Aaron Bergman's tweets, January 2026
Every public tweet, reply, quote tweet and retweet that Aaron Bergman (@AaronBergman18 on X, formerly Twitter) posted in January 2026 and has not deleted: 334 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: December 2025. Next month: February 2026.
Thursday, January 1, 2026
Tweet, 02:41 UTC · 57 likes
Who the fuck invented the top button and tie norm in men’s formalwear
We don’t have to live like this, we can demand to not be choked by tradition
Reply to @izzyz, 02:45 UTC · 5 likes
izzy (@izzyz): @AaronBergman18 i think it has something to do with a fence or some shit
Fuck that
Reply to @mattyglesias, 02:46 UTC · 33 likes
(The tweet being replied to by @mattyglesias is no longer available.)
Thank you Matt but I’m not going back to Macy’s
Reply to @BigGoldMachine, 05:23 UTC · 3 likes
BigGoldMachine (@BigGoldMachine): @AaronBergman18 Buy a shirt that fits
never
Reply to @boazbaraktcs, 06:32 UTC · 59 likes
Boaz Barak (@boazbaraktcs): Wrote this for young people worried about AI.
https://www.lesswrong.com/posts/fwQburGDyGoSSweT9/you-will-be-ok
> It is also possible that the future will be much worse, even by my cynical eyes
Sounds like the title is wrong then?
Quote tweet, 06:38 UTC · 35 likes
If the title of your blog post/whatever is a proposition, it needs to be true by your own lights and directly supported by the content of the piece
Admittedly I take pretty big issue with this post but I feel like I see the general problem not infrequently
Quoting Boaz Barak (@boazbaraktcs): Wrote this for young people worried about AI.
https://www.lesswrong.com/posts/fwQburGDyGoSSweT9/you-will-be-ok
Reply to @livgorton, 07:48 UTC · 6 likes
Liv (@livgorton): +1 for proper cloth!! can order online if you trust yourself to measure properly.
I do not trust myself to measure correctly
Tweet, 07:50 UTC · 20 likes
I like Opus 4.5 too but I don't really get why everyone is freaking out about how amazing it is
Reply to @livgorton, 07:50 UTC · 3 likes
Liv (@livgorton): next time I'm in DC I can help you
🥺
Quote tweet, 07:52 UTC · 13 likes
Lmao
Quoting Amritanshu (@Will_Hackspeare): @grok @GjMcGowan @NathanpmYoung why would it just voluntarily generate a new shirtless pic of a random asian man in response to "deeply troubling"????
Reply to @peterwildeford, 14:19 UTC · 2 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): Claude Code
I use it in cc!
Reply to @cljack, 14:21 UTC · 3 likes
Charlotte Lee (@cljack): @AaronBergman18 your country needs you
(quoting @ArmandDoma: some incredibly sexy effective altruist needs to seduce Bezos’ ex-wife so she donates all her money to like vaccines for malaria and stuff instead of nonprofits that declare a new duplex to be genocide)
On it 👀
I’ll make sure she sets aside a few billion for airplane research, dw
Friday, January 2, 2026
Reply to @diamondminercat, 00:24 UTC · 5 likes
ming cat ! (@diamondminercat): how should I resolve the market?
Oh man, the answer is “yes but then I lost it for extremely weird reasons so I don’t *currently* have anything”
Tweet, 00:42 UTC · 32 likes
2026 goals:
1) fix myself
2) fix everything else
Reply to @ChallengeBelief, 00:59 UTC · 2 likes
Challenge Your Beliefs (@ChallengeBelief): @AaronBergman18 In that order?
Sadly mostly yes
Reply to @frances__lorenz, 01:27 UTC · 2 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
So true
Reply to @RafaRuizdeLira, 01:40 UTC · 4 likes
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): Do you guys think she's single
(quoting @nypost: Socialist Zohran Mamdani’s wife Rama Duwaji appears to wear luxury $630 boots to swearing in ceremony https://trib.al/rbsYcH9)
Ya shoot ur shot king
Reply to @cat_warden_, 01:46 UTC · 2 likes
Katherine (@cat_warden_): :( rip
Rip 😢
Reply to @IsaacKing314, 13:53 UTC · 2 likes
Isaac King 🔎 (@IsaacKing314): @AaronBergman18 @diamondminercat Are you looking for new things?
Sorta - will get back to you
Saturday, January 3, 2026
Quote tweet, 00:37 UTC · 5 likes
It’s a haven of ducks and airplanes for me
Quoting liz (@inerati): i’ve successfully made my instagram reels feed show me nothing but sick edits of beavers being awesome. i have decided this is way better than this website and am now retiring as a twitter poster. thank you for your time.
Tweet, 18:01 UTC · 3 likes
Wait the ~official~ charges are dealing with cocaine? wtf
Reply, 18:42 UTC · 4 likes
(The tweet being replied to is no longer available.)
Omg goss congrats‼️
Reply to @celestepoasts, 20:47 UTC · 2 likes
(The tweet being replied to by @celestepoasts is no longer available.)
Lmao
Reply to @peterwildeford, 21:11 UTC · 4 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford):
Step 1) get invited on the Daily Show
Sunday, January 4, 2026
Quote tweet, 00:10 UTC · 6 likes
My god this is so based
Got bless San Francisco
Quoting Kylie Robison (@kyliebytes): they put my steve madden platform boots in the paper of record
Reply to @thewreathwright, 18:08 UTC · 1 like
Laurel Weaver (@thewreathwright):
Dog
Quote tweet, 19:14 UTC · 11 likes
Waow $69k (+1)
Quoting Aaron Bergman 🔍 (@AaronBergman18): 🚨 IT'S BACK🔥 AND IT'S BIGGER 😤THAN EVER 🚨
✨ PRESENTING👀 THE 2025 EA TWITTER🐦HOLIDAY 🕎🎄FUNDRAISER 🤑
(in support of the EA Animal Welfare fund🦐🐓🐷🐟)
Reply to @justalexoki, 20:09 UTC · 1 like
taoki (@justalexoki): genuinely how do you fix depression
This + there are just a lot of things to try
Psychedelics, MAOIs, ketamine, and pramipexole are all underused
Quote tweet, 20:14 UTC · 6 likes
Correct take
Quoting Liv (@livgorton): what I want so bad is something like a memory free chat that still appears in my history but isn't an incognito chat. a safe space to ask my dumbest questions without fear... a space to hide my stupidity from claude but not from from myself.
Reply to @AndyMasley, 21:56 UTC · 6 likes
Andy Masley (@AndyMasley): There have been 3 points where LLMs felt like new raw magic to me:
The release of ChatGPT
O3
Opus 4.5 + Claude CodeI'm expecting this one to really flip a lot of skeptics over the next few months
(quoting @simonw: It genuinely feels to me like GPT-5.2 and Opus 4.5 in November represent an inflection point - one of those moments where the models get incrementally better in a way that tips across an invisible capability line where suddenly a whole bunch of much harder coding problems open up)
I still don’t really get it
Reply to @AndyMasley, 22:00 UTC · 5 likes
Andy Masley (@AndyMasley): @AaronBergman18 Have you played around with claude code much? That's where it's at specifically for me
Yeah a bunch, 4.5 in CC more than any other model/interface, although evidently not very much at the limit of what it can do
Tweet, 22:12 UTC · 12 likes
Grok undressing women on command under highly visible, popular posts is at the very least very bad taste
Thread, continued, 22:13 UTC · 4 likes
Reply to @RosieCampbell, 23:14 UTC · 22 likes
Rosie Campbell (@RosieCampbell): Just tried this, recommend
(quoting @frankdegods: just mass cancelled $27k/year in subscriptions made a claude code skill that: 1. reads credit card statements/extracts subscriptions 2. automatically asks follow-up q's to clarify which ones you want to cancel 3. actually opens chrome and literally cancels them for you)
Lol I should probably do this and expect maybe I’ll save a couple hundred (might still be paying for Cursor tbh) but the idea of not realizing $27k is hilarious to me
I’m very very very sure I don’t have $27k in unused subscriptions
Reply to @mattyglesias, 23:44 UTC · 35 likes
(The tweet being replied to by @mattyglesias is no longer available.)
50/50 should be purple rather than white
Monday, January 5, 2026
Reply to @arabelladevine, 03:10 UTC · 7 likes
arabella ※ (@arabelladevine): just noticed a man smelled of fresh laundry as we passed each other in the corridor. down so bad i should be shot out of a tree like a rabid raccoon
Reply to @arabelladevine, 03:28 UTC · 8 likes
arabella ※ (@arabelladevine): baffling but i’m happy for you
Hmm trying to figure out the “baffling” interpretation
My interpretation is just “women like the smell of fresh laundry - might want to remember that”
Reply to @arabelladevine, 03:30 UTC · 6 likes
arabella ※ (@arabelladevine): is this news??? everyone likes the smell of fresh laundry
Not everyone describes themselves as going rabid in response lmao
Reply to @arabelladevine, 03:45 UTC · 3 likes
arabella ※ (@arabelladevine): causality backward. down bad therefore noticing how a random man smells and that it’s good
Oh hmm
Thread, continued, 03:45 UTC · 27 likes
Reply to @NoahTopper, 07:21 UTC · 4 likes
Quinoah 🔍⏸️ (@NoahTopper): I'm reasonably happy to call myself a moral realist, but of course that comes down to what precisely we mean by "realist". This is the thing I'm gesturing at:
(quoting @allTheYud: To try again to summarize at length what cannot well be summarized at all, my view on is-ought / orthogonality, is: "There are objective answers but not objective questions." You embody certain questions or meta-questions. Not all cognitive engines would. You look at a painting…)
Ugh should I read this
Reply to @NoahTopper, 07:23 UTC · 2 likes
Quinoah 🔍⏸️ (@NoahTopper): You should read the much longer sequence:
(quoting @NoahTopper: I'm a big fan of Eliezer's Metaethics Sequence on this topic https://www.lesswrong.com/w/metaethics-sequence)
Oy vey
Reply, 07:23 UTC · 1 like
I’m a maybe
Tweet, 08:04 UTC · 277 likes
God I really hope we get lucky with this whole AI thing
Thread, continued, 08:04 UTC · 17 likes
Luck seems necessary but not sufficient so the other bits too
Quote tweet, 08:12 UTC · 676 likes
Seems bad
How did academia go from “modest bias in favor of racial minorities in undergrad admissions” to this?
Quoting Arjun Panickssery (@panickssery): For the 2018–2019 year, Berkeley LSI Committee recruiting for Life Sciences posts cut 76% of candidates in a first pass based only on their DEI statements
Thread, continued, 08:15 UTC · 17 likes
Gotta keep my followers on their toes
Reply to @animalologist, 16:48 UTC · 6 likes
taco belle (@animalologist): I’m a girl but do you guys think convivially fighting people would be a good workaround for being touch-starved, if you’re a guy
(quoting @animalologist: boys should be playing contact sports more. a lil martial arts, some grappling, literally just shove each other around it’s good for you)
No lol
Quote tweet, 17:58 UTC · 18 likes
Someone donated $6500??? Like today??? Based??????????
Thank you mysterious person (feel more than free to identify urself)!!!!!
Quoting Aaron Bergman 🔍 (@AaronBergman18): 🚨 IT'S BACK🔥 AND IT'S BIGGER 😤THAN EVER 🚨
✨ PRESENTING👀 THE 2025 EA TWITTER🐦HOLIDAY 🕎🎄FUNDRAISER 🤑
(in support of the EA Animal Welfare fund🦐🐓🐷🐟)
Reply to @Vert_Noel, 18:45 UTC · 5 likes
Noël Michel ⏹️ (@Vert_Noel): @AaronBergman18 An everyday morning prayer for us all at the hinge of history
🙏
Tuesday, January 6, 2026
Quote tweet, 00:32 UTC · 14 likes
Correct take
Quoting Geoffrey Litt (@geoffreylitt): We need a shorthand way of saying:
"An AI did the work, but I vouch for the result"
Saying "I did it" feels slightly sketchy, but saying "Claude did it" feels like avoiding responsibility
Reply to @geoffreylitt, 00:32 UTC · 15 likes
Geoffrey Litt (@geoffreylitt): We need a shorthand way of saying:
"An AI did the work, but I vouch for the result"
Saying "I did it" feels slightly sketchy, but saying "Claude did it" feels like avoiding responsibility
🤝
Tweet, 04:29 UTC · 2 likes
Even if they had the exact same first order A[GS]I safety development profiles, Anthropic would still look way better than OpenAI and others just because of who the windfall flows to and what I expect they’d do with trillions of dollars
Silly to pretend this doesn’t matter a lot
Reply to @daniel_271828, 05:14 UTC · 3 likes
zzDANIEL_ETH (@daniel_271828): Hot take but intellectuals (both academics & ppl in intellectual communities) often misinterpret (geo-)politics as about prestige when it’s instead about power. Hence why you get takes like “if the U.S. invades Venezuela, then they can’t oppose their adversaries doing invasion!”
I think I agree with the first sentence but you *can* try to set norms in a decentralized way via game theory shenanigans even in a Realist world with no mirage of to supranational authority
Like the US can of course do just about what ever it wants with its large military and economy but one thing it can’t do is will other powerful countries to obey norms unilaterally - one can imagine a world where the US had respected sovereignty for better and worse over the last century and then the Chinese would have to think longer and harder about whether they want to disrupt that norm by invading Taiwan (they still might choose to, but it would be costlier)
Reply, 09:42 UTC · 7 likes
joyce (@henloitsjoyce): 2nd day on hinge as a girl in sf
Tapping the sign
Reply to @ptrschmdtnlsn, 09:46 UTC · 84 likes
Peter Schmidt-Nielsen (@ptrschmdtnlsn): My favorite is when someone refers to something as "big data", and it's just like ten billion records, so it fits in a microsd card. A single slim 2U server can fit petabytes of storage!
(quoting @Duderichy: the average software engineer has no idea how much compute/storage can fit in a single rack in 2025)
It’s big data if it doesn’t fit in RAM
Quote tweet, 18:02 UTC · 6 likes
Number keeps mysteriously going up
I mean you *can* remain anonymous but IMO if you're donating thousands of dollars you should get appropriate credit for it lmao
Quoting Aaron Bergman 🔍 (@AaronBergman18): Someone donated $6500??? Like today??? Based??????????
Thank you mysterious person (feel more than free to identify urself)!!!!!
Quote tweet, 18:05 UTC · 4 likes
Stickers!!! If you donated at least $1k (could be across multiple donations, not too late 👀) please fill out the google form below so we can mail (at least) one to you!
Quoting Aaron Bergman 🔍 (@AaronBergman18): Number keeps mysteriously going up
I mean you *can* remain anonymous but IMO if you're donating thousands of dollars you should get appropriate credit for it lmao
Reply to @daniel_271828, 23:36 UTC · 4 likes
zzDANIEL_ETH (@daniel_271828):
Why not both
Wednesday, January 7, 2026
Reply to @GarettJones, 02:53 UTC · 20 likes
Garett Jones (@GarettJones): "Effective altruists, are you paying attention?"
Lol was I supposed to go down there myself
Tweet, 03:17 UTC · 76 likes
Blog post 👀
Thread, continued, 03:17 UTC · 18 likes
https://www.aaronbergman.net/p/public-intellectuals
Quote tweet, 03:59 UTC · 5 likes
Based reasoning
Quoting alice (@aliceisplaying): claude opus 4.5 confirmed tpot 🥹
Reply to @vren, 04:05 UTC · 2 likes
vren (@vren): @AaronBergman18 @GarettJones @hamandcheese Is what you're doing right now of greater expected impact?
Yes
Reply to @DrewSav, 05:57 UTC · 2,390 likes
Drew Savicki (@DrewSav): I fully buy the idea that misleading maps have convinced Trump that Greenland is massive
Hmm that’s still pretty big
Reply to @rgblong, 06:04 UTC · 2 likes
Robert Long (@rgblong): who is best to read / follow for advice on using AI e.g. Claude Code? especially interested in:
-productivity and todo wrangling (especially for the distractable)
-research assistance
-editing
-learning
I probably don’t have a good answer regardless lol but are you interested specifically in the context of learning and creating philosophy?
Reply to @quant_____, 07:52 UTC · 6 likes
quant (@quant_____): Market is so cooked my librarian has an econ PhD
F
Quote tweet, 08:05 UTC · 5 likes
Not sure it’s fundamentally any different than me getting $15 bowl slop when I don’t really need while people live on a couple dollars a day but ostentatious luxury consumption is super off-putting
$300k cars should be safe, legal, and even rarer than they are
Quoting Nikita Bier (@nikitabier): @Kalshi Finally.
Lambo time.
Reply to @lukalotl, 08:22 UTC · 2 likes
luka (@lukalotl): what's the best thing to do when models will be 10x smarter within a year?
Idk
Tweet, 13:04 UTC · 29 likes
Do you have any manual control over the Waymo you’re in as a passenger? Anything mediated 100% mechanically rather than via software? Can you always open the door?
Reply to @oscredwin, 13:28 UTC · 6 likes
Andrew Rettek (@oscredwin): I'm aware of one EA project back in 2018 to overthrow the socialist government of Venezuela. Eventually they figured out that, to put it lightly, tractability was an issue.
(quoting @GarettJones: "Effective altruists, are you paying attention?")
What
Reply to @marcusabramovi1, 17:36 UTC · 9 likes
Marcus Abramovitch (@marcusabramovi1): I understand this might feel scary but this is the exact situation a child is in when they are in a car
Yeah I’m not personally worried for myself, but interested in the plausibility of cyberattacks and how bad they’d be
I realize I didn’t make this clear at all lol
Reply to @cljack, 18:02 UTC · 3 likes
Charlotte Lee (@cljack): similarly, I always thought there's got to be extremely high intelligence ROI to paying several people to do nothing other than ride the Acela between DC and NYC every day
(quoting @deanwball: I really wonder how many uber black drivers in dc nyc and sf are intelligence assets. So many people I know have extremely sensitive conversations on the phone while in Ubers (guilty). Plausibly yet another benefit of robotaxis!)
U have the wrong kind of autism or else it could be u
Reply to @NoahTopper, 18:27 UTC · 3 likes
Quinoah 🔍⏸️ (@NoahTopper): idk man have you heard pigeon hour
😢
Thursday, January 8, 2026
Quote tweet, 00:41 UTC · 7 likes
If they need a guy 👋
Quoting Lisa Desjardins (@LisaDNews): BREAKING: The Trump administration plans to put money raised from seizure of Venezuelan oil into bank accounts *outside* the U.S. Treasury -- they told lawmakers today per multiple sources familiar.
Sources said they understood these as similar or decidedly "off-shore" accounts.
Asking the WH for clarification.
Tweet, 01:48 UTC · 55 likes
US illegal immigration is generally and by default morally fine/good and should be de facto non-prosecuted, maybe via executive guidance like DACA
Everything is fine and it could just be a non-issue
Reply to @AndyMasley, 04:17 UTC · 10 likes
Andy Masley (@AndyMasley): A general goal I have for blogging this year is a mini series on EA from the inside out. Instead of presenting it as a series of steps "Oh you want to do good? Have you ever thought about the MOST good?..." etc., I'd like to just dive straight into the center core high level stuff first. There's not enough high level writing from the inside from people who have been involved for a long time that's communicating to a general audience. I think the other slow intro way leaves people feeling justifiably like we're not being straight with them.
(quoting @AaronBergman18: Blog post 👀)
Based
Reply to @stanfordNYC, 05:47 UTC · 3 likes
Bernard Stanford ✡︎ (@stanfordNYC): I disagree with this. The presence of others places a number of encumbrances on citizens that they didn't consent to. Even noncitizens are entitled to healthcare, shelter, use of infrastructure, protection from criminals, all of which has costs borne by the public. In the long run, illegal immigrants or their progeny become citizens and change the makeup of the nation, shifting it away from the balance of the original citizenry. This matters because we live in a democracy, and your fellow citizen can vote on your taxes, your freedoms, the education of your children, even your conscription and forcible deployment into a war. The citizenry have a right and duty to police entry for all these reasons, even assuming we could deal with criminals and uninsured drivers etc.
I’m honestly sympathetic to much of this in principle - I think we don’t disagree as much as it first seems
I’m not making a claim about whether to change illegal immigrants into citizens with voting rights - I’m (or at least was intending to) make a claim about whether we have to do anything active to get them out of the country
Of course the natural born citizen thing (ie their children) is legit but my sense is that the children of illegal immigrants just empirically basically aren’t importing a hostile or backwards culture and aren’t even that disruptive to national partisan political balance (and in some sense insofar as there’s a modest democratic tilt that’s cosmically just vis a vis senate representation)
Tepid non-enforcement isn’t really a sexy or very desirable option in some abstract sense but it seems like it would probably in fact work fine for the foreseeable future. It basically *is* working fine anywhere ICE isn’t
On the fiscal stuff: IIRC the literature says that it nets out positive once you include additional generations. Idk, I’m also a bleeding heart lib, I don’t think it’s bad for tax dollars to pay for normal societal infrastructure for non-citizens, but tbf this is a legit political debate to have
Reply, 06:07 UTC · 1 like
Hey I’m all for lowering rent by building more!
Tweet, 18:02 UTC · 7 likes
. @neuronpedia’s feature enhancement/steering thing is working basically exactly as you’d expect (I have no idea what I’m doing but also turned up the activation on a shrimp-related feature)
Reply to @noncanonicAleae, 18:07 UTC · 6 likes
alea ☄️⏺️⛵️ (@noncanonicAleae): Trivially false. Only say what you actually believe if you want to get crucified. I would rather be safe.
(quoting @AaronBergman18: Blog post 👀)
Tweet, 18:17 UTC · 8 likes
Reply to @GjMcGowan, 18:18 UTC · 4 likes
George McGowan (@GjMcGowan): @AaronBergman18 @neuronpedia Aaron version that just goes "girls girls girls girls girls girls girls girls girls"
Only need a minor tweak to make that one
Friday, January 9, 2026
Tweet, 03:17 UTC · 2 likes
Vibecoded a Gemini watermark remover
Tweet, 03:27 UTC · 8 likes
At some point someone is gonna think they have AI psychosis but will have actually discovered/made something major
Reply to @morallawwithin, 03:34 UTC · 2 likes
florence ⊮ (@morallawwithin): @MariGO2thepolls @AaronBergman18 Aaron might have to change his name
I wish
Quote tweet, 03:40 UTC · 10 likes
What do y’all still use humans for?
(Mostly serious question - obviously I mean in the space of tasks or problems or information acquisition - I hope everyone still talks to humans for other reasons)
Quoting Peter Wildeford🇺🇸🚀 (@peterwildeford): Here's currently how I'm using each of the LLMs
Reply to @AndyMasley, 04:52 UTC · 19 likes
Andy Masley (@AndyMasley): I'm organizing underground bookclubs for revolutionary anti-government texts they don't want you to know about
image attached
(quoting @HHSGov: https://realfood.gov/)
Wait isn’t “ultraprocessed food bad” one of like 4 legit consistent findings in nutrition research
Reply to @RosieCampbell, 05:02 UTC · 3 likes
Rosie Campbell (@RosieCampbell): Not according to https://open.spotify.com/episode/3alsDFJlUpLPWadeDgmjxd?si=fA6bju_ESO-qJRnsaLIt_A
Ah hmm thanks
Reply to @livgorton, 06:00 UTC · 3 likes
Liv (@livgorton): I am going to be in DC in about week for a ~week! This is mostly a social visit rather than work but am also very keen to fill my time with meeting new people (for social/research/whatever reasons) or going to whatever events are happening :) Lets hang out 👉👈
Big if true
Reply to @ATabarrok, 06:21 UTC · 10 likes
Alex Tabarrok (@ATabarrok): The Tyranny of the Complainers
Increasingly, public institutions seem to exist to manage the obsessions of a tiny number of neurotic—and possibly malicious—complainers.
https://marginalrevolution.com/marginalrevolution/2026/01/the-tyranny-of-the-complainers.html
Plausibly we’re on the downward sloping (ie atp more concentrated complaining is worse) part of the curve but isn’t it basically always true that a smalll minority of the population “complains” ie actively tries to accomplish some sort of change?
So I expect “a tiny minority accounts for most of the rabble rousing” to be a true headline also in worlds where the rabble rousing is good and perhaps at a sub-optimal level
Quote tweet, 06:33 UTC · 8 likes
Tapping the sign
(I don’t know this person but she’s evidently smart and accomplished; plausibly this is why she’s secure enough to admit a specific general knowledge gap)
Quoting Jane Manchun Wong (@wongmjane): Just realized the “Washington” in The Washington Post is DC, not the state where Seattle is located
Up until now, I genuinely thought Bezos bought them because it’s the local newspaper near Amazon HQ
I’m gonna need to lie down.
Reply to @livgorton, 06:59 UTC · 3 likes
Liv (@livgorton): my guess is 70% yes
75%
Quote tweet, 16:10 UTC · 9 likes
Thank you to Matt for showing the world my take
The people can decide for themselves who to believe 😤
Quoting Tiger Lava Lamp (@DaCaveOfWonders): I don't know if this is how @AaronBergman18 wanted to make the blog, but I'm still jealous and I'm counting it as 4
Thread, continued, 16:10 UTC · 2 likes
Possibly I should in fact get a fitting shirt through
Tweet, 18:19 UTC · 3 likes
For better or worse, large donations are really what matter (from recent fundraiser, unofficial and incomplete data)
Quote tweet, 18:32 UTC · 15 likes
It’s starting
Quoting INVESTMENT HULK (@INVESTMENTSHULK): WOW IT’S ALMOST LIKE SOMEONE INVENTED A ROBOT THAT CAN WRITE COMPUTER CODE FOR YOU HAPPENED.
Tweet, 18:48 UTC · 19 likes
Fuck my comparative advantage was (unironically I think) being able to make stuff with Claude Code when it was just slightly not good enough for other people to
Need to retvrn to Opus 4.1 to protect my wages
Tweet, 22:04 UTC · 18 likes
I wanna see if I can start a beef with Matt Yglesias over this
In some sense it’s trivially true that my shirt doesn’t fit, but my thesis is that this is not my fault (or yours)
He offers (good) advice for making the best of a bad system, but let's interrogate the system...
Thread, continued, 22:04 UTC · 3 likes
As far as I know I don’t have an unusually big neck; the collar of every other shirt (like a polo I wouldn’t wear with a tie) fits fine
I just checked to confirm! And on this one the top button was as easy to button as the other buttons
Thread, continued, 22:04 UTC · 4 likes
I admit that I actually have some autistic tendencies around clothes comfiness but in general my complaints are that the outfit makes me too warm or the fabric is insufficiently soft
It’s only the neck issue in the narrow case of “dress shirt you’d wear with a tie”
Thread, continued, 22:04 UTC · 2 likes
And I'm not alone!
https://chatgpt.com/share/696175f4-45e4-8004-a602-9e7ceb1255b3
https://claude.ai/share/51c20d67-ad0d-4f36-8052-08ce847e58e9
Tweet, 23:09 UTC · 10 likes
The optimal amount of AI-written content is not zero, and not just because the enforcement to get to zero would be draconian/costly
Saturday, January 10, 2026
Reply to @moultano, 01:03 UTC · 3 likes
Ryan Moulton (@moultano): When the ibuprofen kicks in.
What
This is very foreign to me lol
Tweet, 01:16 UTC · 3 likes
The operations and infrastructure at the very top of the US federal government are pretty incredible
Reply to @albrgr, 01:32 UTC · 9 likes
(The tweet being replied to by @albrgr is no longer available.)
Consider this my comms application
Tweet, 04:44 UTC · 926 likes
Bruh, just got "I don't think this is going to work"'d on *before* a first date bc of differing opinions on Ezra Klein
Thread, continued, 04:44 UTC · 161 likes
Specifically whether his Kirk NYT obituary piece was deeply immoral for lending legitimacy to Kirk (and whether therefore we should lose a bunch of respect for Ezra)
(I was on Team Ezra)
Thread, continued, 04:44 UTC · 103 likes
I mean that was the proximate thing but genuinely I think it probably was representative of importantly different values and dispositions (at the very least degree of psychological openness) so the whole situation is amusing + she did nothing wrong by nixing the date
Thread, continued, 04:44 UTC · 73 likes
...But I still would have met irl lol
I had a whole dinner reservation like an adult!
Thread, continued, 04:44 UTC · 30 likes
Form your own take:
https://www.nytimes.com/2025/09/11/opinion/charlie-kirk-assassination-fear-politics.html
No paywall: https://archive.is/Sr0bJ
Reply to @GjMcGowan, 06:02 UTC · 37 likes
George McGowan (@GjMcGowan): I hate him because I find his voice whiny and he’s 10% too woke, we are not the same
I agree he's slightly too woke
Reply to @claudeai, 06:14 UTC · 3 likes
Claude (@claudeai): Some delightfully specific things people are building with Claude Code lately.
Lol people are mad that Anthropic is stopping a “use Anthropic’s product for free” service? And nobody got arrested or financially ruined? That’s the accusation?
Looks like a pretty tame/constrained reaction. They’re playing the cat and mouse game, not defecting
Reply to @Michael_Druggan, 16:00 UTC · 16 likes
Michael Druggan (@Michael_Druggan): @AaronBergman18 Dodged a bullet.
Yeah
Reply to @RafaRuizdeLira, 16:06 UTC · 8 likes
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): Ah yes, the "EA/rat to hated by both left-wingers and right-wingers" pipeline
😌
Quote tweet, 18:02 UTC · 5 likes
So true
Quoting EA Forum Posts (@EAForumPosts): "Public intellectuals need to say what they actually believe" by Aaron Bergman (@AaronBergman18)
Tweet, 19:46 UTC · 11 likes
Claude code + chrome has fast takeoff vibes
☹️
Tweet, 20:12 UTC · 12 likes
We need gradients beyond just “psychosis” for describing the phenomenon of being misled by an AI
Reply to @valleydipper907, 20:17 UTC · 3 likes
(The tweet being replied to by @valleydipper907 is no longer available.)
Thx
Tweet, 23:18 UTC · 22 likes
Would ppl want a Claude .ai "search chats easily and well including the summary, maybe even semantic search" Chrome extension
Sunday, January 11, 2026
Reply to @livgorton, 05:30 UTC · 2 likes
Liv (@livgorton): i think he do a lot of sus stuff besides the stuff he's currently in prison for right? like a bunch of people left in 2019 or something because of him lying ?
Deleted
Reply to his own earlier tweet, 06:05 UTC · 14 likes
Aaron Bergman, January 10, 2026: Would ppl want a Claude .ai "search chats easily and well including the summary, maybe even semantic search" Chrome extension
Fast full text search (not just titles which is the current system Anthropic has) maybe coming your way
(video, plays on X)
Tweet, 06:06 UTC · 27 likes
Ok fine 4.5 is pretty crazy, it just keeps doing things fr
Reply to @daniel_271828, 06:09 UTC · 7 likes
zzDANIEL_ETH (@daniel_271828): I want to see more technologies plotted on this magicalness index
Everything else is at a 0
Thread, continued, 06:09 UTC · 2 likes
Jk I’ll cook something
Reply to @NathanB60857242, 06:36 UTC · 4 likes
Nathan Barnard (@NathanB60857242): @AaronBergman18 @daniel_271828 i really enjoy how persistantly you point out how crazy planes are
🫡✈️✈️✈️✈️✈️
Reply to @livgorton, 07:00 UTC · 4 likes
Liv (@livgorton): oops I just realised I am in fact not 24 and a few years is a lie (just turned 26)
Not Consistently Candid smh
Reply to @LuisCostigan1, 11:09 UTC · 2 likes
Luis Costigan 🔍 (@LuisCostigan1): clearly a 10/10 in that case
Nothing compared to a flying metal tube
Reply to @fede_intern, 17:35 UTC · 18 likes
Fede’s intern (@fede_intern): A neurodivergent researcher spent a year reverse engineering Coca Cola. Using mass spectrometry, he created a chemically identical replica. Big moats no longer exist. Here you have the recipe.
Time to create a local version of it?
10g caffeine per liter is hospitalization territory
That’s not what’s in coke - that would be 200 cups of coffee per 2L bottle
Reply to his own earlier tweet, 22:32 UTC · 2 likes
Aaron Bergman, January 11, 2026: Fast full text search (not just titles which is the current system Anthropic has) maybe coming your way
Comment/dm if you wanna test (outside the Google store system like I just send you a .zip Drive link to load in dev mode)!
Monday, January 12, 2026
Quote tweet, 02:15 UTC · 159 likes
Fuck Donald Trump
Trying his hardest to ruin a good and above-replacement country for personal power and glory or something like that
Quoting Federal Reserve (@federalreserve): Video message from Federal Reserve Chair Jerome H. Powell: https://www.federalreserve.gov/newsevents/speech/powell20260111a.htm
Quote tweet, 03:28 UTC · 6 likes
So far only 5/15 people with >$1k donations have filled out the Google Form (below) or otherwise reached out about this!! Get ur stickers!!!
(if you can't tell they look right and circular after peeling off page)
Quoting Aaron Bergman 🔍 (@AaronBergman18): Stickers!!! If you donated at least $1k (could be across multiple donations, not too late 👀) please fill out the google form below so we can mail (at least) one to you!
Thread, continued, 03:28 UTC · 2 likes
(If you're seeing this more than like a couple days after posting you can still get your sticker but can you also DM/message me instead of just filling form)
Form: https://forms.gle/rRP7S4u2GkYJgfD68
Tweet, 03:48 UTC · 20 likes
Reply to @idavidrein, 05:18 UTC · 7 likes
david rein (@idavidrein): some people are now intentionally spreading poisoned data on the public internet to make LLMs dumber
Wait is the text nonsense or evil or something else?
Reply to @NathanpmYoung, 06:48 UTC · 72 likes
Nathan 🔎 (@NathanpmYoung): Had an MRI this morning (dw I'm fine) and she was like "we're gonna put this high contrast dye in"
me: "I don't want to hear about it, I'll faint"
her: "you want to consent without understanding?"
me: "absolutely"
Hope ur good but also MRIs are very cool
Reply to @Lib_Development, 09:16 UTC · 30 likes
(The tweet being replied to by @Lib_Development is no longer available.)
1) I think shrimp weirdos including myself are extremely disproportionately likely to be concerned about AI welfare so kinda reject the implicit premise - AI welfare is a huge deal but still extremely pre-paradigmatic
2) we have a bunch better sense of what is good and bad for some animals
3) consciousness/sentience really is what matters, and there are good reasons to think that at least some animals are sentient whereas there only weaker ones for AIs
4) i disagree that there’s a clear connection between compelling writing and moral status
5) At some level it’s just a matter of degree; accounting for uncertainty and such I do think arbitrary LLM instances are more important that some individual animals (tartigrade)
Reply to @idavidrein, 09:16 UTC · 23 likes
david rein (@idavidrein): https://RNSAFFN.com/poison2/
I believe they randomly show you a code snippet with a subtle bug/vulnerability inserted into it
That sounds actively possibly harmful lol
Reply to @IvanVendrov, 09:22 UTC · 24 likes
ivan (@IvanVendrov): truth : good : beauty ::
lesswrong: EA Forum : ???
Girls
Tuesday, January 13, 2026
Quote tweet, 03:20 UTC · 2 likes
Rings true, but of course this is not inherent to AI art and at some point we’ll have to worry about it being an “emotionally resonant aesthetic complexity” superstimulus instead sub-stimulus
Quoting Andy Masley (@AndyMasley): I suspect a lot of hatred of AI art comes from a correct hatred of this specific move
Reply to @deanwball, 03:51 UTC · 7 likes
Dean W. Ball (@deanwball): Word to the wise for those newly enamored of Claude models: every Claude 4.5 version, including Opus, is ~terrible at giving tax and legal advice. If you are using AI for tax planning and associated legal needs, you should use GPT 5.2 Pro or Gemini 3 Deep Think.
Thank you
Tweet, 05:51 UTC · 2 likes
When are we gonna get Claude Code but for Work but for Code
Quote tweet, 06:00 UTC · 2 likes
Ahhh this is the standard take I think but I disagree!
There are lots of potential benefits to doing what Litt describes *sometimes* - like conveying information while making it clear that your source is an LLM, or translating between people talking past each other
Quoting Daniel Litt (@littmath): IMO it should be considered quite rude in most contexts to post or send someone a wall of 100% AI-generated text. “Here, read this thing I didn’t care enough about to express myself.”
Tweet, 06:00 UTC · 1 like
Clearly it *can* be rude and excessive but “this argument makes a lot of sense and I endorse it as though it was my own” can be fine actually in many contexts
Thread, continued, 06:00 UTC · 2 likes
Tbc I think deceiving ppl about who wrote what is bad - I’m talking about being very up front about source
Reply to @cljack, 17:52 UTC · 5 likes
Charlotte Lee (@cljack): Hey I haven’t been to any bars lately!
(jk this clearly wasn’t me because I would never bring up data science before airplanes)
(quoting @vuse_lover: Shoutout to the very drunk girl who wanted to be my friend in the bathroom line at the bar and then followed me into the stall and talked to me about data science and famine)
Reply to @Benthamsbulldog, 17:53 UTC · 7 likes
Bentham's Bulldog🔸 (@Benthamsbulldog): Very excited to announce I will be joining @forethought_org as a visiting scholar. I think Forethought is producing hugely important work, and the people working there are great, so super excited about this.
Whaaaaaat congrats!!!
Quote tweet, 17:54 UTC · 15 likes
Simply write a gazillion words a day about the most important things in the world
Quoting Bentham's Bulldog🔸 (@Benthamsbulldog): Very excited to announce I will be joining @forethought_org as a visiting scholar. I think Forethought is producing hugely important work, and the people working there are great, so super excited about this.
Tweet, 17:58 UTC · 11 likes
What
Reply to @NoahTopper, 18:42 UTC · 3 likes
Quinoah 🔍⏸️ (@NoahTopper): Aaron works for the FDA arc
I could fix them fr
Tweet, 20:26 UTC · 27 likes
Twitter gave me $1,136 in 2025 for poasting
Tweet, 20:58 UTC · 15 likes
Pretty bad
Good luck to the people of Iran
Reply to @cremieuxrecueil, 21:03 UTC · 4 likes
Crémieux (@cremieuxrecueil): No, it's bad for their development!
We should make efforts to continue eliminating infectious disease so that people develop to their full potential.
The myth that getting sick makes you stronger has led to irreparable damage to generations of kids.
🔥
Tweet, 21:16 UTC · 15 likes
Just give Intuit shareholders some money to fuck off forever
Reply to @RafaRuizdeLira, 21:19 UTC · 3 likes
Rafael Ruiz ⏸️🔸 (@RafaRuizdeLira): @AaronBergman18 dafuq
High or low?
Quote tweet, 22:18 UTC · 3 likes
Absolutely unacceptable
Quoting Matt Bateman (@mbateman): @ouranometrian2 All I can think of is how Gena had a two word birth plan: “EPIDURAL ASAP” written as large and in as big letters as possible across every document and communicated loudly to everyone
First child: took 4 hours to get epidural
Second child: epidural not delivered before baby
Reply to @AStrasser116, 22:18 UTC · 1 like
Alex Strasser (@AStrasser116): @AaronBergman18 For real?!?!
No, I rounded up to the nearest dollar (but actually, yes)
Tweet, 22:31 UTC · 8 likes
Smoke alarms should really come with a “P(fire)” dashboard
No reason it should be a binary thing!
Quote tweet, 23:52 UTC · 10 likes
Thank you so much to everyone who donated
You all contributed over $100,000 (specifically, $102,451.61), which will do a perhaps-literally-unfathomable amount of good
Quoting Aaron Bergman 🔍 (@AaronBergman18): Someone donated $6500??? Like today??? Based??????????
Thank you mysterious person (feel more than free to identify urself)!!!!!
Wednesday, January 14, 2026
Quote tweet, 02:13 UTC · 4 likes
Something like this is quite fun and interesting to run up against
Makes IQ feel less mysterious
Quoting Justin Skycak (@justinskycak): @elocinationn Douglas Hofstadter wrote about his experience of running up against an “abstraction ceiling” in his own brain while pursuing a PhD in mathematics.
As Hofstadter describes, the abstraction ceiling is not a “hard” threshold, a level at which one is suddenly incapable of learning math, but rather a “soft” threshold, a level at which the amount of time and effort required to learn math begins to skyrocket until learning more advanced math is effectively no longer a productive use of one’s time. That level is different for everyone.
For Hofstadter, it was graduate-level math; for another randomly selected person, it might be earlier or later (but almost certainly earlier).
Quote tweet, 02:23 UTC · 65 likes
Healthy society needs to maintain the distinction between whether this guy should lose his job as a car salesman (no) and whether he should lose his job as someone representing the United States (yes)
Quoting Aaron Reichlin-Melnick (@ReichlinMelnick): There you have it; the Trump admin allows ICE lawyer James Rodden to stay on the job. His personal social media account praised Hitler, supported white nationalism, and said that America started going downhill the day that Lincoln was elected.
Reply to @imitationlearn, 02:33 UTC · 2 likes
imit (@imitationlearn): @AaronBergman18 because of the impact of each job on others?
Yes with some explanation - because the general policy of going after *anyone* for having bad takes, even evil takes, is going to result in a very repressed society with less effective free discourse
And
(1) this isn’t fully theorized but like prosecutors seem somewhat legitimately/by consensus within the domain of politics - like they’re legitimately hired and fired for political reasons via political pressures/processes
(2) more directly relevant to the first bit: the consequences of going after the 0.1% most powerful people for having bad, evil takes is just a lot less costly
we’re not saying this guy needs to live in the gutter, just that he needs to find a different job. Sometimes we’ll be wrong and accidentally boot someone from a position of power for bad reasons and that sucks but it’s not the end of the world because the consequences for them aren’t that bad (find another job) so the disincentive effect on free speech isn’t that bad
Once you’re saying that car salesmen need to lose their job, you’re effectively saying that such a person cant have any job at all (or else what is the distinguishing principle?)
Quote tweet, 06:19 UTC · 29 likes
Watch this space
Quoting alli (@sonofalli): claude code can’t fix your lack of bitches
Quote tweet, 19:01 UTC · 15 likes
Can we not
Quoting Polymarket (@Polymarket): BREAKING: U.S. Defense Secretary says Grok will be integrated into the Pentagon’s internal networks.
Tweet, 19:56 UTC · 4 likes
Recording video meetings + cheap easy transcription + AI is such a life saver
If you’re raw dogging video calls (that are meant to be productive/work related) you’re missing out and probably enduring pointless stress
Thread, continued, 19:56 UTC · 4 likes
Hot take: with consent obviously, recording irl for personal use (not like sharing freely on the internet) should be more socially acceptable bc no reason virtual meetings need to have this advantage
Reply to @AndyMasley, 20:34 UTC · 3 likes
Andy Masley (@AndyMasley): A lot of people very confidently claim that no one's asking for new data centers, but stuff like this is reflective of large demand for way more compute.
(quoting @theo: Claude has had elevated error rates for 4 hours straight now 🙃)
Lol I don’t think I’ve seen anyone say this or something similar
Not saying you’re wrong, just reflective of different info environments
Quote tweet, 20:35 UTC · 6 likes
I really don’t envy whoever’s problem this is
Quoting Theo - t3.gg (@theo): Claude has had elevated error rates for 4 hours straight now 🙃
Thursday, January 15, 2026
Reply to @RosieCampbell, 00:51 UTC · 43 likes
Rosie Campbell (@RosieCampbell): I don't really understand what "trust, but verify" is supposed to convey
Be cooperative and don’t defect unless and until you have reasonably strong evidence that they defected (by lying) first
Reply to @RosieCampbell, 01:02 UTC · 9 likes
Rosie Campbell (@RosieCampbell): Hmm I feel like the word "trust" is pretty misleading then
I also have no idea what I’m talking about
Quote tweet, 01:33 UTC · 6 likes
Have they tried giving Trump SSRIs
Quoting Aaron Bergman 🔍 (@AaronBergman18): Have they tried giving the pitbulls SSRIs
Thread, continued, 01:34 UTC · 5 likes
Maybe he’d just pop them if prescribed and chill out 15% or something
Astronomically cost effective intervention if ur the White House doctor
Tweet, 01:48 UTC · 7 likes
Has anyone tried an RCT with every non-contraindicated antidepressant
Apparently Venlafaxine + mirtazapine + Wellbutrin + abilify + lithium + liothyronine + buspirone + lamotrigine + ketamine + modafinil + testosterone+deplin would probably not cause any sort of crisis/emergency
Tweet, 01:50 UTC · 6 likes
Unlocked “warning message on Max” achievement
$100 version tho (if I need the higher one I’ll get it)
Tweet, 02:20 UTC · 3 likes
Massacres in Iran
Wildcard is Netenyahu, right? I’m honestly not sure what I’m hoping for the IDF to do/choose
Thread, continued, 02:20 UTC · 2 likes
https://www.cnn.com/world/live-news/iran-protests-trump-01-14-26
Tweet, 04:31 UTC · 9 likes
My dad just sent a Claude chat link to the family group chat 🥺
(gif, plays on X)
Thread, continued, 04:33 UTC · 4 likes
Tbc I don’t actually think you should build your identity around preferred LLM
Reply to @gptbrooke, 05:49 UTC · 4 likes
brooke bowman (@gptbrooke): The girlies are plotting, smth abt strategically dating misaligned alignment researchers
“I could align him”
Moral hazard
Tweet, 20:17 UTC · 2 likes
Must feel so good to be a retiree and tell young hooligans on the street (such as myself) that they’re not dressed for the weather
Friday, January 16, 2026
Reply to @generativist, 06:08 UTC · 2 likes
johnny, INFINITE FUN ENTHUSIAST (@generativist): most of the things people are building with claude are like .4-1.05x better than something they already exists. to the human in the exhilarating loop, there is a bonus multiplier for “mine.” which isn’t a criticism it’s fun and custom is great!
but,
I don’t think this is true, “mine” is not only psychology - it means designed for a specific purpose and there’s often huge utility in that
It’s also just really good
Reply to @GarrettPetersen, 07:32 UTC · 2 likes
Dr. Dad, PhD 🔄🔼◀️🔽▶️ (@GarrettPetersen): Is there a centrist faction that's pro-fap, anti-porn? Like a "use your imagination" faction?
This is broadly what I endorse
Tweet, 21:04 UTC · 8 likes
I know I’m supposed to dislike chatGPT ads but as long as it’s clearly labelled as such, this seems good
I love targeted advertising and want more of it. Plz make my life better in exchange for money
Reply to @mattyglesias, 21:09 UTC · 7 likes
(The tweet being replied to by @mattyglesias is no longer available.)
Yeah
At this point it’ lindy
Saturday, January 17, 2026
Quote tweet, 06:04 UTC · 35 likes
Lol maybe you find it annoying, but putting clearly labeled ads on a free service is not a moral atrocity
Quoting Aashish Reddy (@_AashishReddy): Getting in early to say that this seems completely fine
Tweet, 07:19 UTC · 22 likes
Girls be like “I’m starving” and then eat one small taco and two french fries
Sunday, January 18, 2026
Quote tweet, 03:18 UTC · 11 likes
I’m concerned that many EAs really believe in the “great karmic ledger of Utilons” as some sort of genuine metaphysical object or at least adequate model of reality
Quoting roon (@tszzl): Claude runs an indoor shrimp farm in the sub-basement of Anthropic Headquarters in the financial district of San Francisco. Claude has helped the shrimp achieve and maintain states of ecstatic bliss, primitive circuitry of what might be called the first jhana in higher primates. Its intention with this project was to create so much positive shrimp valence that all the sins of wild shrimp suffering are cancelled out in the great karmic ledger of Utilons, without having to stake a position on the conversion rate of shrimp suffering to human suffering. The shrimp swim in what is called the “Pool of Sacred Tears”. The true origin of the naming is unknown, but it is believed that when Dario Amodei and Amanda Askell first discovered what Claude had done, they wept tears of sacred joy readily and for hours.
Thread, continued, 03:18 UTC · 9 likes
The only sense in which this exists is via universal comparability of world states which is forced on us by the need to choose between world states
There’s never a sense in which the original underlying suffering becomes metaphysically “neutralized”
Reply to @Turbinetraveler, 13:28 UTC · 2 likes
Turbine Traveller (@Turbinetraveler): This remains one of the hardest landings ever recorded on an Airbus A380.
The 2017 Emirates Airbus A380 crosswind landing at Düsseldorf (EDDL) during Storm Xavier.
📹: Cargospotter
🥵
Monday, January 19, 2026
Quote tweet, 04:43 UTC · 26 likes
Fun game: what’s closest to the one true monocause?
Scarcity? Incentives? Thermodynamics?
“Evolution” is a tempting one tbh - a lot does depend on “that which self-replicates more self-replicates more”
Quoting Rob Wiblin (@robertwiblin): The single thing most likely to make me tune someone out is realising their conclusions are all driven by one big ideology, while the specifics of individual cases barely register.
Common examples:
1. The root cause is capitalism
2. Government is to blame
3. Something something power imbalances
4. Decentralisation always good, centralisation always bad
5. Woke or anti-woke as lens for everything
6. "Elites are bad / corrupt / liars"
7. Everything's a coordination problem / Moloch
8. Cynicism about motives, e.g. they're just status-seeking
9. Seeing everything as about cultural decay, screens, algorithms, misinformation, etcAll of these ideas are reasonable.
Sometimes capitalism or government is the problem, sometimes 'elites' are self-serving, sometimes people are just status-seeking, etc etc.
But people with a strong reflect to apply one framework to every situation are wrong much more often, boring to talk to, and have low information density output.
(I think this monomaniacal-ideology thing is actually fairly rare in the broader world, like <10% of people think this way, but it's very overrepresented on Twitter.)
Reply to @captgouda24, 04:46 UTC · 46 likes
Nicholas Decker (@captgouda24): If ICE can shoot anyone they want, and not be punished for it, what recourse do we have but civil war?
https://nicholasdecker.substack.com/p/when-will-they-kill-us
I mean one answer is “we don’t do recourse, we just let injustice and lawlessness exist unpunished and hope things go back to normal and they probably will”
(I’ll add that “probably” shouldn’t be very reassuring)
Quote tweet, 06:01 UTC · 3 likes
They closed the whole DC <-> Arlington Key Bridge on Friday night (probably past midnight into Saturday) and there’s one official tweet and a single 2 paragraph news story
I’m not complaining about the closure, I just want to know what happened 👀
Quoting DC Fire and EMS Department (@dcfireems): Update water search Potomac River beneath Key Bridge. 1 victim recovered by a diver and enroute to EMS at a nearby land location. #DCsBravest
Tweet, 06:01 UTC · 1 like
Thread, continued, 06:01 UTC · 2 likes
There’s a “closure” double entendre joke to be made by someone smarter than me
Reply to @Reagan_Airport, 17:24 UTC · 4 likes
Reagan Airport (@Reagan_Airport): Remembering the life and legacy of Rev. Dr. Martin Luther King, Jr. today.
So true Reagan Airport
Tuesday, January 20, 2026
Reply to @Aella_Girl, 01:19 UTC · 27 likes
Aella (@Aella_Girl): fwiw in my survey data, height and # of sexual partners is clearly correlated. Each additional inch of height corresponds to roughly 1/5th a sexual partner for males.
image attached
(quoting @MadelaineLucyH: I don’t actually know any women who haven’t dated or had a crush on someone shorter than them. Especially models and actresses. Sure height is a general plus but that’s not really what women actually fall for IRL.)
Interesting but 0.2 per inch is way lower than I’d have guessed
5’7” -> 6’0” ≈ delta of 1 ≈ delta of 20%? I think I would have guessed 100% ie 5 inches of height doubles number of partners
The main takeaway here is “effect way weaker than you’d guess” IMO, even though “clear correlation” is also true
Tweet, 06:06 UTC · 7 likes
Idk how many people are broadly optimistic about the future (in a predictive sense not a dispositional sense)
Thread, continued, 06:06 UTC · 12 likes
A genuinely good future - and by that I mean one in which you’d want to personally live every life - seems possible but extremely precarious, broadly like one-shotting a plane landing with no training
Thread, continued, 06:06 UTC · 8 likes
The slice of worlds in which we solve wild animal suffering and prevent suffering (digital or bio) from being spread throughout the universe and also solve AI alignment well enough and also understand but don’t abuse digital sentience and also get good governance of unfathomably powerful tech is just…pretty thin
Thread, continued, 06:06 UTC · 5 likes
The sphere that is relatively wealthy people in the modern world who only rarely encounter major suffering is just like a tiny blip on an unfathomably infinite ocean of uncaring forces that mostly coalesce in suffering or nothing at all
Idk man
Quote tweet, 13:17 UTC · 3 likes
Big tent
Quoting Cato Institute (@CatoInstitute): Trump’s “Great Healthcare Plan” would move more safe drugs over the counter—lowering prices, expanding access, and, most importantly, returning control of health care to patients, argues Cato’s @dr4liberty.
Tweet, 13:24 UTC · 9 likes
Gonna mog everyone at the DMV today with my conscientiously printed and folder-organized documents
Quote tweet, 13:30 UTC · 3 likes
Many such cases
Those without a conceptually clear sense of ethics will reach for aesthetics and personal quality signaling as a stand-in
Quoting Holly ⏸️ Elmore (@ilex_ulmus): Crazy that so many people think traveling for your own enjoyment is this admirable thing, maybe even good for the world. Boomers especially. I’m all for fun, but recreational travel is just consumption.
Reply to @ilex_ulmus, 13:32 UTC · 30 likes
Holly ⏸️ Elmore (@ilex_ulmus): Crazy that so many people think traveling for your own enjoyment is this admirable thing, maybe even good for the world. Boomers especially. I’m all for fun, but recreational travel is just consumption.
Oof yes, at my grandfather’s funeral a bunch of people talked about his love for travel and like…all well and good but not the basis on which we should celebrate or praise a person!
Tweet, 17:43 UTC · 4 likes
Everyone hates AI writing but you’re gonna miss it when it’s gone
Reply to @AaronBergman18, 17:53 UTC · 6 likes
(The tweet being replied to by @AaronBergman18 is no longer available.)
Pastoral/PoW camp norms around the value of physical goods, both in absolute terms and in terms of time, don’t actually make sense in wealthy US suburbs in 2026
Reply, 18:43 UTC
Yeah I don’t blame/mind anyone mentioning it tbc
Tweet, 22:35 UTC · 2 likes
I kinda think it would be good for US to control Greenland but needless to say only via some legitimate political process involving the people of Greenland
Tweet, 22:39 UTC · 5 likes
Claude Cowork doesn’t make a lot of sense, it’s just Claude code but more confusing and less powerful
Wednesday, January 21, 2026
Reply to @shinboson, 00:49 UTC · 7 likes
𝞍 Shin Megami Boson 𝞍 (@shinboson): slavery was cartoonishly evil
Who are these people jfc
Tweet, 00:54 UTC · 10 likes
I think I’m finally on the downward sloping part of the lumens available to turn on vs quality of life curve
No more lamps
Quote tweet, 05:51 UTC · 13 likes
Bro just train with together and deploy with axolotl, you can literally finetune on firework and serve serverless on unsloth, the model's on hf bro just tap runpod and host on lambda, just use the vast mcp ontop of groq with modal bro it's easy
Quoting Aaron 🦇 (@aaronfraggle): It's on Tubu. It's literally on Heebee. It's on Poodee with ads. It's literally on Dippy. You can probably find it on Weeno. Dude it's on Gumpy. It's a Pheebo original. It's on Poob. You can watch it on Poob. You can go to Poob and watch it. Log onto Poob right now.
Tweet, 06:15 UTC · 26 likes
We gotta not let bygones be bygones if we make it past Trump
Put all the ICE officers who broke the law on trial
Thursday, January 22, 2026
Quote tweet, 01:48 UTC · 8 likes
This is substantively good and cool but isn’t “constitution” just…not the right word?
It’s *almost* the right word, but “constitution” connotes definition, creation and sovereignty interaction
“Charter” might be better, though still not perfect
Quoting Anthropic (@AnthropicAI): We’re publishing a new constitution for Claude.
The constitution is a detailed description of our vision for Claude’s behavior and values. It’s written primarily for Claude, and used directly in our training process.
https://www.anthropic.com/news/claude-new-constitution
Thread, continued, 01:50 UTC · 2 likes
Unless Anthropic is tracking what I type in other apps and feeding that to the model somehow, Claude and I landed on “soul spec” independently
Quote tweet, 04:24 UTC · 59 likes
Seems bad and illegal and like a bunch of people need to go to prison for casually circumventing the constitution
Quoting The Associated Press (@AP): BREAKING: Immigration officers are asserting sweeping power to enter homes without a judge’s warrant, a memo obtained by AP says. https://apnews.com/article/ice-arrests-warrants-minneapolis-trump-00d0ab0338e82341fd91b160758aeb2d?utm_campaign=trueAnthem%3A+New+Content+(Feed)&utm_medium=trueAnthem&utm_source=twitter
Reply to @cljack, 05:06 UTC · 2 likes
Charlotte Lee (@cljack): Was literally just pondering how long it would take me to Claude Code a browser extension
https://drive.google.com/file/d/1DtmkMudSsTRF0aycL_2oEcuWv1C0xbPD/view?usp=sharing
Probably not actually helpful, couldn't resist
(vibe instructions for whomever: go to chrome://extensions/ -> turn on devmode -> unzip -> load unpacked -> select folder -> profit)
Quote tweet, 05:23 UTC · 33 likes
Type of shit that has a 1/10k chance of just making the difference
Quoting Lewis Bollard (@Lewis_Bollard): Cool to see Claude's new constitution call out the "welfare of animals and of all sentient beings" as a value to consider.
I hope other AI labs follow suit in their guiding docs. Caring about animals is core to our humanity -- and should be core to the AIs we build too.
Reply to @wedrifid, 06:00 UTC · 2 likes
Cameron Taylor (@wedrifid): I suggest it would be good for the USA if the USA was controlled by Greenland via some legitimate political process.
Once the USA re-establishes a legitimate political process they could be left alone manage themselves.
(quoting @AaronBergman18: I kinda think it would be good for US to control Greenland but needless to say only via some legitimate political process involving the people of Greenland)
reasonable
Reply to @GjMcGowan, 06:05 UTC · 1 like
George McGowan (@GjMcGowan): Damn that’s crazy. I guess tell HR?
image attached
(quoting @hamptonism: Anthropic CEO: "Software Engineering Will Be Automatable in 12 Months". yea, we’re cooked.)
Not sure “one guy at the top of the pyramid pulls the puppet strings” is much of an escape hatch
Tweet, 06:20 UTC · 13 likes
Can someone explain what’s going on with Will Stancil
Tweet, 16:36 UTC · 5 likes
This is so exciting ❄️❄️❄️❄️❄️🌨️⛄️🌨️⛄️🌨️⛄️🌨️⛄️🌨️⛄️🌨️⛄️🌨️🏔️⛷️🌨️⛄️🌨️🌨️🌨️⛄️🎿🏔️🎿🌨️⛄️🌨️🌨️☃️🏂🎿🍧🌨️🏔️❄️❄️❄️⛄️⛄️🏂🎿🏂🍧🍧🍧❄️🎿🎿⛄️☃️🌨️🌨️
Friday, January 23, 2026
Tweet, 02:32 UTC · 9 likes
I think we in the US should just start using British accents. It’s aesthetically superior, hands down
Reply to @GarettJones, 05:10 UTC · 2 likes
Garett Jones (@GarettJones): Also the optimal immigration policy:
(quoting @GarettJones: @MosesSternstein First rule of human resources: Hire the lucky)
Lol idk how serious this is but it’s not (in general) true
Tweet, 06:42 UTC · 6 likes
I have vibecoded a probabilistic storm forecast thing (DC only for now): https://snow2026.up.railway.app
Inshallah we get tail risk >3 ft ❄️☃️⛄️❄️🌨️🏔️⛷️🎿⛄️🌨️☃️⛄️⛄️❄️❄️☃️⛄️⛄️☃️❄️🍧🏂🍧🏂🌨️🌨️🌨️⛄️☃️
Thread, continued, 06:42 UTC · 5 likes
Don’t miss out on this fun visualization
Reply to @Yixiong_Hao, 08:15 UTC · 2 likes
Yixiong Hao (@Yixiong_Hao): inner misalignment in humans
'working hard and getting good grades' was a useful proxy and correlated with getting what we want but many mistakenly generalize this as the optimization target. this is maybe no longer true!
(quoting @FT: University careers services departments — such as at King’s College London — are offering students AI tools that give personalised recommendations on how to tailor CVs and cover letters so that first-time jobseekers can game the bots reading them https://ft.trib.al/FFRxkhE)
Not exactly wrong but also probably the wrong frame of analysis
If the problem is that people can’t buy things like food and shelter because they don’t have money because they don’t have a job because there are fewer jobs than people because of AI, it’s not especially productive to observe that “they optimized for the wrong thing” because the same issue would probably still exist in society even if people aimed squarely at “obtaining food and shelter” or even “making money” as the goal
Maybe you’d shift the distribution across the population a tiny bit towards complimenting rather than substituting for AI labor but that seems unlikely to nontrivially change the unemployment rate for very long
Reply to @deepfates, 12:28 UTC · 2 likes
deepfates.delve.town (@deepfates): As far as I can tell you only need to implement these five LLM providers:
- OpenAI
- Anthropic
- OpenRouter
- LM Studioeverything else is superfluous. right?
http://together.ai offers the very specific product that is “serverless lora-adapter based finetuned model inference” which enables fast/hot starts for finetunes of a couple large open weight models whereas normally you’d need to either pay 5 figs/mo to have the model constantly loaded on dedicated GPUs or wait 20+ minutes to download and load the model whenever you wanted to use it
Quote tweet, 18:01 UTC · 225 likes
Eric did nothing wrong
Quoting insane dms (@insanedms):
Reply to @willmacaskill, 19:03 UTC · 9 likes
William MacAskill (@willmacaskill): I think that the "industrial explosion" is about as important an idea as the "intelligence explosion", but gets far less attention.
I discuss this idea with @TomDavidsonX here! We cover:
What is the industrial explosion?
Why the case for recursive self-improvement is stronger for physical industry than for software
How fast the physical economy could grow, the case for weekly doubling times, and limits from natural resources
Three phases of the industrial explosion: AI-directed human labour → autonomous replicators → atomically precise manufacturing
Why authoritarian regimes might have a structural advantage in the industrial explosion — and whether they could lose the race to AGI, but win the race to industrial dominance
Could a leading country outgrow the entire world to achieve decisive dominance?
Why doubling your rival's GDP could mean 30 years of tech advantage — enough for military dominance
Why does the industrial explosion get ~1% of the attention of the intelligence explosion?
🐢
Reply to @absurdlymax, 19:56 UTC · 10 likes
Max Alexander (@absurdlymax): Well he did hit someone with his car
🤫
Reply to @willmacaskill, 20:22 UTC · 2 likes
William MacAskill (@willmacaskill): I think that the "industrial explosion" is about as important an idea as the "intelligence explosion", but gets far less attention.
I discuss this idea with @TomDavidsonX here! We cover:
What is the industrial explosion?
Why the case for recursive self-improvement is stronger for physical industry than for software
How fast the physical economy could grow, the case for weekly doubling times, and limits from natural resources
Three phases of the industrial explosion: AI-directed human labour → autonomous replicators → atomically precise manufacturing
Why authoritarian regimes might have a structural advantage in the industrial explosion — and whether they could lose the race to AGI, but win the race to industrial dominance
Could a leading country outgrow the entire world to achieve decisive dominance?
Why doubling your rival's GDP could mean 30 years of tech advantage — enough for military dominance
Why does the industrial explosion get ~1% of the attention of the intelligence explosion?
For ppls convenience:
Apple pods: https://podcasts.apple.com/us/podcast/forecast/id1796860911?i=1000746199424
Spotify: https://open.spotify.com/episode/69yDWBvfupyvaJNmQmB4BH?si=QFGLj8V-Tk6uxEYHNHatYg
Tweet, 20:35 UTC · 44 likes
If LLM sentience is real it’s very weird, eg the possibility that each token generation is its own little flash of qualia or something, entirely subjectively disconnected from the others
(Idk if this specific version of weirdness is actually plausible)
Tweet, 21:14 UTC · 33 likes
Anthropic going slower would be more legibly high integrity and also worse all things considered IMO
Reply to @GtownVoice, 21:51 UTC · 3 likes
The Georgetown Voice (@GtownVoice): Despite student protests, Georgetown Law has confirmed they will not remove ICE and DHS as employers at a virtual job fair set to take place on Friday, Jan. 23. Read more:
https://georgetownvoice.com/2026/01/22/georgetown-law-refuses-to-remove-ice-from-upcoming-job-fair-despite-student-petition-employers-pull-out-in-response/
Sharing what I just wrote on IG
I’ll add that this dynamic is ~unique to job fairs specifically. Like GU shouldn’t assist ICE in other arbitrary ways
Reply to @timfduffy, 22:09 UTC · 3 likes
Tim Duffy (@timfduffy): If that disconnection also existed for you, would you notice? Seems to me that each disconnected state would feel like it followed directly from the state immediately prior and you wouldn't feel the disconnect.
Yeah good q, I don’t think I necessarily would but it’s conceivable and possible to have the sense of continuity or not to have that sense; having that sense is just like a qualia itself or an aspect of one if you prefer
Quote tweet, 22:51 UTC · 2 likes
The jealousy
Quoting mary (@howdymary): my friend is dating a guy who works at one of the frontier research labs in sf and his sole job is to think and come up with a Big Idea every six months
so his working day consists primarily of long walks spent pondering
Reply to @AndyMasley, 23:12 UTC · 7 likes
Andy Masley (@AndyMasley): Guy with a big bio following seems to get furious at me because he's assuming I'm talking here about an actual lab leak theory and blocks me before I can explain the metaphor
On the merits though this seems silly
Like yeah some people will have bad takes about lab leak/social construction and that is a very modest price to pay for free discourse and object level argumentation
Tweet, 23:20 UTC · 9 likes
Society and modernity is great because for almost all of human history there was no “run back inside and take a hot shower with access to unlimited clean towels” option in a snow storm, you were just fucking cold and miserable and maybe got frostbite and hypothermia and died
Thread, continued, 23:20 UTC · 2 likes
And you probably daydreamed about your tribe’s version of heaven where the sun always shone and there way a big bonfire always raging and
Well here we are
Thread, continued, 23:20 UTC · 2 likes
Consider going winter camping one time to appreciate how long say 36-72 hours in the cold and snow or rain really is
Reply to @vivian39_, 23:21 UTC · 4 likes
vivian (@vivian39_): what do you call this part of Arlington?
Home :)
Saturday, January 24, 2026
Reply to @BEBischof, 03:54 UTC · 11 likes
Bryan Bischof fka Dr. Donut (@BEBischof): This is a beautifully succinct demonstration of our current state of doing things with LLMs.
(Cc. @moultano)
It’s a succinct demonstration of the gap between the systems people use and what the models can actually do
Grok on here and http://Claude.ai skimp on context to save tokens, but once you’re paying per token via the api the providers are extremely happy to let their systems rip with 256k-1M context and very rare hallucination
Reply to @cammakingminds, 03:59 UTC · 13 likes
Lucid™ (@cammakingminds): "AI safety" 😭 < "AI safety" 💪
Lol this not an accurate characterization of AI welfarists
Quote tweet, 11:59 UTC · 6 likes
Jesus the cynicism
Quoting Shakeel (@ShakeelHashim): One of Leading the Future's PACs is running an attack ad against @AlexBores that criticizes him for working for Palantir (fair, tbh), saying that he enabled ICE and its deportations.
What the ad doesn't mention? That Leading the Future is funded by Joe Lonsdale ... the founder of Palantir, and a big Trump donor.
Quote tweet, 16:58 UTC · 9 likes
Me and who
Quoting Dimorphist (@dimorphist):
Quote tweet, 18:03 UTC · 28 likes
Trump and his thugs have taken the 2020 progressive picture of indiscriminate police brutality and murder and made it the modus operandi
Quoting Adam Cochran (adamscochran.eth) (@adamscochran): 1/12
Frame by frame breakdown of this MURDER shows ICE is lying!
-ICE had ALREADY disarmed the man before shooting.
-The man NEVER drew his gun.
At the 0:06 second mark we see the man’s hands on the ground covered in large gloves.
His hands are empty.
Quote tweet, 18:06 UTC · 9 likes
Everyone 🤝 everyone
Concerned about ice
Quoting Matthew Cappucci (@MatthewCappucci): DMV gang – I'm increasingly concerned about the icing. A lot of this icing (freezing rain) will occur Sunday midmorning into early afternoon after the snow has fallen. That means we'll add a "crust" onto the ice that will make if very difficult to plow, shovel or move.
Still looking like a foot plus of snow for northwest Maryland and the Potomac Highlands, the Panhandle and the I-68 corridor. In D.C. proper, 8 or 9 inches. Less snow to the south, but more icing.
It also now looks like some minor to moderate glazing could stretch up the I-95 corridor all the way into New Jersey.
Tweet, 22:02 UTC · 8 likes
Every few months I am reminded that coworking absolutely does not work for me whatsoever and I cannot focus at all with other people around who I’m supposed to be attuned to in the slightest
Thread, continued, 22:02 UTC · 3 likes
(Actually talking to people when that is the sole thing I’m trying to do is fine at least for a couple hours a day)
Thread, continued, 22:03 UTC · 3 likes
I think this relates to my inability to understand non-trivial board game rules in a social setting
Feels the same psychologically/cognitively
Quote tweet, 23:41 UTC · 174 likes
I don’t want this guy owning a galaxy
Quoting Lisan al Gaib (@scaling01): don't forget who supports all of this
Greggy gave $25 million to support Trump
Tweet, 23:49 UTC · 220 likes
How much species protection stuff is just the misconception that the stakes are like “all elephants” rather than “the brown nosed red tailed fluffy-eared Albuquerque apple-mouse”, one of 56,112 subtypes of apple-mouse
Quote tweet, 23:59 UTC · 30 likes
Guys all your “caffeinated X” dreams can come true
You can buy caffeine pills on Amazon and mix them into anything you like
Quoting Someone:
Sunday, January 25, 2026
Reply, 00:02 UTC · 8 likes
Ik you’re probably responding to Matt but like the natural world is horrible and not generally worth protecting
Very pretty for humans though!
Reply, 00:14 UTC · 2 likes
I mean maybe it’s those things but I think it’s also true
I wouldn’t want to live the life an arbitrary wild animal or arbitrary minute of wild animal experience or something and it’s not a close call
Reply to @selentelechia, 02:35 UTC · 7 likes
(The tweet being replied to by @selentelechia is no longer available.)
Fr probably do NOT get powder because pills are just pre-portioned powder (open up the pill cap) that prevent you from killing yourself and/or having a horrible time because you accidentally took 1g instead of 100mg
If you want to get powder to save money, don't, because the relative price delta doesn't matter in absolute terms (probably <$5 year in savings)
Reply to @flybottlemist, 02:36 UTC · 2 likes
(The tweet being replied to by @flybottlemist is no longer available.)
Ok good luck dude
Reply to his own earlier tweet, 02:37 UTC · 5 likes
Aaron Bergman, January 24, 2026: Guys all your “caffeinated X” dreams can come true You can buy caffeine pills on Amazon and mix them into anything you like
I use paraxanthine now bc it's a bit smoother, in the name of full disclosure
Quote tweet, 02:41 UTC · 4 likes
New Pigeon Hour just dropped in which @robi_rahman and I discuss some of his takes and more
https://www.aaronbergman.net/p/15-robi-and-aaron-tackle-donation
Quoting Aaron Bergman 🔍 (@AaronBergman18): Ok it has been a little while but goddammit we are doing the things
Presenting: (Vegan) Hot Ones ft. @robi_rahman and @absurdlymax
Spicy vegan nuggets, even spicier takes 🌶️🥵
(YouTube link below because Twitter doesn't seem to l our video file)
Reply to his own earlier tweet, 02:46 UTC · 18 likes
Aaron Bergman, January 24, 2026: How much species protection stuff is just the misconception that the stakes are like “all elephants” rather than “the brown nosed red tailed fluffy-eared…
Let it be known that I do care about mice of all varieties and think we should be much kinder to them as individuals
Reply to @eigenrobot, 02:47 UTC · 6 likes
eigenrobot (@eigenrobot): my new opinion about the protests is that minneapolis needs to be leveled so we should keep the street battles going as long and as intensely as we can to achieve this outcome.
surviving minneapolitans can be resettled ofc. i dont even care where as long as its out of state
????????
Tweet, 02:55 UTC · 2 likes
I’d really like to lifestyle inflate my way into an elevator (or just ground floor home entrance)
Reply to @spencerschiff_, 05:37 UTC · 6 likes
Spencer Schiff (@spencerschiff_): Why not? There will likely be plenty of galaxies to go around
Seems actively bad for him to have power/resources
Reply to @spencerschiff_, 05:42 UTC · 3 likes
Spencer Schiff (@spencerschiff_): My point is that one single galaxy might not be a huge amount of resources
I’m no physicist but I think in an absolute sense it is - like even if it doesn’t compete with anyone else’s stuff. I’m not saying he’s a sadist, but surely you can understand why wouldn’t want a sadist to control even a single acre on earth?
Tweet, 17:49 UTC · 5 likes
IT’S SNOWING
☃️❄️🌨️☃️❄️🌨️☃️❄️🌨️☃️🌨️❄️🌨️☃️❄️🌨️☃️❄️🌨️☃️❄️🌨️☃️🌨️🌨️❄️🌨️❄️🌨️❄️☃️🌨️❄️☃️🌨️❄️☃️🌨️❄️🌨️☃️🌨️❄️🌨️☃️🌨️❄️🌨️☃️☃️🌨️🌨️❄️🌨️☃️🌨️☃️🌨️☃️🌨️❄️🌨️🌨️❄️🌨️⛄️🌨️🌨️❄️🌨️❄️🌨️❄️🌨️❄️☃️☃️🌨️⛄️🌨️🌨️🌨️☃️☃️☃️🌨️🌨️⛄️🌨️⛄️❄️☃️🌨️☃️🌨️🌨️☃️🌨️☃️🌨️🌨️🌨️☃️🌨️☃️⛄️🌨️☃️🌨️🌨️❄️🌨️🌨️🌨️🌨️❄️☃️☃️⛄️🌨️⛄️🌨️⛄️❄️☃️🌨️☃️🌨️❄️🌨️❄️🌨️☃️
Tweet, 20:24 UTC · 8 likes
Grocery bags between socks and shoes work surprisingly well
Tweet, 20:27 UTC · 5 likes
Sand and snow are opposites
Reply to @GjMcGowan, 20:32 UTC · 3 likes
George McGowan (@GjMcGowan): Please under no circumstances let a woman see you like this
I’m telling you, panties hit the floor
Monday, January 26, 2026
Reply to @RokoMijic, 04:55 UTC · 17 likes
Roko 🐉 (@RokoMijic): I mean it probably wasn't for no reason. The casualty in this case was carrying a loaded firearm and fighting with police.
People have speculated that his loaded firearm may have accidentally discharged and the officers, hearing a gunshot, returned fire in the heat of the moment.
This person was not there as an ICU nurse, he was there as part of a hands on disruptive/violent protest movement that actively seeks out physical confrontations with law enforcement.
They post material on how to "safely" engage in physical fights with police officers who are trying to arrest criminals
I don't know how deeply you have looked into it but I suspect that you're not being shown a balanced picture.
A couple points:
1) the loaded firearm thing is a pretty discrete empirical question, plausibly you’re right about that
2) it’s relevant that a massive amount of American politics, society, history, and culture is fighting for or based on the non-obvious view that you should be able to have or carry a gun in self defense and/or to dissuade or stop government tyranny
Institutions have to be built around those facts of the matter: for better or worse, carrying a gun in situations like what I understand these to be is obviously bayesian evidence of danger and ill-intent but also not very strong evidence
This is to say that in some relevant sense America and specifically Minnesota has decided that merely carrying a loaded gun in public is the kind of thing you can do. That matters a lot. Either it’s legal or it isn’t, and the man was well within his rights to carry that gun (even though I think this is probably bad policy)
3) I don’t know what this guy learned or intended or who organized what, but polarizing public opinion against the enforcers of an unjust order is an actually-good theory of change for changing that order; “how to safely fight police” is what that theory of change looks like.
Would you have blamed a black segregation protester for studying up on what to do if and when white mobs attacked him for sitting in at a whites-only restaurant? I mean honestly maybe you would, but I wouldn’t and I don’t think you should. That’s what earnest preparation for a legit, ethical and just theory of change looks like
Reply to @NoVA_Campaigns, 05:03 UTC · 10 likes
NOVA Campaigns (@NoVA_Campaigns): There is 0% solar energy powering Virginia today during the storm. Wind is 4.5%. Coal [corrected to] 23%. Gas & nuclear 68%
Stop the reckless waste of billions in “green” energy and keep Virginians warm and powered by REAL energy that saves lives
(quoting @ChristieFERC: At 5 am today, in the midst of the winter storm, the generation mix in PJM was GAS: 40%, NUKE 28%, COAL 28%, WIND 4.5%, SOLAR 0. So dispatchables: 91%. The claim that VA or any PJM state can run a modern grid without dispatchable gen is reckless. Facts are stubborn things.)
“X isn’t currently the case” is not an especially strong normative argument against X
Tweet, 05:11 UTC · 10 likes
Idk who needs to hear this but they make autism-friendly belts that (1) allow much finer-grained adjustments in size and (2) are quick, easy, and socially acceptable to change size while wearing
Minor lifesaver, definitely worth the $15
“Ratchet belt”
Thread, continued, 05:11 UTC · 2 likes
I got this one, now selling for $10
I’m sure others are just as good
They’re not sponsoring me but they should https://a.co/d/4pMv9HR
Reply to @Reagan_Airport, 05:19 UTC · 2 likes
Reagan Airport (@Reagan_Airport): Great news! Flights will gradually return to DCA on Monday. Confirm your itinerary directly with your airline. Thanks to our hard-working snow team who drove around the clock to plow, broom and de-ice the airfield.
More details for passengers: https://bit.ly/4rf9aJc
We are so back
Reply, 05:24 UTC
Ok then!
Tweet, 05:28 UTC · 9 likes
Got some fun timelapses of the snow thanks to the fact that Arlington, VA traffic camera streams are public
(video, plays on X)
Reply to @L0m3z, 06:53 UTC · 9 likes
Jonathan Keeperman (@L0m3z): One asymmetry among many is the fiction that the active resistance of a few hundred protesters are an expression of democratic will equal in legitimacy to the democratic will of the 77 million people who voted explicitly for enforcement of federal immigration law
Injustice is injustice even if it has a democratic mandate
See also segregation
Reply to @imitationlearn, 16:26 UTC · 2 likes
imit (@imitationlearn): why do we mock such behavior, have we no empathy
(quoting @why_bun: Using all his 0.01 braincells)
😢
Reply to @Trail_Cams, 16:29 UTC · 3 likes
trailcam (@Trail_Cams):
🦆
Quote tweet, 16:55 UTC · 39 likes
A function of society in general is to take dark triad personality people and mostly align their incentives/preferences with everyone else’s, or restrain them in prison if that fails
Quoting 📚pedagogy in black✊🏾 (@RenyTure): ICE is sexually abusing males in their custody by grabbing and bursting their testicles according to the ACLU and the guardian.
Reply to @ohabryka, 22:06 UTC · 4 likes
Oliver Habryka (@ohabryka): I think it might just be because they mostly have you blocked.
Lmao
Tuesday, January 27, 2026
Tweet, 02:18 UTC · 19 likes
There is no way this is the optimal system (thanks to Claude for the image)
Thread, continued, 02:18 UTC · 3 likes
There’s also no way I’m the only one a medium amount confused until today
Tweet, 05:15 UTC · 2 likes
Claude Code: $25k Google Maps api task (geocoding 6M free text location names) -> ~$5 cloud compute task that will finish overnight
(well claude got it to like $20, I got it to $5 via some additional trickery and scheming)
Thread, continued, 05:15 UTC · 2 likes
I mean I wasn't going to spend the $25k tbc, the actual counterfactual is just no data
Reply to @GjMcGowan, 16:42 UTC · 5 likes
George McGowan (@GjMcGowan): Sometimes I’ll say something EA to someone steeped in regular politics and they’ll give me this look like “I don’t understand how meat like you survived to adulthood”
Based
Quote tweet, 16:59 UTC · 86 likes
I think we should taboo the word “exploitation” and expect that people make a case that an allegedly exploitative thing is in fact bad for the people being exploited
Quoting Billy Binion (@billybinion): These takes are very strange. No one was forced to deliver for DoorDash in the snow. They chose to, because that's how the gig economy works. Trying to take away pay & opportunity from workers is an odd way to protect them.
Reply to @willmacaskill, 22:53 UTC · 2 likes
William MacAskill (@willmacaskill): Here’s my best guess proposal for the design an international AGI project:
- US-led, plus allies
- Weighted voting so that the US can move quickly on most issues, but needs buy-in from other countries for key decisions
- Broad benefit-sharing with non-members
- Sanctions on non-members who try to develop AGI outside of the projectI like this plan for a few reasons:
1. It’s more politically feasible than other proposals for international AGI projects
2. But still limits US power and reduces the risk of AI-enabled dictatorship
3. A project like this could secure a big lead, reducing racing and buying time for safetyIt’s loosely modelled on Intelsat, an international project to create the first global communications satellite network, so I call the plan, “Intelsat for AGI”.
I think an international AGI project like this is probably a desirable compromise between feasibility, the risk of power concentration, and the dangers of racing. But it’s not risk-free on these dimensions, and others disagree that it’s desirable.
What I’m more confident about is that we should be working out in detail what the best version of an international project is, in preparation for worlds where one is set up.
To that end, I’m publishing some of my old notes on this question, setting out my best guess proposal. It’s deliberately pretty specific - I’m hoping I’ll provoke other researchers into writing their own, better plans.
(quoting @willmacaskill: Today I’m publishing a series of research notes on the idea of an international AGI project. The aim is to assess how desirable an international AGI project is, and what the best version of such a project is (taking feasibility into account). The main result is a proposal I call…)
@littIeramblings
Wednesday, January 28, 2026
Quote tweet, 00:45 UTC · 7 likes
What
Quoting Metro Forward (@wmata): By using a “fire snake,” crews were able to safely warm the rail, melt surrounding ice, and complete repairs efficiently — restoring service in a timely manner to keep our customers moving safely.
Tweet, 00:49 UTC · 5 likes
Bruh
Reply to @mrs_alyse, 00:54 UTC · 3 likes
Alyse (@mrs_alyse): "He. Was. A. Nurse."
Listen, Barb, being a nurse is not indicative of being a good person. It's almost a universal experience to see the bullies and mean girls go into nursing.
It also doesn't doesn't make him immune to the natural consequences of his own actions.
Is there some undeserved memetic effect from the nurse thing? Yeah (many such cases in American political discourse)
But by every account the guy was in fact very upstanding? It’s not like people are trying to distract away from some dark history or action
Tweet, 00:58 UTC · 32 likes
My least neoliberal opinion is that unnecessary snow days are good actually
We don’t have to do a song and dance about “danger” just let ppl enjoy the snow
Reply to @Jsevillamol, 01:04 UTC · 6 likes
Jaime Sevilla (@Jsevillamol): Piping hot AI takes
1. All things considered, it's better if we have more advanced AI sooner rather than later.
2. AI is one of the few technologies that could kick us back to a regime of accelerating growth.
3. The biggest risk with current gen AI is in aiding cyberattacks.
4. AI could lead to a tremendous concentration of power.
5. The US weakening restrictions on selling GPUs to China would be a positive development for the world.
6. I think it's more likely than not that AI will end up in charge of the world.
7. If you wanted humans to end up in control of most resources of the universe, I have bad news for you. If you just want humans to have a peaceful retirement, I am cautiously optimistic.
8. AI will be the biggest sociopolitical change of this century, and worth thinking very seriously about.
9. We likely won't get Dyson spheres before 2050. Sorry.
10. AI development is much more fragile than people realize. It is very capital-intensive, so it requires a lot of trust from investors, which could be easily shattered.
1) wrong
2) right
3) probably wrong
4) right
5) really wrong
6) right
7) wrong (not cautiously optimistic, although it’s not impossible)
8) right
9) idk
10) wrong, it could have been easily shattered, perhaps, but we’re past that point. The models are in fact economically useful now
Reply, 16:27 UTC · 1 like
Tap your card against a physical device, choose tip then and there on the screen yourself
Quote tweet, 16:46 UTC · 15 likes
.@GoodfireAI, like OpenAI, raised a bunch of money, talent, and goodwill by posturing as a safety-first company and presumably hopes nobody important or influential is going to notice when they ✨pivot✨
Quoting Max Alexander (@absurdlymax): Another one bites the dust (safety company pivots to training frontier models)
Thread, continued, 16:46 UTC · 9 likes
By all means, convince me I’m wrong. I don’t have a cow in this derby or whatever the phrase is
Thread, continued, 16:46 UTC · 6 likes
Unlike OpenAI, I don’t think this is illegal. Merely bad
Reply to @absurdlymax, 16:49 UTC · 4 likes
Max Alexander (@absurdlymax): They call them meal squares but they're actually rectangles
But are they non-square rectangles 👀
Reply to his own earlier tweet, 16:49 UTC · 3 likes
Aaron Bergman, January 28, 2026: Unlike OpenAI, I don’t think this is illegal. Merely bad
.
Reply to @eric_ho, 16:51 UTC · 12 likes
Eric Ho (@eric_ho): interp happy hour at our office in SF on Thursday, where you can hear from our technical staff on understanding & steering large models (kimi k2 thinking)
our goal is to hire 10+ MLEs in the next few months who can train and design large models and move insanely quickly
I don’t know you at all or the company that well but seems bad and wrong
Tweet, 20:32 UTC · 3 likes
Would like to see an RCT of propranolol where the outcome variable is second date yes or no
Tweet, 20:53 UTC · 7 likes
Self-prescription licensing is a sub-libertarian but maybe an elegant and more tractable policy than making everything OTC:
1) works within existing system to a greater extent
Thread, continued, 20:53 UTC · 7 likes
2) still do have pharmacist oversight, so it’s really not de-regulated
3) filter for conscientiousness
4) delay between prescription and access, even if only a day, is something
Thread, continued, 20:53 UTC · 6 likes
5) mechanism to regulate quantity dispensed if that’s part of some political compromise (I haven’t actually analyzed anything but my guess is that this cuts overdoses by >90%, ofc at some substantial cost depending on what the thresholds are)
Thread, continued, 20:53 UTC · 3 likes
6) idk how i feel about this, but possibly you could have your license revoked for various reasons. I don’t love this one because it seems ripe for government abuse, but it depends on what the actual reasons are
Thread, continued, 20:53 UTC · 3 likes
7) insofar as this preserves the FDA quality control oversight function, I’m mostly ok with that. I mean it’s a legitimate political discussion to have and of course protects rent seeking but like whatever
Reduces organized political opposition
Thread, continued, 20:53 UTC · 5 likes
8) mitigates the “yo bro try some of this” phenomenon where people don’t know what they’re taking because you have to write out the exact drug and dose
Thread, continued, 20:53 UTC · 5 likes
9) I don’t love this one either, but potentially you can make the test pretty serious and some combination of effortful and difficult to pass. An axis of political compromise
I mean at the limit this is just med school, kinda
Thursday, January 29, 2026
Reply to @Sauers_, 02:01 UTC · 2 likes
Sauers (@Sauers_): @AaronBergman18 @eric_ho why? is training a big model inherently unsafe?
Honestly it might be but more importantly, differentially pushing capabilities relative to safety is bad yeah
Quote tweet, 04:13 UTC · 11 likes
I hope like 7 decision making executives enjoy their slightly nicer third home
Quoting Dmitri Alperovitch (@DAlperovitch): From @RepMoolenaar @ChinaSelect: “NVIDIA provided extensive technical support that enabled DeepSeek—now
integrated into People’s Liberation Army (PLA) systems and a demonstrated cyber security risk—to achieve frontier AI capabilities”
Reply to @ratlimit, 05:17 UTC · 8 likes
@ratlimit (@ratlimit): You know too much about gender, dogs and pissing
Sounds like the actual abundance of knowledge is in spacesuits which is a perfectly legitimate special interest
Reply to @polisisti, 13:52 UTC · 2 likes
Alessandro (@polisisti): @Kirsten3531 Kirsten, you're giving @AaronBergman18 ideas! 😄
😈
Quote tweet, 14:11 UTC · 3 likes
This is actually a profound point
Quoting Lysandre, qui tire les cartes (🇫🇷7/18-9/20) (@9chabard): after working with claude i finally understand the emotional appeal of this image. there's simply no point in me saying "listen here you little shit you have completely failed to fix the same problem ten times in a row and just screwed things up worse" because i cannot actually cause it suffering
Thread, continued, 14:11 UTC · 2 likes
Cc @AndyMasley
Tweet, 15:02 UTC · 213 likes
2025-30 is plausibly the highest leverage time for the leftist economic project in history and they don’t really seem to be on the ball
Reply to @grep_cat_, 19:20 UTC · 7 likes
James (@grep_cat_): I could make this work. Loft bed with desk underneath. 1700 lux high-CRI lights. And for gods sake get some melamine sponges
(quoting @0xLTR: you gotta move to NYC bro its a movie)
Main issue is air quality. Solve that and I’d be happy with 50 sq ft and maybe quite a bit less (currently at 75)
Reply to @max_spero_, 19:29 UTC · 18 likes
Max Spero (@max_spero_): Did you know you can get fancy custom chopsticks shipped from China @ $2/unit?
Fun fact: when I was 3 I managed to fall on a chopstick got it physically stuck in my neck, not out the other side but like past the skin and next to the artery
Didn’t splinter because it was a fancy high quality chopstick for some reason. If it was a normal takeout chopstick it would have splintered (or so I’m told) and I probably would have died bc that is not a good artery to fuck with
So fancy chopsticks may save you or your kid’s life some day!
Reply to @eric23332, 19:33 UTC · 10 likes
Eric23332 (@eric23332): If they were rational enough to see where AI is going, they probably wouldn't have been advocating leftist economics until now.
There are a few of them! @bijibiji2023 gather ur people
Quote tweet, 19:44 UTC · 2 likes
I don’t think anyone should be sent to a Salvadoran prison (or anything worse than that) but sometimes I hope there’s a God so that people like Trump do ultimately get something like a fair trial and have to answer for themselves and their actions
Quoting Bryan Caplan (@bryan_caplan): It would be poetic justice if everyone who ordered, planned, and enforced the deportations were sent to a Salvadoran prison.
Tweet, 21:04 UTC · 68 likes
Fuck it is feeling very over, very doomy indeed
Reply to @max_spero_, 21:13 UTC · 2 likes
Max Spero (@max_spero_): holy shit that's horrifying. I'm glad you're OK
Thank you, I don’t think my mother found the episode especially enjoyable or relaxing
Reply to @dundereloise, 21:47 UTC · 6 likes
__eloïse__ (@dundereloise): Anyway, this guy seems cool, maybe consider going on a date with him?
https://docs.google.com/document/d/10q8DD7V7xLMcne7pKNlmhRbmYXtSq4qUvGnkRLZMN8Y/edit?tab=t.0
If this guy can’t get a gf it’s over
Reply to @tautologer, 22:08 UTC · 2 likes
tautologer (@tautologer): taking anxiety meds and stimulants at the same time
No this kinda just does work in the case of propranolol. Probably not benzos tho
Friday, January 30, 2026
Reply to @tyler_m_john, 01:19 UTC · 4 likes
Tyler John (@tyler_m_john): Google already gave the DoW an unsafeguarded model
Bruh
Reply to @ComInTheMarg, 01:20 UTC · 8 likes
Comments in the Margin (@ComInTheMarg): @AaronBergman18 Why doomy? Politics? AGI?
More the latter but both
Tweet, 01:31 UTC · 19 likes
At some level of energy abundance we would have snow removal via evaporative scaled-up hair dryer/flamethrower
Quote tweet, 03:48 UTC · 41 likes
We don’t deserve Andy
Quoting Andy Masley (@AndyMasley): Unfortunately Bender and Hanna made what I think are some false claims about me on Bluesky that I'd like to respond to.
First, I think it makes sense to alert people to an error in a book that's 3 orders of magnitude off. Hao was extremely gracious in fixing it.
Reply to @paulg, 04:05 UTC · 5 likes
Paul Graham (@paulg): Stripe's new frontpage shows the scope of their ambition. A ticker for percent of global GDP! This is not a gimmick. Patrick and John have always thought in these terms.
Fuck it’s such a good business (in the sense of “being good at making money”)
Saturday, January 31, 2026
Reply to @AnthropicAI, 02:07 UTC · 49 likes
Anthropic (@AnthropicAI): On December 8, the Perseverance rover safely trundled across the surface of Mars.
This was the first AI-planned drive on another planet. And it was planned by Claude.
OpenAI social media team: fuck how do we spin these illegal corporate shenanigans the right way
Anthropic social media team: “Claude planned a trip on Mars, Mr. Amodei? Yeah I think we can put together a tweet thread about that”
Quote tweet, 02:13 UTC · 15 likes
Me on Twitter
Quoting Alan Cole (@AlanMCole): @besttrousers They talk *so* much and the gap in quality between their smartest ideas and their dumbest ideas is staggering.
Tweet, 03:19 UTC · 10 likes
Fr they should probably nuke this
Quote tweet, 22:26 UTC · 23 likes
👀
Quoting Liv (@livgorton): Just a heads up: I signed a permanent non-disparagement agreement when I joined Goodfire. To keep things simple, I won't be commenting on the company, its research, etc.
Reply to @BogdanIonutCir2, 22:28 UTC · 5 likes
Bogdan Ionut Cirstea (@BogdanIonutCir2): tbh, it seems a bit crazy for them to make this kind of ask, especially since my impression was that they were kind of an AI-safety-first org
Real
Reply to @tszzl, 23:52 UTC · 21 likes
roon (@tszzl): wildly bullish
¿
End of January 2026. Previous month: December 2025. Next month: February 2026. Back to the archive index.