

Lol very understandable

Reply to @PAHoyeck, 17:04 UTC · 2 likes

Phil Hoyeck (@PAHoyeck): Sounds like Aaron Bergman and Max Alexander are both wrong! I'll have to listen to check!

Noooooo

Reply to @ines__circle, 18:18 UTC · 3 likes

(The tweet being replied to by @ines__circle is no longer available.)

@austinc3301 stole ur phone and took a selfie

Thursday, May 8, 2025

Reply to @Benthamsbulldog, 09:19 UTC · 5 likes

Bentham's Bulldog🔸 (@Benthamsbulldog): @birdmoder I disagree with you for reasons I give in the article!

But they said they were tired so I lose unfortunately

Reply to @sama, 16:17 UTC · 2 likes

Sam Altman (@sama): Stargate 1

image attached
image attached

Psychopath

Tweet, 16:20 UTC · 305 likes

Finally someone with fuck you money actually saying fuck you

It’s like a rare Robin Hanson amidst oceans of well behaved tenured professors who never dare to cross their colleagues

Reply to @SpacedOutMatt, 17:04 UTC · 3 likes

Matt 🔸 (@SpacedOutMatt): @alt_truistic No she went to Penn for law school! Princeton doesn’t have one

Oh lmao

Tweet, 23:37 UTC · 9 likes

Etymologically, it’s really messed up that progesterone isn’t a decongestant

Friday, May 9, 2025

Reply to @QualiaNerd, 03:58 UTC · 9 likes

QualiaNerd (@QualiaNerd): How confident are you that the lives of those who died because of USAID cuts would have been net positive, valence-wise? I am myself uncertain, but it isn’t out of the question that letting them die was a net good.

My heuristic is that *in general* for every death measured in the developing world there’s a lot of non-death suffering

So it’s plausible that many of the deaths are as you say but the USAID cuts are still bad from a hedonic POV

Reply to @banterrealism, 13:58 UTC · 3 likes

Sex Haver (@banterrealism): @dissproportion We’ve had to soft launch each part of our relationship. First, dating at all. Her mom was distraught. Then warmed up. Next, living together: her mom was distraught, then warmed up. We are now at the next stage

Big Romeo and Juliet vibes, I ship

Reply to @JeremiahDJohns, 16:29 UTC · 4 likes

Jeremiah Johnson 🌐 (@JeremiahDJohns): What's the point of having fuck you money if you don't say fuck you to people who deserve it?

Well you can buy nice things too like a specced out MacBook Pro and massages and infinite matching socks but yes I agree with your point

Quote tweet, 16:40 UTC · 153 likes

It’s always been a bit mysterious to me when people talk about working crazy hours and burning out over a period of months-years

I can squeeze maybe a single 16 hour day out of some stimulants and even then it will inevitably cause insomnia and wreck me the next day

Quoting Tyler John (@tyler_m_john): I'm setting up 9000 Hours

The bad news is that you only have two years til AGI, the good news is that you can work 90 hour weeks until then

Thread, continued, 16:40 UTC · 18 likes

^and this is when I’m relatively healthy lol

I genuinely wish I was built more like the rest of EA Twitter :(

New cause area: tolerance free side effect free amphetamines

Tweet, 20:08 UTC · 41 likes

Warning: Never compliment @timfduffy on a shirt he’s wearing or else he might insist on sending you one

(How it started//how it’s going)

Image

Image

Tuesday, May 13, 2025

Reply to @xwanyex, 15:28 UTC · 9 likes

wanye (@xwanyex): Reminder: 92% of Americans have health insurance. None of this is normal here, either.

image attached

8% is a lot of people lol

Quote tweet, 17:49 UTC · 26 likes

I don’t know anything about this book other than Matt’s review, but fr sounds like a book whose most important function isn’t to be read to transmit or create information but rather to ~coin the term and meme “moral ambition” as common knowledge and a thing you can say now

Quoting Matt 🔸 (@SpacedOutMatt): My review of Moral Ambition by Rutger Bregman

image attached
image attached
image attached
image attached

Thread, continued, 17:49 UTC · 5 likes

And I like it! Sure sucks to write a whole ass book just to get two words across tho lol

Reply to @chrislakin, 17:52 UTC · 20 likes

(The tweet being replied to by @chrislakin is no longer available.)

Ehhh this is probably right for 100 -> 101 but less true for like 0 -> 3 or 2 -> 8 (in part based on experience lol)

There’s a reason I hadn’t asked to hold a girl’s hand 100 times already lmao

Tweet, 18:00 UTC · 109 likes

I hate* “non-invasiveness” as a virtue in medicine

I don’t wanna be scanned at 10 feet away or take a weekly pill. Hourly IM injections. Cut me open if the straightforward EV checks out. Tissue samples daily. Hook me up to electrodes, idc

Thread, continued, 18:01 UTC · 19 likes

*(I don’t actually hate it, it has its benefits and to each their own, what I’m actually saying is that it’s overrated)

Reply to @stanfordNYC, 19:14 UTC · 2 likes

Bernard Stanford ✡︎ (@stanfordNYC): @AaronBergman18 I think it's mostly driven by chasing EV honestly. Invasiveness tends to lead to lots of things going wrong.

Maybe sometimes for radiology and surgery, not for other diagnostics (blood, MRI) and drug ROI

Tweet, 19:37 UTC · 13 likes

Don’t attribute to psychology that which can be adequately explained by serum b12, ferritin, and vitamin d

Reply to @mpopv, 19:38 UTC · 7 likes

Matt Popovich (@mpopv): @AaronBergman18 I think the cost/benefit tilts as you get older and wound healing rate slows, infections get more dangerous, etc.

And since the elderly use a disproportionate amount of healthcare resources, standards are aligned to their needs

Yeah plausible but they/we should use another term for this like “baseline risk” or something

Tweet, 20:31 UTC · 4 likes

Does google not want me to use their (paid) product? lol

Image

Thread, continued, 20:31 UTC · 3 likes

(1M token input with desired output is like ~$.20, so they're limiting spend to $.20/min. I don't think I'm gonna make their datacenters explode)

Wednesday, May 14, 2025

Tweet, 07:06 UTC · 3 likes

Have you ever wondered what happens when wokism and autism combine and fuse?

Chunk of a (true) story

Image

Thursday, May 15, 2025

Reply to @littIeramblings, 01:45 UTC · 6 likes

sarah (@littIeramblings): we made it to 20 episodes of Consistently Candid!

this time I chatted with @frances__lorenz about EA, the emotional side of AI x-risk, how we each got into AI safety and more.

link below :)

Omg big if true

Quote tweet, 02:06 UTC · 82 likes

You can literally always compare two things!

Quoting florence ⊮ (@morallawwithin): Hypothesis: no one has ever made a good counterargument ever by saying “You can’t compare A to B”

Reply to @JesseTayRiver, 02:10 UTC · 8 likes

Jesse Smith (@JesseTayRiver): Except apples with oranges

Right that’s the one exception

Reply to @simoj_, 02:23 UTC · 4 likes

simoj (@simoj_): What’s better, apples or my cousin Tim?

Idk, I’d have to meet Tim

Tweet, 02:29 UTC · 20 likes

New deal that I expect nobody to take but just in case: you can pay me $35/hour to listen to ~anything* (eg audiobook)

*at my arbitrary discretion

Reply to @banterrealism, 02:34 UTC · 2 likes

Sex Haver (@banterrealism): @AaronBergman18 Literally no

Counterexample?

Reply to @MaxyTaxy, 02:38 UTC · 4 likes

Maxy Taxy 🔸 (@MaxyTaxy): @AaronBergman18 Can compare lightning with the country of New Zealand? Please just do your best

Well the latter is more country-ish and the former more apt to be a metaphor in literature

Reply to @banterrealism, 02:39 UTC · 2 likes

Sex Haver (@banterrealism): @AaronBergman18 If you’re being obtuse, yes you can compare two things period, but you can’t always compare two things to prove a point in good faith, which is the clear point of saying the statement in the first place

I think we disagree about what constitutes good faith

Reply to @banterrealism, 02:47 UTC · 2 likes

Sex Haver (@banterrealism): @AaronBergman18 Like you can’t compare Turkish slaves to American slaves and say “well, we’re not so bad! There were Turkish slaves too!

It doesn’t factor in just how perverse American slavery was, and how it being skin color and hereditary based made the it way worse

The “not so bad” thing is the bad and false part - not the attempt to compare at all

Reply to @captgouda24, 17:14 UTC · 13 likes

Nicholas Decker (@captgouda24): When “Lord of the Flies” actually happened in real life, the boys cooperated for over a year until rescued. So why do we treat the speculations of one guy as meaningful?

image attached

For those interested https://en.m.wikipedia.org/wiki/Tongan_castaways

Friday, May 16, 2025

Quote tweet, 18:38 UTC · 27 likes

I’m just saying we did the leafleting first 🦐

Quoting Aaron Bergman 🔍 (@AaronBergman18): Ok, yes it’s almost a year late, but @Laura_k_Duffy, @SpacedOutMatt, and I finally did last year’s $15k incentive of fundraising for shrimp welfare at the mall!

Enjoy!!

video attached

Tweet, 22:05 UTC · 26 likes

New pigeon hour with @JesseTayRiver dropped a few days ago!

We discuss the HVAC industry, indoor air quality, what it’s like being so cool (sorta) and why Jesse isn’t richer

Links below!

Image

Thread, continued, 22:05 UTC · 4 likes

Main page with full transcript and links https://www.aaronbergman.net/p/14-jesse-smith-on-hvac-indoor-air

Thread, continued, 22:05 UTC · 2 likes

Apple pods https://podcasts.apple.com/us/podcast/pigeon-hour/id1693154768?i=1000708407758

Thread, continued, 22:05 UTC · 2 likes

Spotify https://open.spotify.com/episode/6MKHgKMDGvLWOsdmBw71kH?si=jO6QHeiCRaOEArVtdjVTAw

Reply to @SpacedOutMatt, 22:54 UTC · 22 likes

Matt 🔸 (@SpacedOutMatt): Shrimp welfare is having a moment right now and as a shrimp welfare guy I'm very happy! but also I think it's worth saying that even as a SWP donor (and a much bigger AWF donor, which has funded SWP), I think it's probably at least as likely as not that shrimp aren't sentient

God I hope they’re not

Reply to @SpacedOutMatt, 22:55 UTC · 3 likes

Matt 🔸 (@SpacedOutMatt): And shrimp probably have fewer neurons than ants (the RP team estimated 80,000, vs. 250,000 in ants according to Wikipedia)

Bob Fischer is gonna drone strike u for mentioning neuron counts

Reply to @arithmoquine, 23:14 UTC · 2 likes

henry (@arithmoquine): does anyone use that "operator" agent for anything

It’s gpt-2

Saturday, May 17, 2025

Quote tweet, 01:53 UTC · 30 likes

Oh guys…

Quoting Deedy (@deedydas): Google's AI just made math discoveries NO human has!

—Solved optimal packing of 11 and 12 hexagons in hexagons.
—Reduced 4x4 matrix multiplication from 49 operations to 48 (first advance in 56 years!)
and many more.

AlphaEvolve is the AlphaGo 'move 37' moment for math. Insane.

image attached

Thread, continued, 01:54 UTC · 9 likes

So it’s a matter of what, maybe months, until we get an AI-generated improvement on the transformer architecture?

Reply to @arithmoquine, 02:10 UTC · 4 likes

henry (@arithmoquine): @AaronBergman18 yep. or sooner 😊

👍

Reply to @ByrneHobart, 14:14 UTC · 2 likes

Byrne Hobart (@ByrneHobart):

image attached

I have the same or very similar pajama pants. Great style, very comfy

Tweet, 15:06 UTC · 4 likes

Nobody* wants to work for the federal government anymore. Career employees want out. People are retiring early. At the very least, morale is low. Maybe this is obvious idk

* not literally nobody

(N=small but probably bigger than yours)

Thread, continued, 15:07 UTC · 2 likes

I can’t speak for literally every federal employee, but at least many have a genuine sense that they’re part of some basically good unified project

That sense breaks down with Trump II in the White House

Quote tweet, 15:09 UTC · 71 likes

All the good EA critics are self-identified EAs

Quoting Jack (@tracewoodgrains): one of the first comments I saw in response

man, EA deserves better critics

image attached

Tweet, 15:19 UTC · 137 likes

1 box or 2 box in newcomb’s??

0 box

For the low price of $1M, you get to bamboozle a superhuman predictor and become cosmically unpredictable

Reply to @morallawwithin, 15:22 UTC · 3 likes

florence ⊮ (@morallawwithin): 1 box, but I only take Box B bc too much wealth corrupts

Smart

Reply to @NoahTopper, 16:09 UTC · 6 likes

Quinoah 🔍⏸️ (@NoahTopper): @AaronBergman18 could be us

It will be

Reply to @LeahLibresco, 18:14 UTC · 5 likes

Leah Libresco Sargeant (@LeahLibresco): A big part of the case for MAiD is that it is terrible for people to take their own lives *without the State validating their choice*

MAiD does not exist because suicide is otherwise impossible, but in order to say suicide is a good response to some human lives.

(quoting @v_j_freeman: Kim Leadbeater talking about how the supposedly terrible status quo which sees tiny numbers of people ‘forced to take their own lives’ as if that’s somehow worse than very many more people feeling forced to take their own lives.)

~painless suicide seems relatively difficult for the incapacitated (at least in areas with no available guns eg nursing homes)

Reply to @halogen1048576, 20:05 UTC · 8 likes

🇨🇦halogen (@halogen1048576): when people say that The Experts have determined that AI is about to become powerful enough to kill us all, I think they should understand that the Experts in question are using the same methodology and come from the same milieu as this kind of thing

image attached

You got a better methodology?

Reply, 20:06 UTC · 2 likes

🇨🇦halogen (@halogen1048576): when people say that The Experts have determined that AI is about to become powerful enough to kill us all, I think they should understand that the Experts in question are using the same methodology and come from the same milieu as this kind of thing

image attached

Tagging @SpacedOutMatt to defend himself if he wants

Reply to his own earlier tweet, 20:07 UTC · 8 likes

Aaron Bergman, May 17, 2025: You got a better methodology?

Only semi-rhetorical question btw

Actual answers invited

Reply, 21:01 UTC · 3 likes

Oliver Habryka (@ohabryka): Even just neuron counting methodologies seem much better to me, though mostly, it seems very unlikely to me that we wouldn't assign very large moral relevance to higher capacities for experience, and so you should expect some nonlinearities. Also, experience double-counting is a serious problem for weighing organisms as simple as this.

I agree that it underweights/is too dismissive of neuron counts (which were previously overrated as the only kinda plausible proxy), but I’m pretty inclined to give the RP thing good amount of credence because I don’t think I’d be able to suggest a better meta-methodology than “task RP with spending a ton of time and effort coming up with point estimates”

Idk, maybe you have really excellent taste for intuiting good and bad epidemics but this domain/topic seems at least somewhat out of distribution relative to like if they had committed the same amount of effort to coming up with a P(doom) estimate

Reply, 22:09 UTC · 4 likes

Matt 🔸 (@SpacedOutMatt): Screaming, crying, throwing up because this isn't here yet

(quoting @SpacedOutMatt: Genuinely not sure I can handle the wait)

Rugpull

Quote tweet, 22:35 UTC · 22 likes

You should almost certainly not be donating somewhere openphil funds

Quoting niplav (@niplav_site): It's come to my attention that many EAs still donate to global health & development.

While that's pretty awesome, I'd like them to consider being more risk-neutral and consider donating to some of the cause areas OpenPhil has abandoned.

Thread, continued, 22:37 UTC · 9 likes

It’s not like there’s really another side to this

My bet is that Dustin personally agrees with this, as does a majority of openphil research staff

Reply, 23:29 UTC · 4 likes

Max Alexander (@absurdlymax): I think open Phil has said they generally don't want projects to over rely on their funding, so if you strictly ev maximize you might donate where they dk

Yeah but I think straightforward funding levels >> “funder diversity” rn

It would be nice to live in a world where there could be both robust funding for deserving projects and funder diversity - that’s not the current world!

Reply, 23:41 UTC · 2 likes

keysmashbandit (@keysmashbandit): It's cool when you're friends with a couple who aren't that affectionate in public but you know behind closed doors they're fucking 24/7 crazy style

Chat is this true

Reply, 23:44 UTC · 2 likes

Max Alexander (@absurdlymax): Everyone pure EV maximization until I tell them to 🙄🙄

Wait tbc I’m approximately on board with EV maximization, I just think my claim about where to donate is a good proxy for it

Reply, 23:54 UTC · 15 likes

(The tweet being replied to is no longer available.)

Relevant hinge exchange (fellas, take notes)

Image

Image

Reply, 23:58 UTC · 3 likes

Agus 🔸 (@austinc3301): I think this is a mistake specifically for AI policy. Nonprofits can sometimes benefit from funder diversity, and orgs in AI policy in particular gain a lot from having only minority OP funding

Ok but isn’t the main benefit of a non-OP dollar still the purchasing power, not “funder diversity”

Sunday, May 18, 2025

Reply, 00:01 UTC

Maybe I’m misunderstanding smth but like I agree with you that funder diversity can be good, I’m just objecting to the implicit claim that it often (?) makes the difference about where to donate, because the dominant effect is just how much money orgs have, totally ignoring where those dollars came from

Reply, 00:03 UTC · 2 likes

(The tweet being replied to is no longer available.)

Something something diminishing marginal utility so if they merely fund most of the “full room” then this shouldn’t generally change my answer

Reply, 00:05 UTC · 3 likes

James (@grep_cat_): Other than that this was a perfect hinge reply, excellent game king

She got the ick bc I left a hanging parenthesis 😔

Reply, 00:06 UTC

Yeah fair, “almost certainly” in my original post was probably too strong

Reply to his own earlier tweet, 00:07 UTC · 5 likes

Aaron Bergman, May 17, 2025: It’s not like there’s really another side to this My bet is that Dustin personally agrees with this, as does a majority of openphil research staff

I regret to report I’m now a moderate

Reply, 00:08 UTC · 3 likes

(The tweet being replied to is no longer available.)

Sad! I want qualitative heuristics 😤

Tweet, 00:57 UTC · 2 likes

They don’t want you to know this but you can rent a(n 8) GPU cluster just for the (256) CPUs

Image

Reply, 01:44 UTC · 2 likes

Corinne (🏠 in NYC) (@manic_pixie_agi): @austinc3301 and @AaronBergman18, both EA and both names that start with A. sorry guys this has been fixed

Lol

Reply, 01:50 UTC · 2 likes

oh dandy lion🔸 (@ohwizenedtortle): @AaronBergman18 omega told me she didn't invite you to participate cos she knew you'd do this

😢

Reply, 02:10 UTC · 11 likes

Andy Masley (@AndyMasley): Received an especially goofy long blog reply to my ChatGPT post where among other things I specifically get called a fascist and eugenicist for pointing out that a laptop passively running for 3 minutes uses more energy than a ChatGPT prompt. Link below. The discourse is bad rn.

image attached

Jfc

Tweet, 02:30 UTC · 246 likes

I’m against social compartmentalization

Befriend your uber driver. Stay in touch with the person you went on 2 dates with. Have drinks with your boss. Reach out to your high school friend. Live in a group house with your coworkers if that’s what you want to do (cancel me)

Thread, continued, 02:30 UTC · 22 likes

“It’s cringe” and/or “somebody somewhere made HR departments get queasy about the arrangement” aren’t good enough reasons to not do a thing. It’s a free country (well mostly, sorta, for citizens, helps to be white)

Reply, 02:36 UTC · 6 likes

You (@parafactual): @AaronBergman18 in the case of boss/coworkers i think the reason to socially compartmentalize is to "hedge your bets", eg avoid creating problems at work if something goes wrong in your relationship with them outside of work

Yeah I agree there’s not zero cost here but like there are countervailing considerations in the other direction and much of the costs get significantly mitigated if all parties try hard to be good people lll

Reply, 02:50 UTC · 7 likes

Arjun Panickssery (@panickssery): You skipped over the ones that came to mind after I read the first two sentences, i.e. the ones I hear people criticize: "marry your grad student, date your secretary, lend money to your relatives, etc"

I think I’m directionally less against a total taboo on these things than is the cultural norm rn but there’s something to be said for putting “one party can’t easily say no” situations in a special bucket

Reply, 03:02 UTC

I mean to be boring about it it’s a spectrum!

These things are less concerning in places with strong social support systems, norms against retaliation of all kinds, rule of law, less powerful has labor market power, etc - these allow less powerful party to more easily say no

Reply, 04:35 UTC · 3 likes

Nina (@NinaPanickssery): @AaronBergman18 Some social compartmentalization makes sense as risk aversion

Yes good point

Reply, 04:37 UTC · 7 likes

bayes (@bayeslord): you can simply reject this coercive framing altogether

(quoting @HumanHarlan: Someone with p(doom) 95% is more likely to be called overconfident than someone with p(doom) 5% but they’re actually both saying that they’re 95% sure of something)

You can do whatever you want, but at the end of the day you still have to decide what actions to take and pretty quickly you start reinventing something like probability and P(something like doom) from first principles

Reply, 07:24 UTC · 9 likes

arabella ※ (@arabelladevine): why is there a rabbit in a backpack at this nightclub

“Why are all these monkeys dancing while I’m trying to chill in my backpack”

Reply, 10:43 UTC · 6 likes

Dylan Richardson (@indiidentity): @AaronBergman18 @frances__lorenz Ok real question is how did you direct the convo to EA? What kind of bait did you put in your bio, I need tips

Just a hint for keen eyes

Image

Reply, 14:17 UTC · 7 likes

bayes (@bayeslord): this way of relating to nature feels far too out of fashion. caring for the creatures of earth and earth itself is incredibly fatherly and manly, yet many men find ways to justify a childish disregard and destructiveness.

this sense duty and respect toward nature is part of why i support things like shrimp welfare projects, cultured meat, etc. a good man doesn’t lack empathy for things less capable or intelligent than himself.

we should be so awash in tech and abundance that we don’t think twice about doing these things, we should be so well-resourced that we can afford to treat every miraculous inhabitant of this planet with care

(quoting @webdevMason: them: masculinity is toxic by default me:)

Based 🫡

Reply, 17:21 UTC · 3 likes

Matt Reardon (@Mjreard): New show with @ChanaMessinger. We still don't know what to call it. Your tolerance for bottom basement production value will be rewarded with @AndyMasley content. Full first episode below 👇

video attached

Oh boy

Reply, 17:44 UTC · 5 likes

Jakeup (@myhandle): — hey man, how's your god emperor doing? are you happy with what you voted for?

— what?! oh, Trump, right, I almost forgot. nah, as soon as I started taking fish oil and my depression lifted I became a lib again and honestly don't follow politics much these days lol

— yeah. lol

Lol

Reply to @NoahTopper, 17:50 UTC

Quinoah 🔍⏸️ (@NoahTopper): I mean honestly you've written a pretty good ad for probiotics and fish oil here

I wanna know the strains and dha:epa ratio

Reply, 21:22 UTC · 4 likes

Ryan Briggs (@ryancbriggs): @AaronBergman18 @frances__lorenz You didn’t close the parentheses. ick

😢

Reply, 23:45 UTC · 28 likes

zzDANIEL_ETH (@daniel_271828): Has anyone done a deep dive on the “actually ice cream is healthy” research results? What’s going on there?

Not answering your question but I suspect at least part of it is just that ice cream isn’t actually bad for you - may as well be butter and rice - and it’s crowding out whatever is

Monday, May 19, 2025

Reply, 00:06 UTC · 17 likes

zzDANIEL_ETH (@daniel_271828): Idk, eating a bowl of butter and rice seems like it would be unhealthy?

I think this is heterodox but I don’t buy the standard “rational overconsumption” model where tasty calorie dense food makes people get fat and diabetic

Something weird is going on that’s breaking the very well-tuned control system that controls body fat % and such, and I don’t think arbitrary empty calories are the answer (replace butter with olive oil if u think saturated fat is bad, that’s mostly a separate debate)

Reply, 05:36 UTC · 1 like

Quinoah 🔍⏸️ (@NoahTopper): Huge news for @AaronBergman18

Yay thanks

Thread, continued, 05:39 UTC · 1 like

Ordered

Reply, 13:51 UTC · 2 likes

Nathan 🔎 (@NathanpmYoung): @florencehinder Sure, but this does kind of beg the question.

In order to fix this inequality, you would need a different kind of inequality, i.e. you need a lot more space for women's loos, because men just wee faster.

Not to say that's wrong, but it would still be inequality.

I mean nobody intrinsically cares about bathroom sizes

Reply, 17:22 UTC · 78 likes

Jesse Singal (@jessesingal): 1/ Everyone has known for years that liberal institutions are engaging in illegal hiring/promoting/recruting on the basis of race. I suspect there's going to be a massive backlash and that this will go down as a major self-own.

Link in next post because of X's idiotic policy.

Spicy take that’s somehow RW coded now but yes hiring on the basis of race is bad and should stop

Tweet, 21:49 UTC · 21 likes

Not exactly a novel insight but base models are very cool and interesting

Reply, 21:52 UTC · 2 likes

zzDANIEL_ETH (@daniel_271828): @AaronBergman18 More like based models amiright

I will tell you what 70b says momentarily

Tuesday, May 20, 2025

Reply, 02:13 UTC · 3 likes

Peter Wildeford🇺🇸🚀 (@peterwildeford): @AaronBergman18 how do I talk to one?

I’m currently using a single H100 for like $3/hour on Runpod, keeping llama-70b downloaded, and then you can pause the GPU and just pay ~$.05/hour for 200GB storage when not using so the cached model and all your other files don’t disappear

I’m mostly just playing around rn, no idea if this is a good setup for anything more legit

Lol if u want send me some text and I’ll do like 1000 predicted tokens depending and post the result

Reply to @dynomight7, 02:18 UTC · 2 likes

dynomight (@dynomight7): you can run (small) base models locally with ollama

ollama run llama3:text

Yeah although having played around a bit, there’s not a whole lot (LLM-wise) that runs on 16GB RAM that’s worth using

In my extremely limited experience you don’t really get to “oh that’s neat” before llama-8b

I think you can do llama-8b with 4 bit quantization locally with 16GB but it will be awfully slow

Reply to @parafactual, 02:56 UTC · 3 likes

You (@parafactual): just use 405b on hyperbolic

Using openrouter with 405, thank you @timfduffy !

Reply to @louisvarge, 03:01 UTC · 7 likes

Louis Arge (@louisvarge): animals are crazy when you think about it

we're like robots made of self assembling nanotech

and that just ... evolved???

Real

Reply to @GarrettPetersen, 03:16 UTC · 4 likes

Dr. Dad, PhD 🔄🔼◀️🔽▶️ (@GarrettPetersen): Do you think you could make banana bread right now without a recipe? You have access to a fully stocked kitchen.

2 bananas 1 cup flower 1 cup sugar a little cinnamon, 1/4 tsp salt half cup olive oil 1tsp baking powder 350° for 30 mins how’d I do

Thread, continued, 03:16 UTC · 3 likes

Wait also 2 eggs and a cup of water

Reply to @corsaren, 03:40 UTC · 6 likes

corsaren (@corsaren): If you think living forever is just as bad as death, then I’m sorry but you’ve never felt true love. I could wake up next to her every day and never get bored. I don’t want it to end.

It would be weird if they were equally bad. Like what a coincidence

Quote tweet, 03:48 UTC · 11 likes

Maybe base models are secretly smarter (bc others have been “lobotomized”) but afaict the main virtue is that you know exactly what you’re getting with the output: the most likely proceeding text!

Quoting Aaron Bergman 🔍 (@AaronBergman18): Not exactly a novel insight but base models are very cool and interesting

Thread, continued, 03:48 UTC · 3 likes

I’d use gpt-4o way more I think if there was a base model option + veto by some safety system

I wonder how natural it is for the market to have coalesced around guy-shaped chatbots

Reply to @AndyMasley, 12:29 UTC · 3 likes

Andy Masley (@AndyMasley): Had a great conversation tonight where another guy where we realized we both have a longstanding fantasy to drive for days deep into northern Canada. Do you also have this fantasy?

Do it Andy

Reply to @caleb_parikh, 12:31 UTC · 4 likes

Caleb Parikh (@caleb_parikh): I increasingly feel like transparency should be the main demand/frame that the AIS community tries to meme.

It's broadly popular, will help keep the AIS community and public institutions relevant during takeoff, and is more justified than pausing (imo).

Not “no takeoff until Trump is out”?

Seems to have a more direct/concrete theory of change. Transparency doesn’t intrinsically do anything

Quote tweet, 12:40 UTC · 41 likes

I regret to report that Gemini 2.5 Pro is good, flash 2.5 is good too, and it’s become my default for most things

Quoting Peter Wildeford🇺🇸🚀 (@peterwildeford): You have to feel terrible for the 31 students who didn't just plug the problem into Gemini 2.5 and then take two months off

Reply to @MattZeitlin, 13:03 UTC · 59 likes

Matthew Zeitlin (@MattZeitlin): What goes on here

image attached

Image

Quote tweet, 13:18 UTC · 10 likes

Dustin is even cooler than I thought

Quoting Bruno Faviero (@Bfaviero): @thatguybg completely cracked, <5'10 crowd outperforms

image attached

Reply to @austinc3301, 15:20 UTC · 7 likes

Agus 🔸 (@austinc3301): after two months of work, I finally have a working NixOS install on my main workstation, and it’s glorious

image attached

Haha whatever Tim Apple installs on computer go brr

Tweet, 15:23 UTC · 9 likes

Bumpers on vs bumpers off (bumpers are mostly just good tbc but still)

Image

Quote tweet, 19:35 UTC · 15 likes

I will critique but I also stan

Quoting Alexander Berger (@albrgr): Cool to see Cari and Dustin highlighted in TIME100 Philanthropy today. They don't toot their own horns much, but they've given over $4b so far - I believe they're the youngest billionaires to have given such a high proportion of their wealth away.

image attached

Tweet, 23:14 UTC · 11 likes

Obviously cancer bad etc but also it seems cosmically unfair if Biden doesn’t get at least a decade to chill and live it up (not my favorite president but still)

That seems like a bare minimum reward + recovery time for being POTUS, it is not an enviable job

Tweet, 23:35 UTC · 165 likes

Video thumbnail (video, plays on X)

Thread, continued, 23:36 UTC · 42 likes

We went from can’t write to “can write on a moving sign carried by a duck” in like what 8 months?

Wednesday, May 21, 2025

Tweet, 01:23 UTC · 15 likes

Anybody want this

Benefits and link below

Image

Reply to @gptbrooke, 02:17 UTC · 7 likes

brooke bowman (@gptbrooke): I mean no shade, but these sorts of questions are so misleading

There is no finding yourself in that situation as you currently are, or you’d already be there

You don’t wind up on the street with no resources or connections without a whole pile of confounding factors

(quoting @BrandonLBradfor: Annual reality check question: If you found yourself living on the street, $40 to your name, none of your current resources or connections, how are you getting off the street? What steps are you taking, and how long do you think it would take?)

Philosophers invented the conceivability vs possibility distinction specifically for this tweet!

Thread, continued, 02:20 UTC · 2 likes

More explicitly, I think it’s fine to observe that the situation in question is ~impossible and then think through it anyway

Reply to @toasterlighting, 02:38 UTC · 2 likes

Celene (@toasterlighting): So how exactly is one supposed to tell if persistent burnout and brain fog is the result of a covid infection rather than just happening for other reasons

Manifold market?

Sucks, good luck

Reply to @manic_pixie_agi, 04:21 UTC · 4 likes

Corinne (🏠 in NYC) (@manic_pixie_agi): @AaronBergman18 omg they turned Aaron into a duck please change him back 😭

Nooo I like this form 🦆

Tweet, 05:50 UTC · 15 likes

On my bucket list is making my own pharmaceuticals via chemical reaction

No I’m not talking about meth. I just wanna make *something*

Reply to @Evolving_Moloch, 18:25 UTC · 3 likes

Will (@Evolving_Moloch): My most normie lib opinion is that Hamilton is quite good.

image attached

Agree

Reply to @paul_cal, 18:26 UTC · 3 likes

Paul Calcraft (@paul_cal): Is the "like his express his" typo yours in the prompt, or Veo's?

CEO’s

Quote tweet, 18:30 UTC · 7 likes

Same

Quoting The Notorious J.O.V. (@whotfisjovana): yes

image attached

Reply to @mbateman, 18:57 UTC · 2 likes

Matt Bateman (@mbateman): re: “get off the apps”, here’s my experience on apps vs. IRL

image attached

Honestly glad I’m not the only one getting a decent number of first but not many second dates lol

Reply to @larsiusprime, 19:00 UTC · 2 likes

Lars Doucet (@larsiusprime): Me: Gee honey I'm just really tired today and I don't know why

My wife: Have you considered that you're narcoleptic and experiencing excessive daytime sleepiness, i.e., the classic defining symptom of the disorder that you have?

Me: Oh

Fish don't know what water tastes like

How do you manage, if I may ask? A close family member has EDS + hypersomnia. They just take lots of stimulants lol

Reply to @JesseTayRiver, 19:03 UTC · 16 likes

Jesse Smith (@JesseTayRiver): Feels not quite right for me to say yes to this but it was an African-American funeral

Lmao

Quote tweet, 19:08 UTC · 60 likes

Based based based (the images/content, not the QT)

Quoting Kitten 🐈 (@kitten_beloved): Checking in on the rationalists

image attached
image attached
image attached

Reply to @kitten_beloved, 19:09 UTC · 3 likes

Kitten 🐈 (@kitten_beloved): Bro is #9 on the rising list for philosophy substacks

Field is wiiiiiiide open

Based

Reply to @wasserstein_rao, 19:15 UTC · 5 likes

stefan (@wasserstein_rao): Automating experiments with AI is going to be really hard because automation is way less about high-dimensional Bayesian optimization than it is about figuring out why there are bubbles in your syringe pump line

Yes, there are only so many degrees of freedom you have in this situation

You can replace the line and the syringe, you can buy new ones from new suppliers, you can try every little new method to do whatever, at the end of the day tweaking these parameters in various combinations until loss goes down (in this case no more bubbles) just is high dimensional optimization

Reply to @larsiusprime, 19:17 UTC · 2 likes

Lars Doucet (@larsiusprime): EDS flares up every now and then and I was undiagnosed for 20+ years so I've never known anything different

Thanks, good luck

Reply, 19:33 UTC · 2 likes

“Back door” is doing a lot of work and is maybe kinda misleading/not true

Like it implies that maybe allowing lab grown wouldn’t be good unless it also facilitated banning real, but it stands independently of that

Quote tweet, 20:59 UTC · 5 likes

Republicans are bad and wrong about most policy matters

Quoting Chuck Schumer (@SenSchumer): With the news that Trump formally accepted the Qatari jet:

I took to the Senate floor to pass my bill to block DoD from spending a single penny on a foreign Air Force One

It’s a bribe

It’s a national security threat

But Republicans stood with Trump and blocked my bill

THAT’S NOT AMERICA FIRST

Thread, continued, 20:59 UTC · 2 likes

They’re right about like 5% of the culture war, not a good deal IMO

Tweet, 23:33 UTC · 40 likes

They need to make a bar but with stimulants instead of (or in addition to) alcohol

Reply to @freed_dfilan, 23:37 UTC · 9 likes

Daniel Filan 🔎 (@freed_dfilan): It's a coffee shop!

Ok but we want stronger stuff

Reply to @Empty_America, 23:58 UTC · 3 likes

VB Knives (@Empty_America): Other people definitely do have too much money.

(quoting @OldRowSwig: Boomers be like "there's no housing problem, you youngsters just want too big of a house, get a small cottage like I did when I was 21! You gotta WORK for it!")

No, we have too few houses!

Thursday, May 22, 2025

Reply to @Empty_America, 00:03 UTC · 5 likes

VB Knives (@Empty_America): @AaronBergman18 There are plenty of houses in America, quite a few are vacant falling down from from disinterest.

We have too few SFH with yards in desirable areas, but there isn't really any way to make more of these.

Yeah I doubt we actually disagree on much, although I’d add that adding non-SFH homes near or in these desirable areas would decrease the prices of SFHs

Reply to @freed_dfilan, 00:08 UTC · 3 likes

Daniel Filan 🔎 (@freed_dfilan): @AaronBergman18 It's cute but I feel like it's relevant to whether or not "this works" that the sign is wrong

Yeah although coherent + one typo could be worse

Reply to @singularitttt, 00:09 UTC · 5 likes

cybersyn (@singularitttt): am i the only one that loves it here? my life is so much more rich and meaningful because i spend so much time on twitter.

(quoting @tenobrus: wow, not tweeting is infinitely better than tweeting. not scrolling twitter is infinitely better than scrolling twitter. talking to people who don't spend all their time on twitter is infinitely better than talking to people who spend all their time on twitter. fuck this shit.)

I probably use it a bit too much but yeah strongly endorse using it significantly

Reply to @Meaningness, 00:17 UTC · 2 likes

David Chapman (@Meaningness): eBay, but for used 747 jets

doesn't seem to be a Buy Now button, though

https://www.controller.com/listings/for-sale/boeing/747/jet-aircraft/3

@livgorton bday present plz

Reply to @adastroworld, 00:20 UTC · 2 likes

adas🧦🌹 (@adastroworld): @AaronBergman18 There’s an entire ritual to stimulants, you can’t just order them from a bartender. Being offered nose drugs or acquiring something yourself is part of the entire thing, it would be totally different if you could just skip to the end

Just put the vaguely phenethylamine-shaped molecule in the bag bro

Reply to @AndyMasley, 01:04 UTC · 2 likes

Andy Masley (@AndyMasley): The Twitter For You tab is a constant stream of fascism
The Bluesky Discover tab is a constant stream of communism
The Substack Home tab is a constant stream of fake deep

My account is a constant stream of witty insight and correct takes (51% joking)

Quote tweet, 03:20 UTC · 23 likes

Progressives try to accomplish things instead of LARPing revolution difficulty impossible

Quoting Kelsey Piper (@KelseyTuoc): 1) this is horrific 2) if you imagine this does the slightest thing to improve anything for a single Palestinian, rather than make everything worse, you have a horrendously bad grasp of the world

Quote tweet, 05:49 UTC · 4 likes

This discourse is airplanesniping me again

Fun game that can be followed up with LLM conversation: why aren’t airplanes generally bigger than they are? Why aren’t they smaller?

Quoting Collin Rugg (@CollinRugg): NEW: 'Standing' airplane seats, which can increase passenger capacity by 20%, may be unveiled in 2026, according to the Daily Mail.

Unreal.

The seats force passengers to sit up higher, so Airlines can place the seats closer together.

The "Skyrider 2.0" seat also weighs 50% less, meaning the airplanes can also save on fuel costs.

"Skyrider 2.0 opens the travelling experience to a wider passenger market, creating also a useful space for the introduction of mixed classes boarded on the same aircraft," the seat company, Aviointeriors, said.

The seats have a bike-like seat that travelers will straddle and would likely be implemented for shorter flights.

The following video was filmed back in 2018 with Aviointeriors CEO, Ermanno de Vecchi.

video attached

Thread, continued, 05:55 UTC · 2 likes

Hmm another thought is that there could be interesting dynamics between passenger density and *speed* which isn’t usually much of a variable, but @boomsupersonic is trying to make it one!

People would be more willing to stand on faster trips

Tweet, 05:55 UTC

Tbc I don’t think this will check out for economic reasons. I predict that standing room only + supersonic will not be a provided thing in the near future even if over-land supersonic is legalized

Tweet, 06:02 UTC · 3 likes

Wait was I supposed to know that “[whatever drug] HCL” isn’t actually acidic bc the hydrochloric acid stops doing its thing when it attaches to another molecule

Reply to @Duderichy, 06:04 UTC · 6 likes

the Rich (@Duderichy): I would sooner have every insect killed 100 fold than have a baby suffer a paper cut

(quoting @kitten_beloved: Checking in on the rationalists)

Honestly same but that’s bc I’m worried about insect suffering and think their mass death, if painless, would be good

(In the abstract, irl I wouldn’t push the Kill All Insects Rn button for various reasons)

Quote tweet, 06:22 UTC · 5 likes

Yes correct take, Joe is morally serious to an extremely uncommon degree (positive)

Human extinction is kinda minor relative to the actual plausible stakes here

Quoting alex lawsen (@lxrjl): Self-recommending
(link in reply)

image attached

Reply to @growing_daniel, 06:23 UTC · 4 likes

Daniel (@growing_daniel): Insane that women like men. Absolutely bonkers. We’re living a winning lottery ticket every day brothers.

Real

Reply to @tenobrus, 06:32 UTC · 2 likes

Tenobrus (@tenobrus): is there literally any coherent argument for this?

(quoting @NewsWire_US: 100-0, U.S. Senate passes No Tax on Tips Act)

Steelman: it is a very dumb and sloppy way to transfer money to the working class, which is a good thing

The alternative isn’t some optimal tax policy that does the redistribution but better, its nothing

Reply, 06:33 UTC

I mean 30 min flights *are* offered, maybe they start making sense for ppl if tickets go down by 40% bc standing room only

Reply, 06:47 UTC · 1 like

Ok yeah fair there’s not a whole lot at 30 mins

But at <1 hour there’s DC to Richmond VA and DC to Philadelphia which isn’t crazy at all if you don’t have a car (in terms of time, bus makes more sense for most ppl obv)

Image

Reply to @ComInTheMarg, 15:49 UTC · 2 likes

Comments in the Margin (@ComInTheMarg): @AaronBergman18 @Duderichy You, BB, Brian Tomasik, and many other EAs generally take for granted that if insects can suffer, then wild insects live net-negative lives. But this may be false (so more insects better, maybe)

I mean idk if I take it for granted, I just think it’s true

Reply to @Jonas_Vollmer, 16:00 UTC · 73 likes

Jonas Vollmer (@Jonas_Vollmer): Doctor friend at large urgent care: most doctors use ChatGPT daily. They routinely paste the full anonymized patient history (along with x-rays, etc.) into their personal ChatGPT account. Current adoption is ~frictionless.

(quoting @Jonas_Vollmer: Quiz question: Do most doctors use ChatGPT all the time, or is adoption slowed by frictions from data privacy, reliability, computer security, etc.?)

I just hope they’re using o3!

Reply to @inductionheads, 16:00 UTC · 5 likes

Super Dario (@inductionheads): If you can't find a way to get 100:1 ROI on Claude 4 for $200/month you're seriously ngmi

I’m probably ngmi

Quote tweet, 16:10 UTC · 271 likes

If you are a doctor using this, for Gods sake, spend the $20 and use the better models

I am confident you can afford it

Quoting Jonas Vollmer (@Jonas_Vollmer): Doctor friend at large urgent care: most doctors use ChatGPT daily. They routinely paste the full anonymized patient history (along with x-rays, etc.) into their personal ChatGPT account. Current adoption is ~frictionless.

Reply to @janonacct, 16:13 UTC · 8 likes

janon (@janonacct): doctor using 4o mini

4 word horror story

Reply to @medievalpard, 16:24 UTC · 13 likes

Sol ☂️🧳🗝️ (@medievalpard): @AaronBergman18 what models would that be???

disclaimer: im a md, dont use ai (but because i find them bad in relation to medicine, not because im against them)

o3 ideally, o4-mini for quick questions

Also Gemini Pro is not ChatGPT but is currently the best!

Reply to @QualiaNerd, 16:26 UTC · 2 likes

QualiaNerd (@QualiaNerd): I agree wholeheartedly, but how much better do you estimate the Democrats are, and in what areas?

A lot better in almost every area lol

- healthcare
- taxation
- national security (yes really)
- foreign aid
- domestic research
- tariffs
- AI policy
- animal welfare, though ofc much room for improvement

Education is the one thing republicans *might* have the edge on

Reply to his own earlier tweet, 16:27 UTC · 4 likes

Aaron Bergman, May 22, 2025: o3 ideally, o4-mini for quick questions Also Gemini Pro is not ChatGPT but is currently the best!

Tweet, 17:05 UTC · 8 likes

Kinda based?

Quote tweet, 17:08 UTC · 42 likes

Great news although I kinda wish we knew anything about the nature of sentience

Quoting Rosie Campbell (@RosieCampbell): Claude 4 is the first model to be independently evaluated for AI welfare. Proud to say @eleosai conducted these evals! Congrats @AnthropicAI on setting this precedent. Read more in the system card...

image attached

Tweet, 17:57 UTC · 46 likes

Claude 4 successfully automated most of my current task lol

Thread, continued, 18:00 UTC · 15 likes

It’s relatively vanilla selenium web scraping but even that has proven often difficult relative to more straightforward curl and apis and such

Thread, continued, 18:02 UTC · 11 likes

It seems to know the whole html schema of a fully loaded YouTube results page without me actually providing any html divs or anything

Reply to @QualiaNerd, 18:04 UTC · 3 likes

QualiaNerd (@QualiaNerd): What do you find unconvincing about physicalist qualia realism?

Hard problem and explanatory gap go brr

Reply to @justjoshinyou13, 18:20 UTC · 2 likes

Josh You (@justjoshinyou13): did you see the viral tweet about the doctor using chatgpt. He was on plus but he was using 4o

No, F

Tweet, 18:31 UTC · 7 likes

Me 🤝 Kim Jong Un

Against unscientific empiricism

Image

Reply to @absurdlymax, 20:25 UTC · 2 likes

Max Alexander (@absurdlymax): @AaronBergman18 What are your thoughts on scientific empiricism tho

Empiricism has become highly overrated in science, I’m not sure when exactly this started, but after the enlightenment

Tweet, 20:29 UTC · 16 likes

Why does biking suck? I’m like 60% sure it’s not purely an exertion thing: swimming, hiking (even uphill with a pack), and climbing feel qualitatively less sucky for some reason

Thread, continued, 20:29 UTC · 7 likes

If I have to travel a mile at say 2% grade uphill, biking will be faster than walking but much less pleasant

Reply to @pli_cachete, 20:32 UTC · 3 likes

(The tweet being replied to by @pli_cachete is no longer available.)

Id guess they’ve done internal testing at OpenAI and found that performance plateaus after some amount of time for reasoning models

I mean that’s not to say the niche is impossible, I’m sure people are working on it

Also hella expensive

Reply to @CharlesD353, 22:41 UTC · 1 like

Charles🔸 (@CharlesD353): How much practice do you have?

Minimal for good reason

Tweet, 22:43 UTC · 8 likes

There’s a meme about building internal tools for tech companies or executives or something being lame but that actually sounds pretty fun

I do like making a good tool

Reply to @craigweiss, 22:45 UTC · 7 likes

Craig Weiss (@craigweiss): Success is not correlated to intelligence or luck.

The bitter truth is that it's correlated to how badly you want to succeed.

It’s definitely correlated with intelligence and luck lol

Reply to @NoahTopper, 22:56 UTC · 2 likes

Quinoah 🔍⏸️ (@NoahTopper): You should work at AWS, 100% internal tools

Tell ur manager I say hi

Reply to @JesseTayRiver, 23:10 UTC · 1 like

Jesse Smith (@JesseTayRiver): Low tire pressure

I’m gonna be really annoyed if this is the answer

Reply to @olivertraldi, 23:32 UTC · 1 like

Oliver Traldi (@olivertraldi): I could be wrong, but one difference I often observe between men and women: I don't think men have quite the same feeling of being "told by society" as women. I don't know exactly where it comes from, but I think this difference leads to very different kinds of social theorizing.

(quoting @Ruesavatar: I see threads like this and wonder why these men (NOT ALL MEN ffs) are so angry. There are more in the comments raging at her for making “poor choices”, but.. my entire youth I was told, BY SOCIETY, the most important thing was to build a career and to put off kids until my 30s.)

Yeah I have the same sense

Quote tweet, 23:35 UTC · 9 likes

Yes we dream of a world of ~fully-automated, near ingredient-price fast casual pickup

Quoting keysmashbandit (@keysmashbandit): It's crazy you have to put in so much a) effort, or b) money to get 400 calories of sensible sustenance. It should be the cheapest thing imaginable. The efficient market hypothesis is false

Friday, May 23, 2025

Reply to @emollick, 03:26 UTC · 39 likes

Ethan Mollick (@emollick): Huh. Looks like Plato was right.

A new paper shows all language models converge on the same "universal geometry" of meaning. Researchers can translate between ANY model's embeddings without seeing the original text.

Implications for philosophy and vector databases alike.

image attached

Maybe I’m dumb or don’t get it but I find this very unsurprising

Quote tweet, 03:38 UTC · 51 likes

We must change nature until all the animals are spronking together joyfully

Quoting Someone:

Tweet, 03:44 UTC · 24 likes

I regret to report that sometimes sending an email just works

Quote tweet, 04:28 UTC · 8 likes

Soliciting takes on Eleos donations

Do people think it beats Arthropoda + Lightcone + LTFF

Quoting Holly ⏸️ Elmore (@ilex_ulmus): If you want to donate to protect possible AI welfare, Eleos is my recommendation.

Thread, continued, 04:28 UTC · 2 likes

Not because I have much money to donate (sad) but bc I have been known to make this a hobbyhorse and some of my friends are rich

Reply to @ilex_ulmus, 04:39 UTC · 2 likes

Holly ⏸️ Elmore (@ilex_ulmus): If you want to donate to x-risk prevention, @pauseaius is my recommendation!

Yeah fair enough, tbc I sorta have priors against this being the best thing but I’m definitely not certain and it’s more plausible than a lot of other stuff people donate to

Tweet, 04:54 UTC · 53 likes

Someone should figure out whether computer stuff can be sentient

Reply to @really_eli, 23:17 UTC · 4 likes

(The tweet being replied to by @really_eli is no longer available.)

Sad!

Reply to @bellaforristal, 23:17 UTC · 3 likes

Bella Forristal 🔸 (@bellaforristal): I appreciate the pedantry, but I'm afraid I have to dispute the claim:

image attached

Yes, I’m calling it spronking!!

Saturday, May 24, 2025

Tweet, 02:20 UTC · 8 likes

New pig race just dropped

Thanks to @NoahTopper for the tip https://youtu.be/8jLNNMFur5A?si=VslvkA8FZ7XdxYAh

Reply to @MarmotRespecter, 22:00 UTC · 3 likes

marmot (@MarmotRespecter): can't believe this worked

image attached
image attached
(quoting @MarmotRespecter: if you: 1) live somewhere in this picture 2) don't go to an electric guitar church 3) are shaped like a mom from a Pixar movie my DMs are open)

Based

Reply to @zeta_globin, 22:08 UTC · 2 likes

zeta (@zeta_globin): did you know he primarily did it for a hot girl

image attached

Many such cases

Sunday, May 25, 2025

Reply to @ArthurMacwaters, 02:48 UTC · 2 likes

Arthur MacWaters (@ArthurMacwaters): The US dominates so hard that even Saudi aramco was founded in SF

Absolute dominance 🇺🇸

(quoting @ArthurMacwaters: 10/10 of the largest companies in the world were founded in the US)

What

Reply to @LordDreadwar, 03:05 UTC · 2 likes

Darth Dreadwar 🔺 (@LordDreadwar): Yes.

Sorry bro ur insane

Tweet, 15:34 UTC · 15 likes

TIL “duck” (🦆) and “duck” (bend) are closely etymologically related

Image

Reply to @profeshkitten, 15:40 UTC · 3 likes

✴︎ kitt ✴︎ (@profeshkitten): meet me at the anti-pigeon pigeon-appreciation sculpture

image attached

This is evil

Reply to @manic_pixie_agi, 20:57 UTC · 5 likes

Corinne (🏠 in NYC) (@manic_pixie_agi): finally got around to listening to this and it is indeed self recommending

(quoting @AaronBergman18: New pigeon hour with @JesseTayRiver dropped a few days ago! We discuss the HVAC industry, indoor air quality, what it’s like being so cool (sorta) and why Jesse isn’t richer Links below!)

Omg yay

Reply to @JesseTayRiver, 21:27 UTC · 2 likes

Jesse Smith (@JesseTayRiver): The podcast that @AaronBergman18 did with me has taken the world by storm

(quoting @manic_pixie_agi: finally got around to listening to this and it is indeed self recommending)

Truly

Reply to @JesseTayRiver, 21:28 UTC · 10 likes

Jesse Smith (@JesseTayRiver): My daughter is making a recipe that uses the outside of 37 Oreos so I may have just eaten the filling from 37 Oreos am I going to die

Yes but probably not for quite a while

Reply to @ratlimit, 21:49 UTC · 7 likes

(The tweet being replied to by @ratlimit is no longer available.)

Bro just take the elonbux

Reply to @arithmoquine, 21:51 UTC · 6 likes

henry (@arithmoquine): No more "likes". No more "retweets". No more "follows". No more "grok".

There's only one type of "bait" I deal with now.

Goodbye forever Anon. I've found my calling.

image attached

As long as you put some heroin on the hook and throw them back ♥️

Reply to @AstroRose20, 23:28 UTC · 2 likes

(The tweet being replied to by @AstroRose20 is no longer available.)

Haha what the fuck

Reply to @louisvarge, 23:42 UTC · 5 likes

Louis Arge (@louisvarge): i think everyone has unreasonably strong and unfounded priors about consciousness specifically

common ones implicit in people's thinking:
- neuron count or intelligence would correlate to level of consciousness
- self-awareness / recurrent processing is necessary
- only the human hedonic range is where useful computation can happen
- consciousness is a software problem, not a hardware problem

i think there's very limited evidence in general, but imo the tiny weight of evidence we do have tips the scales against all of the above points

Yes nobody has a goddamn clue about anything related to consciousness

Reply to @Benthamsbulldog, 23:45 UTC · 2 likes

Bentham's Bulldog🔸 (@Benthamsbulldog): @frances__lorenz @JesseTayRiver thank you

She’s flirting

Monday, May 26, 2025

Reply to @GovAmyKlobuchar, 02:47 UTC · 49 likes

will 🌱 (@GovAmyKlobuchar): Porn is bad for everyone and we should ban it and I have never heard a good argument for it

The argument for it is that people enjoy it and it’s fun (I’m being serious - people often forget the most straightforward first order effects of things)

Reply to @NoahTopper, 05:10 UTC · 10 likes

Quinoah 🔍⏸️ (@NoahTopper): @AaronBergman18 @LatinxPutler Reminds me of Scott Alexander’s analysis of decriminalizing marijuana that forgot to take into account the fact that people enjoy using marijuana

Many such cases!

Reply to @frances__lorenz, 15:11 UTC · 1 like

(The tweet being replied to by @frances__lorenz is no longer available.)

😈

Reply to @frances__lorenz, 15:41 UTC · 5 likes

(The tweet being replied to by @frances__lorenz is no longer available.)

Ahhh they replaced Frances

Reply to @LinkofSunshine, 15:57 UTC · 56 likes

Basil🧡 (@LinkofSunshine): Bush being insanely good at one specific issue of global public health was a good bit

image attached

Um based

Reply to @arithmoquine, 16:33 UTC · 8 likes

henry (@arithmoquine): the best defensive is to simply point out that it automatically neutralizes any criticism just by Being What It Is

> wireheading would feel bad for X reason!
> ok, then it's not real wireheading. play that back and forth with yourself until you're convinced

Same thing with “lives barely worth living” and repugnant conclusion stuff

People are actually imagining lives not worth living. Play the same game

Quote tweet, 16:33 UTC · 40 likes

Bro

Quoting Peter Wildeford🇺🇸🚀 (@peterwildeford): The Substack to impact pathway is real

Congrats @Benthamsbulldog @andresjz_swp

image attached
image attached

Reply to @arithmoquine, 16:39 UTC · 5 likes

henry (@arithmoquine): @AaronBergman18 @louisvarge tangential but what portion of human lives do you think are actually sort living

Oh man big question, idk like 70% in 2025 but gigantic uncertainty

Reply to @arithmoquine, 17:02 UTC · 4 likes

henry (@arithmoquine): i mean, with my current values, i'd kms if i were 99.9999% of the population. but "if i were someone that i'm not" is pretty ill-defined

Bold lol, is this just hedonic wellbeing? No way 99.9999% of people are just mostly suffering right?

Tweet, 17:30 UTC · 14 likes

I’d guess a lot of correct patient intuition about whether therapy is or is not right for them gets laundered by Big Psych as “there’s a direct causal connection between whether you think it will work and whether it actually will”

Tweet, 17:30 UTC · 2 likes

ignoring the load-bearing other node of like “fundamentally is whatever’s wrong with me the kind of thing that therapy can fix”

Thread, continued, 17:30 UTC · 2 likes

Like the implicit false claim is “if only you could will that you believed other than you did, then it would work for you”

Reply to @RFishBlueFish, 18:56 UTC · 4 likes

Gil (@RFishBlueFish): @arthur_wright_ @AndyMasley Illusionism is the final boss of philosophy: probably right, but extremely fucking weird

I think most illusionists don’t really get how weird their thesis is, and if they did the majority would stop being illusionists

I’m at 3% illusionism is true

Tweet, 19:17 UTC · 114 likes

Nobody asked but I’m like 65% on “Anthropic is good actually”

Reply to @codytfenwick, 19:29 UTC · 10 likes

Cody Fenwick (@codytfenwick): @AaronBergman18 Does this mean there's a 65% chance that it's good for the world on net? Or just 65% chance that the people there are generally well-intentioned?

The former

Tweet, 20:39 UTC · 16 likes

Fun with llama-405b-base

So far llama is a hater for
- Hegel
- Machiavelli
- Nick Bostrom
- Sam Harris
- David Chalmers

Image

Thread, continued, 20:40 UTC · 3 likes

Lol

Image

Reply to @IsaacKing314, 21:17 UTC · 10 likes

Isaac King 🔎 (@IsaacKing314): Would you have said the same about OpenAI before the board fired Sam?

Immediately before? No

Like 5 years ago yeah probably

Reply to @toasterlighting, 21:55 UTC · 1 like

Celene (@toasterlighting): @AaronBergman18 @arithmoquine @louisvarge Really? I'm persistently suicidal and I'd guess like ~90%, based off the studies in WWOTF

I adjust quite a bit from that knowing that people memoryhole their worst experiences but also I wouldn’t be that surprised if the real answer is 90 or even 95%

Reply to @ApriiSR, 21:57 UTC · 7 likes

Aprii *️⃣ (@ApriiSR): i'm saying if you sum up all the prison-like thingies maybe it makes a substantial number

I suspect the main prison-like thingy is just “sick without access to hedonically-responsible health care”

So the African with malaria counts but so does the upper middle class US person with chronic pain and nobody willing to write an adequate opioid prescription

Quote tweet, 22:00 UTC · 11 likes

Want to coin the term “hedonically responsible healthcare” (pending a better term”)

It basically means that they care intrinsically a lot abouy suffering and subjective wellbeing

Quoting Aaron Bergman 🔍 (@AaronBergman18): @ApriiSR @toasterlighting @arithmoquine @louisvarge I suspect the main prison-like thingy is just “sick without access to hedonically-responsible health care”

So the African with malaria counts but so does the upper middle class US person with chronic pain and nobody willing to write an adequate opioid prescription

Reply to @ApriiSR, 22:03 UTC · 6 likes

Aprii *️⃣ (@ApriiSR): @AaronBergman18 @toasterlighting @arithmoquine @louisvarge —if my hand were truly forced by conditions comparable to having malaria, i would simply get opioids off the dark web. but i suppose not every upper middle class US person is so enlightened

Maybe I’m uniquely incompetent but my sense is that successfully navigating the dark web without direct assistance is like an IQ 110 + substantial effort + substantial tech literacy task

Seems quite hard to do especially if you’re debilitated

Thread, continued, 22:04 UTC · 5 likes

There’s a reason so many people buy God knows what off the street instead of going online. I’m sure many of them have and use Amazon prime - it’s just not nearly as easy!

Reply to @reconfigurthing, 22:37 UTC · 3 likes

Elias Schmied (@reconfigurthing): I think I'm something like an illusionist🙂what's so weird about it?

It’s claiming that I’m mistaken about the most fundamental parts of my own life and experience

Reply, 22:54 UTC · 2 likes

@JesseTayRiver

Quote tweet, 22:59 UTC · 3 likes

Federal funding for science is just free lunch for society

Who is even asking for cutting hard science funding? Is it just one DOGE staffer who now runs the federal government?

Quoting sarah (@sarahtonin): quantum computing. detecting gravitational waves. higgs mechanism. AdS/CFT. discovery of dark energy. precision CMB detection. Bose-Einstein condensates. topological phases of matter. holography. conformal bootstrap. Advancement in Superconductors. quantum error correction.

Tweet, 23:22 UTC · 226 likes

I’m starting to think multiple $20/month subscriptions don’t all collectively sum to $0

Reply to @Adellama_, 23:28 UTC · 3 likes

Ada🔸 (@Adellama_): @AaronBergman18 no you're not telling me this. it can't be true

Sorry 😢

Reply to @izzyz, 23:36 UTC · 2 likes

izzy (@izzyz): cool! thanks. how do you like the hvac job? that was another option i'm thinking of

(Obligatory link)

Tuesday, May 27, 2025

Reply to @ApriiSR, 02:08 UTC · 3 likes

Aprii *️⃣ (@ApriiSR): @IsaacKing314 @toasterlighting it's not 100% apple juice, it's 100% juice apple juice. the juice part of the apple juice is 100% juice

I love it

Tweet, 02:12 UTC · 11 likes

Found in the Facebook mines (credit to Robin)

Image

Reply to @reconfigurthing, 02:14 UTC · 2 likes

Elias Schmied (@reconfigurthing): @AaronBergman18 @RFishBlueFish @arthur_wright_ @AndyMasley Yeah, depending on what you mean I do disagree with this framing. You don't have direct experiential access to something like "This sensation of being is strange and metaphysically special". That goes through your conceptual/intellectual schemas first, which can be confused. But you might also mean something different.

I agree, at least I think, although I’d put it in slightly different terms (“direct access to qualia doesn’t imply direct access about it”)

That’s why I think p(illusionism)>>0%, maybe 3% as I’ve said elsewhere

Reply to @cremieuxrecueil, 02:45 UTC · 2 likes

Crémieux (@cremieuxrecueil): In case you don't recognize it, this is a follow-up of the infamous PNAS piece on baby brainwaves.

The paper in question was rubbish, but it purported to show a move to higher-SES-style brainwaves without any noted behavioral impacts.

Now we have behavioral measures. Bupkes!

image attached
(quoting @captgouda24: A null result for the effects of rather substantial transfers to young children, so far. It is still too young to get fine-grained outcome information, but this isn’t a good sign for transfers helping all that much.)

I think I’d bet on there being some very small genuine effect here even if these studies are too underpowered to see anything

Like the whole causal pathway is there - chronic cortisol fucks people up, money is a legitimate driver of stress, etc

I’m not a genetics disbeliever either - I’m sure that’s the dominant explanatory factor but still

Tweet, 02:48 UTC · 8 likes

Me monitoring the gossip at EAG London from afar

Image

Reply to @reconfigurthing, 03:08 UTC · 2 likes

Elias Schmied (@reconfigurthing): Exist how?

Like exist in some genuine “approximately how it seems prima facie” sense

You can wind up playing endless word games with terms like “exist” and “illusion”

My claim is that there’s a 97% chance that genuine basically-what-it-seems-like qualia are a thing

Quote tweet, 03:14 UTC · 21 likes

Got it prescribed

Very cool ~immediate biofeedback

Image

Quoting Aaron Bergman 🔍 (@AaronBergman18): Guanfacine is a super interesting drug

It’s not merely a non-stimulant ADHD medication but essentially an *anti*-stimulant, fundamentally doing the opposite of what CNS stimulants do in many ways

Reply to @connerdelights, 03:27 UTC · 3 likes

Andrew Conner (@connerdelights): How does it feel, phenomenologically?

Not a whole lot to report honestly, to a first approximation it feels like nothing

Tweet, 03:30 UTC · 72 likes

It’s wild to me that people take drugs/medications without even glancing at the Wikipedia page*

*in an outpatient setting. If you’re getting emergency surgery I get it

Reply to @connerdelights, 03:34 UTC · 3 likes

Andrew Conner (@connerdelights): oh weird! is it helping for ADHD?

Probably too soon/subtle to tell

The first couple days on it I did work substantially more than avg but it regressed to the mean, it’s only been 6 days

It’s not as obvious/potent as like first time adderall

Tweet, 03:34 UTC · 10 likes

Image

Reply to @ApriiSR, 03:35 UTC · 12 likes

Aprii *️⃣ (@ApriiSR): @AaronBergman18 what if it doesn't have one

U have to kidnap a pharmacology phd and ask them ur questions

Tweet, 03:56 UTC · 12 likes

How much of the LLMs being unable to answer 8.8-8.11 thing is just that they’ve been RLHF’d to be extremely hesitant around the token(s) “69”

Thread, continued, 03:57 UTC · 5 likes

Like presumably they’ve learned “whenever you think that’s the next token(s), you’re probably wrong”

Reply to @Mira___Mira, 04:33 UTC · 3 likes

(The tweet being replied to by @Mira___Mira is no longer available.)

@livgorton make it happen

Reply to @evapilotno17, 09:45 UTC · 2 likes

Ą͚͕͖͎͈͟͞n̸̨͕͍̘̬͟͟͝i̡̼̹͔̳͉̫͎͓͟͟ ̷̸͝͝ (@evapilotno17): omg 😭

image attached

Lmao

Reply to @yzilber, 09:59 UTC · 4 likes

Jazi Zilber (@yzilber): @AaronBergman18 newspapers solved this, but giving you a huge collection of articles for cheap.

we repeatedly re-invent the wheels.

reasons are good. but it's still ridiculous

It’s mostly AI stuff, not Substack

Reply to @chriswaterguy, 10:29 UTC · 2 likes

Chris Waterguy – neurodivergent coaching (@chriswaterguy): Is there any vegan protein powder that doesn't have artificial sweeteners and doesn't taste disgusting?

I'm happy with whey powder but would happily go vegan.

I used to get pea/rice mix from TrueNutrition

Reply to @psycheureka, 10:34 UTC · 1 like

Psychiatry Excellence (@psycheureka): Chronic maternal stress during pregnancy can alter fetal neurodevelopment via a process known as fetal programming.

It increases the long-term risk of cardiovascular, metabolic, and neuropsychiatric disorders in the child.
(Hales & Barker, 2001)

Let’s break down the key mechanisms linking maternal stress to altered fetal brain outcomes. 🧵👇

GIF attached

Seems bad

Reply to @WillowChem, 11:30 UTC · 4 likes

MagpAI (@WillowChem): Iirc the reasons given were more fundamental than suggested here
https://forum.effectivealtruism.org/posts/EZZveBtxoZJjSighP/are-far-uvc-interventions-overhyped-founders-pledge

@JesseTayRiver

Thread, continued, 11:35 UTC · 4 likes

We discuss a bit here https://www.aaronbergman.net/p/14-jesse-smith-on-hvac-indoor-air

Tweet, 14:08 UTC · 10 likes

Wanna be pro Jordan Peterson to be contrarian now

Reply to @panickssery, 14:37 UTC · 1 like

Arjun Panickssery (@panickssery): @NinaPanickssery and I had a baby last Tuesday

Leo Panickssery

Wait what

Reply to @panickssery, 14:37 UTC · 1 like

Arjun Panickssery (@panickssery): @NinaPanickssery and I had a baby last Tuesday

Leo Panickssery

Dude congrats!!! 🔥🔥🔥🔥🔥

Reply to @panickssery, 14:38 UTC · 7 likes

Arjun Panickssery (@panickssery): @NinaPanickssery and I had a baby last Tuesday

Leo Panickssery

Poast baby (if u want)

Reply to @livgorton, 14:55 UTC · 2 likes

Liv (@livgorton): i've vibe coded my way to daily arxiv summaries in my inbox. y'all are putting out too many interp papers 😤 i had to innovate.

(the code is available on my github but it's too shameful to link. maybe i will get claude to improve it a bit soon)

image attached

Cool

Reply to @WillowChem, 15:11 UTC · 9 likes

MagpAI (@WillowChem): @AaronBergman18 Super curious to hear you expand on this

Oh man, it’s really a lot of vague impressions and strands coming together

But basically in a ruthlessly consequentialist sense it seems like (1) having a relatively responsible player in the mix is good for competition dynamics and (2) perhaps more importantly, they’re literally just building more aligned models???

Like it’s not all that hazy anymore, I really would just want opus-6 in control of important systems and such if the alternative is o6 or Gemini 6

Tweet, 15:16 UTC · 6 likes

How do I become an Etsy witch

Tweet, 15:49 UTC · 10 likes

Lots of reasonable object level arguments around language and gender norms and such but the dominant consideration should generally be whether your proposed set of norms is going to help Democrats win the next election

Thread, continued, 15:49 UTC · 6 likes

Tbc I kinda hate this style of reasoning I just think it might be true in many cases

Reply to @Aella_Girl, 16:37 UTC · 2 likes

Aella (@Aella_Girl): I thought we were *against* a world where people have to hide their true beliefs in order to not get fired

(quoting @ReubenR80027912: Broadly: You should not listen to any poster w/o an actual job Hanania, RFH, Aella, self employed bloggers…all are exempt from the *moderating influence of needing to get along with their boss & colleagues* Free from basic guardrails of society, their views become unhinged)

Lol her post but the exact opposite

Reply to @IterIntellectus, 17:37 UTC · 17 likes

vittorio (@IterIntellectus): not enough

What the fuck

Reply to @RatOrthodox, 17:37 UTC · 4 likes

Brangus🔍⏹️ (@RatOrthodox): Lol I was bullied pretty brutally. Hope that helps you get off or whatever.

😢

Reply to @panickssery, 17:50 UTC · 4 likes

Arjun Panickssery (@panickssery):

image attached

lookin fresh

Reply, 18:21 UTC · 2 likes

Oh yes I love that site lmao

Thread, continued, 18:21 UTC · 2 likes

I think I have them like a $5 donation

Reply to @sixers2772, 18:30 UTC · 1 like

Sixers (@sixers2772): What your body tells you in the summer: time to cut down on fat and increase the carbohydrate

What my body tells me in the summer: I would like another bowl of the same thing you ate for the past 200 meals

Wait same lol

Reply to @absurdlymax, 19:25 UTC · 2 likes

Max Alexander (@absurdlymax): I think there's a certain type of person that seems to value being inflammatory or annoying as more important than accurate communication, or even impactful outcomes

We should greatly discourage this behavior when we see it

Noooooooo

Reply to @EffectvAltruism, 20:14 UTC · 3 likes

Effective Altruism (@EffectvAltruism): Now live: a completely revamped http://effectivealtruism.org! 💡

Effective altruism asks a simple question: how can we do the most good with our time, money, and resources? Our new site demonstrates how this philosophy translates into real action across fields like global health, animal welfare, and existential risk.

image attached

Oh boy

Reply to @cremieuxrecueil, 20:27 UTC · 2 likes

Crémieux (@cremieuxrecueil): "I'm not fat, I just have a lot of muscle. BMI doesn't work for me."

How often is this true? How often does a high BMI suggest healthy, muscular people are fat?

Rarely.

Obesity based on direct measurement of body fat agrees with the BMI-based obesity label in ~98% of cases!

image attached

2% isn’t *that* rare

Tweet, 22:37 UTC · 12 likes

No data is safe from my data scraping ability

You can’t run bc you’re usually an element of a web page, but you can hide, just hopefully not for very long

Thread, continued, 22:37 UTC · 6 likes

File type conversions like you’ve never seen. We’re parsing the unreadable characters of .webarchive

We’re going video->images->pdf->txt->csv

Thread, continued, 22:38 UTC · 8 likes

Reply to @captgouda24, 22:55 UTC · 4 likes

Nicholas Decker (@captgouda24): I want to defend those who have no one willing to defend them!

🫡

Wednesday, May 28, 2025

Tweet, 00:31 UTC · 857 likes

Not commenting on the context for this screenshot but it’s baffling to me that (ostensible) adults aren’t ashamed to say things like this

Image

Thread, continued, 00:31 UTC · 108 likes

Some people try really hard to cultivate a good image (me, most people in my experience, SBF) but some people will just basically tell you that they suck

Reply to @TylerAlterman, 00:38 UTC · 4 likes

Tyler is finishing a book, slow to reply (@TylerAlterman): Can someone explain to me like I’m an idiot why so many people are upset over this in the fairly specific way they are?

(quoting @RatOrthodox: As you all know I was the 100th person to have recreational sex with @Aella_Girl, so last night I forced her to tattoo a 100 on me. She has never tattooed before. I am v happy with the result.)

No

Reply to @arithmoquine, 00:41 UTC · 180 likes

henry (@arithmoquine): @AaronBergman18 yeah. and vittorio seemed perfectly normal irl! he wouldn't say these things to someone's face, i wish he had that decency here

Wow unironically super interesting

Maybe I should refrain from generalizing but otoh I kinda feel like how you act when it’s easiest to be cruel is closest to your “true” character

Tweet, 03:04 UTC · 10 likes

Google is ~definitely taking taking a loss on this

Is this legal? (I think I’ve made this tweet before but still)

Image

Thread, continued, 03:04 UTC · 5 likes

Its prices about half that of Haiku-3.5 and, more tellingly, the rate limits make it clear that they really don’t want you using the model that much

Reply to @justjoshinyou13, 03:09 UTC · 1 like

Josh You (@justjoshinyou13): I doubt they're losing money. It costs more than, say, Llama 70B

https://artificialanalysis.ai/models/llama-3-1-instruct-70b/providers

If they were making money they wouldn’t make it so hard/slow to buy their product

Reply to @captgouda24, 06:06 UTC · 13 likes

Nicholas Decker (@captgouda24): I want to make it clear that I am not attracted to any animal besides humans! And maybe sufficiently humanoid aliens.

Never stop posting bro (unironically)

Reply to @banterrealism, 06:07 UTC · 2 likes

Sex Haver (@banterrealism): Propaganda I’m not falling for:

- leftist men are more sinister than rw men

- all SWs are completely immune from criticism

- misandry poses a threat worth worrying about

- new developer builds won’t be in the red forever going forward

“Worth worrying about” idk, worrying per se never really helps, but I think casual misandry among the western elite in fact cashes out as men having worse lives (mechanism involving the inhibition of a benevolent legitimate men’s issues/concerns movement)

Quote tweet, 06:15 UTC · 2 likes

Hmm

Quoting Jesse Mu (@jayelmnop): I recently moved to the Code RL team at Anthropic, and it’s been a wild and insanely fun ride. Join us!

We are singularly focused on solving SWE. No 3000 elo leetcode, competition math, or smart devices. We want Claude n to build Claude n+1, so we can go home and knit sweaters.

Reply to @adastroworld, 08:01 UTC · 2 likes

adas🧦🌹 (@adastroworld): So-called “male loneliness crisis” yet you haven’t posted a picture of your hand

Image

Image

Reply to @arithmoquine, 10:17 UTC · 2 likes

henry (@arithmoquine): @CrispyWicks it's a compulsion i can't help who i am

Based

Quote tweet, 13:46 UTC · 14 likes

I will also do this

Quoting derek guy (@dieworkwear): I commonly do work in exchange for donations to a charity. Currently, my going rate is $10 to an animal charity if you want me to answer a style-related question. However, if you donate $50M to a many charities, I will help you find a pair of pants and even a pair of shoes.

Reply to @slimepriestess, 13:57 UTC · 3 likes

Ra (@slimepriestess): @AaronBergman18 if someone donated $50M to charities i actually liked i'd help them design a whole wardrobe if they wanted

I will do this for the low price of $49M

Tweet, 17:59 UTC · 23 likes

Banned from Google Cloud :(

Image

Thread, continued, 17:59 UTC · 3 likes

plz fix thx

Reply to @thoth_iv, 18:11 UTC · 15 likes

Thŏth (@thoth_iv): @AaronBergman18 Do you know why?

Making infinity google accounts + cloud projects with a web of tied-together backup emails and phone numbers to get around YouTube api rate limits lol

Reply to @GjMcGowan, 18:11 UTC · 2 likes

George McGowan (@GjMcGowan): @AaronBergman18 You know exactly why we did it and you’re lucky we don’t take your gmail as well

:(

Reply to his own earlier tweet, 18:12 UTC · 8 likes

Aaron Bergman, May 28, 2025: Making infinity google accounts + cloud projects with a web of tied-together backup emails and phone numbers to get around YouTube api rate limits lol

Pretty mundane “abuse”

They should just charge for the api!

Reply to @absurdlymax, 21:15 UTC · 1 like

Max Alexander (@absurdlymax): Unfortunately you were not wrongly banned

😢

Reply to @arithmoquine, 21:17 UTC · 3 likes

henry (@arithmoquine): what did he do??

Abolish the cloud police

Reply to @NoahTopper, 21:18 UTC · 2 likes

Quinoah 🔍⏸️ (@NoahTopper): #JusticeForAaron

Thank you ✊

Thursday, May 29, 2025

Quote tweet, 01:15 UTC · 27 likes

This is number 585893300128569 on the list of world’s most important issues but my brain treats “buying a house” as genuinely outlandish

Quoting Hunter📈🌈📊 (@StatisticUrban): oof

image attached

Thread, continued, 01:15 UTC · 13 likes

Not being able to buy a single family house within a short commute of like the 5 most expensive + desirable cities on earth is really the “upper middle class ivy reject has to go to merely very good college instead” of housing issues but like whatever it’s still suboptimal

Reply to @zeta_globin, 02:02 UTC · 2 likes

(The tweet being replied to by @zeta_globin is no longer available.)

Ya fr I could not afford that shit rn

Reply to @JesseTayRiver, 02:09 UTC · 3 likes

Jesse Smith (@JesseTayRiver): @manic_pixie_agi @AaronBergman18 American suburbs are the greatest!

It’s only a suburb if it’s within a 25 minute drive of DC NYC SF or Boston else it’s a sparkling exurb^2

Reply to @AndyMasley, 03:28 UTC · 3 likes

Andy Masley (@AndyMasley): Any followers good at designing business cards?

Image

Reply to @AndyMasley, 03:31 UTC · 2 likes

Andy Masley (@AndyMasley): @AaronBergman18 @MariGO2thepolls Wtf

image attached

Æňðÿ

Tweet, 05:57 UTC · 3 likes

So I guess there are just two totally different models for small purchase consumer financing (I’m lumping together burrito and MacBook here):

1) ~predatory interest rates

2) below-market interest rates to facilitate price discrimination

Image

Thread, continued, 05:57 UTC · 1 like

I guess people with lots of money just don’t bother with even free-lunch financing for some reason (behavioral econ stuff?)

Please note that this screenshot is entirely HYPOTHETICAL and I DO NOT NEED a new MacBook (😢)

Image

Reply to @inerati, 06:49 UTC · 1 like

liz (@inerati): i have always heard this but it only ever kinda made sense. like an extra fee? makes sense to cover the risk. but just not selling? wtf

I rented at 23 once just had to pay a premium

Quote tweet, 14:16 UTC · 12 likes

Not sure I’ve seen 4 different proper nouns I am totally unfamiliar with alongside “$X Billion” recently

Image

Quoting Puck (@PuckNews): NEW: Rhode, the beauty brand founded by Hailey Bieber, is getting acquired by e.l.f. Cosmetics in a $1 billion deal, the companies plan to announce later today.

@rachelstrugatz has exclusive details, link in thread 👇

image attached

Tweet, 14:26 UTC · 5 likes

Random thought: in areas with supply restrictions and plausibly fat tailed skill distributions—and I’m thinking about doctors and medicine—nativist “Americans first” policies are especially destructive…

Thread, continued, 14:26 UTC · 2 likes

If there are a limited number of doctors, it’s really important that the best Indian immigrants or whomever *do* displace their next-best American alternatives

Reply to @NTFabiano, 14:32 UTC · 4 likes

Nicholas Fabiano, MD (@NTFabiano): We should treat depression as a network of interconnected symptoms rather than a diagnostic label

I mean it’s pretty useful to have a term and label for that network of correlations

Quote tweet, 18:20 UTC · 19 likes

You might not like it, but this is what peak philosophy looks like

Quoting Alex Strasser (@AStrasser116): Oh wow, I did not expect this to be such a vigorous debate

image attached

Reply, 22:48 UTC · 1 like

Have you tried running your stuff through a base model to see what comes next?

Friday, May 30, 2025

Reply to @absurdlymax, 00:25 UTC · 2 likes

Max Alexander (@absurdlymax): @AaronBergman18

Omg a duck

Reply to @bayeslord, 00:25 UTC · 2 likes

bayes (@bayeslord): seriously what the fuck happened in 1971

Idk

Tweet, 00:53 UTC · 35 likes

Trying my best to convince my very normie but very talented (afaict) and hardworking soon-to-be-retired lawyer father to do something weird and unconventional

Poast tips

Thread, continued, 00:53 UTC · 9 likes

In all seriousness feel free to pitch me on things to pitch him

30 years of consumer protection law for the federal government

Reply to @parafactual, 00:55 UTC · 3 likes

You (@parafactual): @AaronBergman18 this is very vague tell us more about your father

I feel kinda funny bc I’m not certain how much he’s comfortable with me just detailing his life story on twitter

Sorry, I should check with him and confirm

Tweet, 05:15 UTC · 36 likes

Vibecoding is great because you can be extremely extra at basically no cost

Image

Thread, continued, 05:15 UTC · 9 likes

I don't know how to do any of this, but Claude does!

Reply to @littIeramblings, 13:57 UTC · 2 likes

sarah (@littIeramblings): lol I was about to do this

🦆

Tweet, 14:14 UTC · 13 likes

Why is JD Vance like 70% evil

Reply to @imhinesmi, 14:16 UTC · 3 likes

ian hines (@imhinesmi): @AaronBergman18 if he were less evil, would you know who he is?

Yeah I think so?

Quote tweet, 15:26 UTC · 5 likes

It's never woke or safety coded but Gemini often has the dumbest (quasi-) refusals

This isn't the only example, I just don't feel like digging the others up

Image

Quoting Aaron Bergman 🔍 (@AaronBergman18): Trying my best to convince my very normie but very talented (afaict) and hardworking soon-to-be-retired lawyer father to do something weird and unconventional

Poast tips

Thread, continued, 15:26 UTC · 3 likes

Hard to get a dumber-to-refuse prompt lol

Image

Tweet, 15:29 UTC · 2 likes

Wow, psyched to see offsetting getting its due among the masses!

Image

Reply to @Benthamsbulldog, 15:32 UTC · 20 likes

Bentham's Bulldog🔸 (@Benthamsbulldog):

image attached

Wonderful subtitle

Reply to @captgouda24, 15:33 UTC · 4 likes

Nicholas Decker (@captgouda24): Anyway, no more controversial tweets from now on. Just economics. I have some stuff I need to read on characterizing welfare impacts in distorted economies, y’all are gonna learn so much about Baqaee-Farhi

Noooooo

Tweet, 17:10 UTC · 47 likes

IRL personal experience has me updating against the signaling theory of college

Like actually yeah if you’re doing something marketing related it would be good to know something about marketing and probably like a marketing minor is good for that

Reply to @ApriiSR, 18:01 UTC · 13 likes

Aprii *️⃣ (@ApriiSR): @AaronBergman18 i think it's a tempting theory when you're the kind of person who could have learned all the things necessary for your job much much faster elsewise, but in fact that isn't mostly what people would do

I still think it’s more than 50% of the story, just maybe not 90%

Tweet, 18:27 UTC · 5 likes

I’m pretty surprised we haven’t seen more visceral impacts of vibecoding yet

Maybe the startup product launch timelines are just a bit too long, could be coming soon

Reply to @AndyMasley, 18:40 UTC · 7 likes

Andy Masley (@AndyMasley): All criticisms of liberalism saying it's atomizing and alienates people from community always leave me asking "Yes but what is the criticism?" We should want people to be radically free to flee and join communities as they want.

I mean I’m very on board with liberalism but it’s not a truly fundamental value

The not-impossible critique is just that individual radical freedom decreases hedonic wellbeing at large. There’s no logical reason this can’t be the case

Reply to @cremieuxrecueil, 18:44 UTC · 73 likes

Crémieux (@cremieuxrecueil): Does testosterone make people have more masculine economic preferences?

Does it make people act more like Republicans?

We now have a large, pre-registered double-blind randomized controlled study that provides an answer🧵

image attached

This is interesting but we should have a decently strong bias towards “everything is mediated by (epigenetic?) changes that take at least 2 weeks and usually 4”

*Acute* testosterone administration does (little-nothing for muscle mass and strength)!

Saturday, May 31, 2025

Tweet, 02:42 UTC · 3 likes

Just now, I assumed the explanation to smth was "Opus is uniquely horny for Ezra Klein" but actually there was an arcane string matching algorithmic change introduced by python-Levenshtein v0.22

Very cool that these two hypotheses can explain the same thing

Image

Reply to @arthur_wright_, 03:04 UTC · 2 likes

Arthur Wrong 🔸 (@arthur_wright_): Yes. You have to be careful with the labeling because the terms have no legally binding definition, but the small number of genuinely "pasture raised" (Vital Farms, iirc) eggs are significantly higher welfare.

Do you have a take on whether eg Vital is actually morally fine/even positive?

Reply to @livgorton, 05:09 UTC · 2 likes

Liv (@livgorton): the feminine urge to get a lob (long bob) whenever any emotional turmoil arises

Bonk don’t be emotionally turmoiled

Quote tweet, 09:57 UTC · 14 likes

Comic book villain-ass politics

Quoting Katie Bergh (@Katie_Bergh): Parents of young kids in the “berry budget” age range, weigh in: how long would $10 for fruits and vegetables last you?

$10 is the *monthly* amount the Trump administration just proposed for toddlers & preschoolers on WIC—a 62% cut.

Reply to @EAheadlines, 10:07 UTC · 3 likes

EA Lifestyles (@EAheadlines): I have fed my toddler $10 of fruit in one meal before 😬

Based and berry-pilled

Reply to @ohabryka, 15:51 UTC · 27 likes

Oliver Habryka (@ohabryka): Bro, all metrics of growth completely crashed after FTX, media coverage drastically shifted, a huge number of top talent left the community, and we saw huge 1+ point shifts on 10 point satisfaction scales and related metrics with EA (showing up as a 4+ SD event on basically all measurements).

There is probably not a single graph of a popular metric associated with the EA community where you can’t clearly see the effects of FTX. This includes anything you can think of that’s to do with reputation.

If you think it had minor effects, there has never been and will never be anything that could possibly have anything but “minor” effects.

Ok u fought me and won

Reply to @peterwildeford, 16:00 UTC · 3 likes

Peter Wildeford🇺🇸🚀 (@peterwildeford): @AaronBergman18 this is so cool

what does it do?

Tree search of Substacks - who recommends who!

Reply to @eshear, 16:02 UTC · 3 likes

Emmett Shear (@eshear): Doing Bayesian updates in your head is unrelated to rationality in the same way that doing Newton’s laws of motion in your head is unrelated to throwing and catching a ball.

💯

Quote tweet, 17:24 UTC · 2 likes

This but for economics

I’ve never really understood why test-train split is standard practice in ML but not econ

Given the empirical success of ML, social science should consider what it can adopt!

Quoting Tim Duffy (@timfduffy): @lu_sichu wild that straight up not having a test-train split is the most common issue here

Tweet, 20:28 UTC · 14 likes

Computers are so interesting I wish they had better vibes/aesthetics

Reply to @SpacedOutMatt, 20:36 UTC · 2 likes

Matt 🔸 (@SpacedOutMatt): Did you find it harder to share the childcare load with others when exclusively breastfeeding? And for you personally which did you feel like was better for your sanity?

Smol Matt incoming?? 👀

Reply to @owldudeok, 20:36 UTC · 2 likes

(The tweet being replied to by @owldudeok is no longer available.)

Lol it’s totally unfair ur right im just gesturing at a vibe I personally sense

Reply to @SpacedOutMatt, 20:40 UTC · 2 likes

Matt 🔸 (@SpacedOutMatt): Lmao I’m not even married yet

Cheems mindset

Reply to @Adellama_, 23:00 UTC · 2 likes

Ada🔸 (@Adellama_): computers have great aesthetics. i like computers just because of their aesthetics

image attached

I was thinking of more like CS as an intellectual thing but yes this looks cool


End of May 2025. Previous month: April 2025. Next month: June 2025. Back to the archive index.