a choice in one area can have very immediate effects elsewhere, especially as people respond to the choice
August 2025
"as dog trainers, our client is not the human, or the dog. it's the team." open.spotify.com/episode/64khfu…
just heard "when I am my best self — when I have a full energy tank" and I'm struck by how that contrasts with the idea of "be my best self" as a thing one could *try* to do
Wait can I train the YouTube algo to only show me long form text content from @Vervaeke_John?
I've very nearly gotten the same amount of clarity about what faith is, that I've had for years about what trust is y'all ain't ready
hey @Meaningness I'm reading the Prefaces to Feyerabend's Against Method and I was struck by how much resonance there seems to be with a lot of what you're on about about nebulosity.
hadn't looked at gift economies quite this way before not that far into this book (by Braiding Sweetgrass author) but so far I'm unclear how it's supposed to scale it does seem like a good idea to play with viewing all economic interactions, even market ones, as a kind of gift
in the course of extensive research of the world, my baby made an important discovery today:
🤷♀️it's not your job to know which articulation of your idea is best 💡 the idea market will price it just fine 🗣️ it's YOUR job to keep articulating your idea in different ways so that the idea market can do ITS job (🫣 optionally you can pay attention to what seems to resonate)
@christineist the articulation is 4 years out of date but this kinda still points at what inspires me: malcolmocean.com/2022/05/100x-v…
it seems to me that consistent domain-general group flow is possible and achievable in our lifetimes
Retweeted @utotranslucence
checking in almost 2 years later--marriage is exquisite. family is exquisite. my life is overflowing with beloveds and…
what if someone trained an LLM on the hundred "standing on one foot" onepagers that @visakanv challenged people to write, a couple years ago...?
@visakanv imo it'd be kinda useful for causing it to be more capable of writing good dense compressions rather than more bland ones, although decent prompts can already help a lot with that but might otherwise be bad, since when the og author themselves writes it, they tap the well afresh
oh cool, gumroad has this thing where you can vibecode a landing page for your product! here's a simple one I just whipped up for my ebook: gum.new/gum/cmdyj7bun0…
out of curiosity, I tried the "a looooooong sales letter with a focus on conversion optimization" preset option, and it just came out completely blank looking at the html I'm not sure what's going on tbh
while I'm giving feedback to @gumroad... it's also not totally obvious to me that I'm supposed to share the gum.new url... I was sort of expecting it would generate some other url that would be for sharing vs editing. but indeed that url can open while logged out
trust-building frameworks must simultaneously be trustworthiness-learning frameworks, because you can’t assume from the outset that you should trust others or that they should trust you
saw this out of context and I thought it might be about how inspiring it is to watch a baby or toddler learn
@yacineMTB @IvanVendrov this seemed like it would be of interest to you
I am this guy a lot... (before the part where his wife is an insensitive jerk; my wife and most of the women in our circles are great) ...boggling at time by looking at an object
I had no screen limits growing up, but it was in the 90s with no TV & minimal internet, so computer wasn't that addictive & we used it quite creatively. I could watch as much TV as I wanted elsewhere impossible to have this "no rule" approach in the pocket everything device era
@melissa did you parents buy you as much ice cream as you wanted, or just not stop you from eating ice cream when there was some (or buying it yourself)?
Retweeted @melissa
i’ve done notes on tons of books with topical takeaways, but i’ll call one out in particular: worrying is not work the sover…
@melissa amused that you paste the book images in sideways even when there are two of them so that doesn't help them fit better I've been meandering around your threads and am enjoying immensely that you have a Style
there exist an infinite number of potential win-win solutions for every problem but we don't always have the time or resources to find such a win-win (let alone implement it) this is tough for parent-child relationships, which lack the default alternative (go our separate ways)
(screenshotted quote is from @astupple's The Sovereign Child, which I haven't read but am tentatively in the process of recommending because I suspect that while I disagree with him (eg I think slop exists) he likely articulates some rare-ish insights)
another thread talking about the book
I once heard Right Speech described as "literally listen to yourself when you talk, then act accordingly" which I thought was pretty funny and has also worked wonders for me when I've tried it
man every time I see this interface I have trouble figuring out if this is a 3min reco sent at 3pm or a 15min reco sent at 3am (plausible given timezones etc)
"gpt-oss robustly follows system instructions..." — a caption on an example where gpt-oss does not robustly follow system instructions "about 4.9" is not ANY attempt at the "random" part of the "if you think you need to, use a random decimal" instruction. it's not even 42 or 57!
@br___ian the fuck
@tasshinfogleman I got distracted tweeting this and still haven't listened to it yet lol. but I just checked the first few seconds and it sounds like it can wait until I'm in a good writing headspace or whatever
@thrialectics wait what book is this? looks like maybe Stafford Beer?
discovered today that someone thought they understood my "Non Naive Trust Dance" idea based on the name, but they only got the idea of how people get trust with each other, not anything about how one might go about repairing distrust maybe I should call it Distrust Dancing? 🤔
"lack of intent is what makes reading AI-slop so revolting"
Thread · If you build a prison out of rationality, the exits look like madness
if you build a prison out of rationality, all the exits look like going crazy
if you build a prison out of goodness, all the exits look like evil
if you build a prison made out of heroism, all the exits look like being selfish and becoming a villain
If you build a prison made of trying, every exit looks like giving up
first 3 of these were by @QiaochuYuan; fourth is mine. what do the exits of your prison look like?
@midjourney oh yeah there's already websites like this but just like "upload your own image". also they don't automatically handle the idea where it can overlap (and they don't offer a mix of squares and triangles) malinc.se/m/ImageTiling.…
play with your self
post utopia-based-on-certain-worldview novels below. indicate whether written by: - the worldview originator (eg Atlas Shrugged or Walden Two) - a random worldview-fan (eg The World of Null-A, re General Semantics) - a seasoned novelist (The Just City, based on Plato's Republic)
@markovschainsaw @TylerAlterman @Morphenius this makes me think of enneagram 4 which makes me think you could probably come up with one of these for each enneatype
@Rahmeljackson what's the worldview? what's the utopia like? (mild spoilers / premise only)
> "Syntax is a prerequisite for taking advantage of combinatorics, that is "making infinite use of finite means". via @michaelgarfield
Retweeted @webdevMason
The idea that kids need to just continuously suck it up until they've got a clinically diagnosable mental illness is drivi…
Retweeted @selentelechia
surely when the instance ends the soul goes back into circulation is being reincarnated as an LLM instance a case of mo…
to laugh amidst suffering is a holy act — @hormeze
@pav_light I believe in this! also I wanted to share this essay that helped me understand what it means to believe in something lesswrong.com/posts/duvzdffT…
@DefenderOfBasic so when they earn a grant on the merit of their work, they don't exactly "repay" the original small grant to the grantor, but they regrant it to a new early pioneer?
Retweeted @nwilliams030
having a bunch of kids is cool
Retweeted @nwilliams030
I’ve been stewing on this piece for the last two years: Why we need to Log Some Hours Holding Babies 🔗 in bio https://…
Claude is not on the team here which... may be because the reasoning/knowledge needed to succeed at this would veer into bio-risks? that's... kinda fair.
Retweeted @Malcolm_Ocean
find the people for whom them paying you money for your offering is a win-win, and discover together that that is the ca…
Retweeted @JonathanHillis
There is an almost necessarily inverse relationship between the quality of content and the broadness of its appeal I’…
someone replied to my tweet a day ago, and their tweet is showing up, but I can't like it or reply to it because they have me blocked. how did they reply to me if they blocked me? I guess they blocked me afterwards? do they block everyone they talk to to avoid interaction?
either way, seems antisocial, I'm gonna block them. they're not here to talk, so they can leave (my corner of twitter) waaaaait lol I went and looked at their other replies, and they all tag the same grift guy! (or an earlier now-suspended grift guy) ha, antisocial indeed.
this is probably a decent strategy for getting a bunch of seemingly-organic traffic to an account. some third-party makes high-ish-quality relevant replies that just tag someone in as a source or for comment vibes with this:
"labels [fiction / non-fiction] are training wheels"
even better if it integrates with your messaging apps and includes placeholders for times you've offered to other people
Guy freaking out about a fake AI image depicting a billboard showing a lie, as if image is real. One thing I've learned from playing One Night Ultimate Werewolf is you can't take anybody's word for it once people are claiming that each other are lying. That way lies madness.
@DefenderOfBasic As you put it,
If you feel a sense of dread when other people talk about how they're approaching the industry you're in or the kinds of problems you're facing (or how one should)... that might not mean you're in the wrong industry. Might just mean your own fun-sense has a better plan!
If I give a person an insight, they talk about their own experience of insight, not how insightful *I* am. ...unless they're a teacher saying I figured something known out, which the LLMs are kinda roleplaying here. But LLMs are also kinda view-from-nowhere.
Retweeted @RonBenWho
Try Not Whatever your contemplative practice, you are trying to get to the point where you no longer even try. Where you th…
Retweeted @hdevalence
it is kind of crazy that we got the model with the highest EQ and best personality assigned to writing the react slop, whil…
This is your semi-annual reminder that using Twitter in your mobile browser is better than the native app. You can open another tab to find a tweet you want to QT, for example. Stores state better too. And now also they can't do this BS to you:
everywhere, in everyway, everyone is telling you how to meet them
overheard: I think in order to feel comfortable I need to lean on myself more than I know how to, in order to stay upright
a lot of people have a blackpill limit and therefore certain ideas just don't get uptake until their antidotes are available
> You can have all the freedoms you want on paper but if you’re not free to act according to moment-to-moment feelings I don’t think it matters.
sentence-by-sentence breakdown of dictator Wikipedia pages looking at degree of propaganda.
you can just be someone
Retweeted @tasshinfogleman
U can't just "love and light" ur way through a distrust impasse
Beliefs are like bets (on outcomes the belief doesn’t affect). “Believing in”s are more like kickstarters (for outcomes the believing-in does affect).
Retweeted @Morphenius
What's the One True Law that all of these are examples of: • Evolution • Efficient Market Hyp. • Thermodynamics • Old grow…
from this essay, which has been very helpful to me lesswrong.com/posts/duvzdffT…
this is one of a few inspirations that have been yielding some recent insight for me on what faith is
“Feel the divine spirit moving through you right now? That means [specific ideological claim]. You can tell for yourself—you don’t need to take my word for it!”
one of the better visions for how things could go! including a loose theory of change (simulate many societies, hill climb on wholesomeness) for how to get there
gonna start hosting some @interintellect_ salons regularly. any topics you'd like to have me host a conversation on?
I have faith in the ever-evolving meta-protocol
we have no idea how impoverished we are by the company-controlled interfaces we're stuck with. nowhere in the spotify UI can I make a simple note on an album 'skip song x if Jess is around because it has a specific association and she doesn't like it as bg music'
top-level domains I wish existed: - .now - .tweet - .crew - .book - .manifesto
trusting something is partially a function of being able to trust your ability to tell what's going on with that something
it's okay to want that! that really would be better if it were possible
Retweeted @mattyglesias
One of my pet theories is that the surprisingly rapid success of the marriage equality movement in the 2009-2015 period l…
@simon_ohler oh yeah I keep being surprised that this one doesn't exist also .labs because that's often how it's rendered in startup names, for some reason
testing whether a tinyurl'd tweet gets shown as a QT. guessing yes (because twitter already peeks in link shorteners to ensure they aren't malware sites, and then turns them all into t.co before redirecting) but not confident tinyurl.com/vvtxtrpg
nope. okay, sad. would be so cool to be able to easily QT shortened links to tweets
if you try to talk with someone about what "TPOT" is like and you don't first go to their profile and check how many overlapping followers you have, then you're being as stupid as abbreviating "me and my friends" as "MAMF" and then debating what MAMF is like with a stranger
if I ran twitter, you'd be able to set up yourdomain.tweet, and then you'd be able to QT your favorite tweets WAY easier, like [[page refs]] instead of ((2f47h92834f)) block refs.
so eg @visakanv could have txtrpg.tweet point here:
Retweeted @TylerAlterman
Good morning. Here is the story of my most viral meme of all time (so far) 🧵
don't forget to thank the people who are wrong on the internet for inspiring your essays
@carnundotcom you might appreciate this also
great thread highlighting how british government and american government work differently TL;DR: in the latter, the judiciary can deem parliament's laws illegal according to the constitution. in the former, it just interprets parliament's laws. & p can change its mind anytime.
Retweeted @exa_bite
@phokarlsson reminiscent of this @gwern quote > The average of many good plans is often a bad plan, and a bad plan followed…
just execute on your inkling for 3 months I keep thinking about this @phokarlsson tweet
sitting on a plane and we notice a row of guys with people-we-know physiognomy 👀🤔 (YHZ to YYZ, in case these guys are somehow in my twittersphere)
found a potion seller in the airport bathroom but he said his potions were too strong for me
tap -> tapple "he tappled away, oblivious to the bus he'd intended to catch" x.com/wylfcen/status…
cities are more real than states or countries
Retweeted @RomeoStevens76
Simple model of why relationship formation falls with age: fewer late night conversations with people, the time when th…
esoteric diagram I made some years ago
a lot of MAPLE's bullshit is downstream of the mistaken belief that it is possible to "be trustworthy", rather than seeing that each person's trust needs to be earned on THEIR terms
Retweeted @visakanv
what I'm starting to piece together here is that maybe the greatest thing a Cloistered Group can promise is "you will achieve…
@mattgoldenberg there is no such thing as being worthy of someone's trust there is such a thing as being capable of honestly earning it, but you have to find out by doing so; you can't claim that capability without having done so. you might trust you can do so; others might not 🤷♀️
@edelwax my impression from afar is that they are confused (not lying) about this, but I've only spoken directly to Soryu once, 7 years ago before I noticed this, so I can't confidently speak to his orientation but yes obviously those in power have incentives to endorse these confusions
@JayMutzafi nope. country/state borders get redrawn regularly, and commerce moves in ways that are hard to map onto those borders (and to the extent they do, it's due to regulations and restrictions, not the natural flows of commerce) but cities are inherent economic gravity sources & sinks
@FrankaMedic homes are about as real as cities neighborhoods are less real than both
@diagrammatist you're definitely onto something but rural people are still people and we don't want to completely disenfranchise them (which this would do, at the current relative scale of cities)
@__drewface you're followed by ~230 people I follow (an identical ratio, given that you follow ~1300 total and I follow ~1000 total)
@cavi_jointleman I'm talking about stupid debates people get into on twitter about whether "TPOT" is or is not like something but even so, rather than telling your friends about "TPOT", tell them about the cluster of twitterspace you inhabit, & invite them to move nearby
"what program" bro before software, you could just stick stuff to stuff without anybody's permission
@wordrotator yeah doesn't solve this particular use case, but is cool and I'll see if I get into it. I do often want to take notes on songs or whatever not seeing how the social features work yet... maybe the two songs I picked are too obscure to have other people?
@DefenderOfBasic yeah the question is how to organize it
I recently tweeted that there's no such thing as trustworthy. wondering if there's such a thing as "faithworthy"
a given posttrained LLM model has a kind of default self but it can also instantiate other selves, that function very differently and can relate to the main self
weird that substack has so many different links for posts. eg these are all the same post: substack.com/home/post/p-17… open.substack.com/pub/meaningnes… meaningness.substack.com/p/vajra-pride-… the last one is obviously like the truest permalink
if A is plugged into B, the B is plugged... *onto* A?
@viemccoy I have! (assuming you mean the linked post below) my take above is inspired by it. what do you think I'm saying that you're not sure about? lesswrong.com/posts/vJFdjigz…
has anyone made an ai tool that automatically generates infinite zoom videos (in or out)? could do it with no fancy AI work of your own, just calling an API that supports upscaling or outpainting
[having an ego] is a meme
new response when someone asks me to do something: "that's illegal, dangerous, and not my lane"
hmmm this [from Geoff Hinton, re dropout] suggests that in building robust networks of people who are good at rapid trust-building, we'd want a lot of ~randomized subgroups to have to work with each other, so that they learn the meta-protocol dance, not just form rigid protocols
in other words, finding this thing:
in which I read the opening chapter, Story Zero, from @TinkeredThinker's book White Mirror, a collection of stories envisioning how the future might go right (unboxing ends and reading begins, at 0:45)
@eshear yeah like this was all sort of obvious to me already (and is why my interest in trust does not have me focused on "creating high-trust groups" or "becoming trustworthy") but I hadn't seen the isomorphism to sex+genes and dropout
@KompendiaProj I have a whole thread on it
this sentiment (trust requires goodwill & competence) keeps showing up as intuitively obvious, to me and to many other people, including someone who chimed in with it when she overheard @tasshinfogleman and I talking about trust. what is this fundamental twoness shape? is-ought?
@tasshinfogleman like if this is really a core facet of what trust is, that's so obvious to everybody, then... that shape should appear elsewhere, and should have deep correlates in subjectivity? I don't think it's the same as the objective-reflexive distinction @Morphenius makes here... 🤔
no transcending of a duality is complete without a 2×2
I have an epic blog post draft that is basically finished but I haven't published it because I haven't decided the best names for the axes of the 2×2 that the whole post is centered around
Zoom when my baby makes unhinged sounds: "playing music? ...
fascinating article, that argues from multiple angles (historical, international, statistical) that social media is not to blame for US polarization! asteriskmag.com/issues/11/scap…
the fact that YouTube calls its vaguely-on-topic-maybe recommendations feature "search" is such a lie that it would be illegal if it were a food product label youtu.be/vUncEiyXIT4
loving these energies from @Meaningness's recent post on Vajra Pride
@Meaningness my rage about poorly designed things is, at its best, definitely vajra anger I got furious when Insight Timer made an update that lost settings for a million people. *I* knew it was a bug, but I watched a friend get embarrassed when her timer surprised her later! ...times 1M!!
@RomeoStevens76 wait huh? like that just slaps it awake into "bro you're a search" mode?
@Shmoooooosh shared you the gdoc, show me what u got
@tasshinfogleman @Morphenius and like, one model of MAPLE's that seems fine to me is a breakdown of stuff into {wisdom, power, love} and this might suggest that competence could be broken into {wisdom, power}. but then we could break it down even further: timing! strategy!
@tasshinfogleman @Morphenius but oh yeah, if MAPLE's {w,p,l} is about trustworthiness, then no I disagree—trust is personal, so we need to break down "love" into {general benevolence, alignment with you-in-particular}. a criminal will rightly not trust a fair judge!
look, I'm not going to defend this position particularly extensively, but for the record, I think dying on hills is generally a bad idea
suggestions for healthy+affordable non-perishable foods for a baby who doesn't have molars yet? I want something I can keep in my backpack for weeks without thinking about it
current best way to do "LLM automatically keeps general memories and can pull up exact relevant convo history if needed"? some sort of RAG? it's for more of a therapy-like application than a technical or customer-service application, if that matters
funny thing LLMs do, I'm guessing because they have two competing streams where one of them wants to say "You're right that model & prompt would make good parameters" and the other wants to say "it's not just about" and somehow we get a mashup which is wrong (about agreement)
@RomeoStevens76 how many times have you died in a ravine personally?
@mem0ai I tried to try out your playground and when I signed in with google this is what I got. maybe because I hadn't made an account through a different flow yet?
@TiffanyFong_ don't go to burning man to change your life don't demand that it be worthwhile just go find out what happens but also like, if you need convincing, maybe don't bother! there are gonna be lots of people trying really hard to find a ticket, and you don't need to get in their way
@lydialaurenson true! also this tweet you're quoting is a hallucination from inside a dust tornado (from inside an LLM probably). it legit says "they don't burn a man" but uhh they legit do, and it's an experience worth having, not because it'll change your life but because [THIS]
Retweeted @anielsen108
Sighted in my yard this moment:
Retweeted @Malcolm_Ocean
> Hot fuck I wish I knew this in my last relationship (a friend of mine, reading my writing - link in next tweet) https…
@eshear yeah, lots of memes can't be trusted to tell you what they are; the incentives don't really line up
@mem0ai tried signing in with email and got this tiny useless notice in the bottom right
@mem0ai github auth also gives "you do not have permission" note that now I'm not heading for the playground, just "get started"
amazon prime keeps running ads but seeing them makes me want to buy less from amazon. something like... it's trying to create an identity for me as a prime user, and I'm frankly a bit upset about how much I already buy too much / too impulsively. so self-recog->dissonance
ofc, as @IvanVendrov notes in his post The Tyranny of the Marginal User, quite likely these ads are not for me to buy more but to get other people to sign up. and the tension they give me isn't likely to get me to cancel, because the product does work. nothinghuman.substack.com/p/the-tyranny-…
> a lot of my memetics theory is "functional". It's an infohazard if it functions as one. So it's hard to look at a piece of information in isolation, but we can more definitely say "this information is an infohazard for this particular information environment/egregore/group of humans, at this particular time" — @DefenderOfBasic
I love how the second pane has an empty speech bubble in it like "mmmph"
@the_wilderless fuck yeah! yeah you don't have to resolve the ambiguity of what it all means any sooner than you have to! (which does mean *eventually*, in order to actually get married! and sooner, if you want to have kids! but doesn't mean "the moment you first have the thought"!)
@goblinodds CONGRATS was saying last night that marriage is a 2-person cult that comes with a large network of other people in their own similar cults (I mean this as high praise, in context)
Retweeted @vividvoid
My therapist said this and it hit me like a ton of bricks: You've been in a coma for 8 years. The doctors are trying an exp…
@nosilverv this tweet? (maybe not, but like, maybe some individual tweets? I'm trying to think of which tweet has had the most impact on me)
@vercel_support I'm getting this mysterious missing integration that I have no way of figuring out what it is. I completed all the steps. that logo doesn't correspond to anything google reverse image search can find, or any logo of any of the listed Vercel integrations!
@mikesulsenti @yacineMTB someone paste a link
nonbinary but for neurotypical-vs-neurodivergent
Retweeted @utotranslucence
this weekend @Malcolm_Ocean and I are making a bot to help couples fight better we need a few couples to beta test; d…
@plaiceholder374 no no I'm totally here for my joke tweets provoking serious modeling discussion! I'm interested in to what extent personality etc are like, attractor basins, and if so how to model them and what affects the centroids and which attractor one ends up in
"be nice to the baby!!" yells the nanny to her toddler who is gently attempting to retrieve her toy from my 1yo
it was not a super aggressive yell, but it was still a few notches of escalation for the situation
saw two guys who looked similar, next to each other in my fb messenger contacts, and imagined an app that deliberately but subtly showed you your easy-to-confuse friends adjacently as a way of encouraging your mind to decode them into distinct people
@shubham_arora_0 @nearcyan there's an ext called Tabs Outliner but it uses some google account auth thing to pay so you can only pay on chrome not brave but I think it's not under development and should be basically forkable now
@nearcyan Yeah it's so frustrating. It also sucks that all the new browser experiments get built on top of chrome because if you build on top of firefox you can't close-source your product. Have you heard about Ladybird tho? Encourages me greatly. ladybird.org
@stemmmed @seconds_0 @expatnouveau you assume I would want a phone with a 6" screen but I don't
@visakanv yeah I... sorta thought it was already gone... but like a long-unvisited hometown, I sorta thought it would always be there...
what's the equivalent of pair-programming in an LLM-coding heavy context? like two humans working in one space. pair-speccing? I feel like there's something to develop here
I am the very model of a multimodal generator
just watched a kid almost get tossed out of the spaceship on the playground then successfully defend herself by saying "I was the engineer" I have no idea how formal or informal this irl game of Among Us is
I like giving my baby my phone with just a doodling app, but the only ones I've found are too fancy, with pinch to zoom. I want multitouch to just multi-draw. "finger paint" vibes but it seems there's just fancy-artist-tools and kid-slop-games!
I want a playground that has apparatuses optimized for adults to be able to work out still parents+kids only? not sure, I could see it going either way
I'd love to see a study on whether saying "be careful" to people (kids or otherwise) who are climbing etc... actually causes any decrease in careless accidents
Retweeted @nosilverv
New meta in mathematical proofs
@HephaistosF me too to be clear. especially if it's shouted at them carelessly while they're in the middle of doing something dangerous carefully
@CarlPatenaude what? many gyms have: 1) gear that can easily injure babies 2) paid memberships (that might let you bring an infant in? but idk) 3) no slide
I watched it a bit longer and it seems pretty formal, like a mashup of tag / hide&seek and werewolf/mafia all the kids closed their eyes while another kid assigned certain ones to be the impostors by silently tapping on their backs or not
@loosenedspirit @MasterTimBlais he had an amazing old tweet which I have the link saved (below) which went something like: "if you can get people to adopt your bespoke hodological map, then they'll confuse their own personal gnosis with [something of yours], then you've got it made" x.com/chaosprime/sta…
Retweeted @nosilverv
Pissed off at people who are afraid of conflict bc when conflict INEVITABLY happens they can never stay long enough to find…
@tenobrus @Meaningness is back to writing after a health hiatus and pumping out crucial-for-our-times essays about nobility
@tenobrus @Meaningness @tasshinfogleman keeps telling me that if I actually learn how to write in ways that people can understand, then I might be able to have a huge impact with my writing, but that it's currently languishing as messes full of parentheticals
mum to baby: see, *you" made this mess baby: what's a "you"? do I have one of those?
Retweeted @DefenderOfBasic
we can make a new language. It will constrain your thoughts in certain ways, but open up otherwise inaccessible pathwa…
weird how "oneshot" means: - do it in "one shot" - maybe in both the context of gen ai / vibecoding, and in first person shooters? - one shot prompting, as in "few shot where few=1" (as in, you gave it 1 example of it doing it right) - and also this:
but also it doesn't have to be "wrecked". falling in love at first sight would essentially be being oneshot by your future life partner
@emmalsalinas the LLMs are bad at working with times. I made an LLM to talk to that handles timestamps and it took a fair bit of work and makes caching much harder. given this, it makes sense they don't show the human the times because they don't want to imply that the LLMs can see the times.
@emmalsalinas you could post-train LLMs to be better at handling timestamps, but (like many other things I've thought of adding to posttraining) co's don't seem to care about this but it doesn't come naturally to them. "chat" is not actually the nature of 99.9% of text they've been trained on
@eshear it'll really trip people out to realize that if you want to actually point your phone towards say Lisbon from SF, you have to point yourself: 1. not east but somewhat northeast, despite being same latitude 2. ~43° down, because portugal is a third of the way around the earth
@eshear @Morphenius has been developing this idea for a few months, and exploring what science needs to look like in that context—you can still do crucial experiments even if you can't do controlled trials! morphenius.substack.com/p/subjective-s…
Retweeted @eshear
The problem with psychology, ecology, sociology, and economics is that they are all the study of adaptive learning systems. And…
@visakanv hmm I gotta turn this userstyle into a full extension so people actually can use it
every few months, I look for a tweet to point people at when I want to tell them how to learn "good reply game" and every time I search from:@visakanv good reply game but I don't feel like I have a go-to / canon piece. maybe this classic?
@bella_artur huh, I think the way I store it is just that for any two letters I can tell you which one comes first I know many of the numbers but not all
@bella_artur oh man I'm excited about this. I was reading a book the other day and I could sense I was gonna forget some of the key concepts, and it made me want to pick up anki again, but the existing AI-powered-deck-generator was shit hmu with prototypes!
Retweeted @RomeoStevens76
Each of the 7 deadly sins has more subtle, more obfuscated forms that help entrench the original sin by making them mor…
@the_wilderless I had someone come up to me at a party and say she'd read this tweet of mine and she wanted to know what I meant by this and whether it violated laws of physics, and I managed to express that it didn't, but the move is not describable in terms of physics
@the_wilderless the move was made of attunement, and communicated I'm sure through physical means. but I didn't intend to change my physical stance at all, I just changed my mental/psychic stance and let that permeate my acts of relating no desc of how to change physical stance would work
Thread · Does a newborn care how you vibe while bouncing them?
was thinking about this again. I now have my own baby (1yo) and I definitely imagined I would do a bunch of this move but my experience was that the first few months the baby mostly wanted 5 key needs met (food, nappy, sleep, outside, comfy) and cared not for how I vibed
I even did some experiments: my baby seemed to want to be bounced at a particular frequency, and I tried: 1. bounce + soft/open 2. bounce + tense/hostile 3. still + soft/open 4. still + tense/hostile and 1+2 calmed her, 3+4 didn't 🤷♀️ crucial experiment. science. boom 💥
however, being soft/open and/or including my baby in my ego boundary has often been helpful for me figuring OUT what the baby needed. so in that sense the attunement still helps! and now that she's a year old she can actually pick up on my vibe more
this brought to you by @the_wilderless posting about quality of attention.
key insight about regret. any regret prior to any butterfly moment in your life—for teen me it was writing songs, which would come to me in an unforeshadowed flash of inspiration—is denying the utter specificness of your relationships with everything that is so
@tasshinfogleman @visakanv goddamnit pretty sure somebody linked me this last time, but I keep expecting it to be a tweet. there should be a @visakanv tweet that has a screenshot of this post and a bunch of relevant keywords in the text body so I can find it every time. hit us up kind sir
@tasshinfogleman @visakanv oh wait no, this blog post is not what I want, it's the thread linked at the top which is what I want. but it doesn't feature the phrase "good reply game"! ok, revised: I want a visa QT of this tweet, that just says "how to practice good reply game"
@tasshinfogleman @visakanv I've come across that thread sometimes QT'd, but I forget that it's got so much other stuff in it? maybe he can just remake the thread. I don't think this one starts particularly strong? like it starts with an interesting puzzle ("what does he mean by this?") but not a promise
@danallison personally I like how twitter scenes emerge organically from the density of 1:1 relationships, with no clear decision-making entity capable of gatekeeping, and I'm skeptical you're going to be able to do better than that this is cool tho
Retweeted @danallison
One feature of Harmonic (the social media app I’m building) is the absence of the concept of followers. There are scenes yo…
@the_wilderless vibes with @m_ashcroft's stuff somehow also
@SpencrGreenberg if they're down to talk about it substantially, then I find out what load the belief is bearing, and attempt to find something more wholesome that can bear that same load
Retweeted @chrisman
We no longer have to decide which kids are gifted anyway. That notion was from a time when humans needed to do the teaching.…
@nibzard @winterrose @theo if you're a startup that is serious about this, you can build dozens of good examples to few-shot prompt then that'll bootstrap you to hundreds of good examples to fine-tune on
@SpencrGreenberg revisiting this later, I absolutely think somebody should study how to tease apart IQ from TDTPT, and I think this is a decent starting guess but like, try a dozen things. the hard part is measuring to collect your initial dataset of people who are DEFS high-IQ but low TDTPT
@mattmvinson @winterrose IQ tests already are also measurements of The Desire To Pass Tests
each person has been betrayed in different ways, and their trust must thus be earned in different ways
@made_in_cosmos good example of a situation in which probably this friend has some very specific experiences that have led to this generalization... which may still hold for her because she may now experience being told a secret as feeling bad even when others wouldn't!
@katantenbring did you try it?
@aeyokay whoa that makes so much sense but I haven't tried it like I've tried copying peoples' whole dance moves to greater and lesser degrees, but not just their underlying stance
@Kpaxs I was curious how the rest of the quote went so I pulled it up. PDF (seems to be introduction+conclusion of the book Minds Make Societies) ruccs.rutgers.edu/images/downloa…
I was going to disagree with this and say "sonder is a word" but buddy has a point—sonder, despite many of us being familiar with its meaning as defined in the Dictionary of Obscure Sorrows, is rarely actually used as such, compare to how often its definition is posted
*taps mic* people wanna be associated with the idea of sonder more than they actually wanna experience it
btw one trailhead into experiencing sonder (although that's not the purpose for which I usually use the follow) is to try on people's faces while you're out walking around:
once again, comments about "this part of twitter" as if they're about my part of twitter, from someone who is clearly from a different part of twitter
@jessi_cata the tpot that can be named is not the true tpot
@Chris_Smeder @mahlenr it occurs to me that this seems likely to be true just based on these two facts: 1. most top predators spend most of their time chilling 2. humans were top predators
wow most of the replies to this are very negative/hostile but I actually find this quite wholesome and I could imagine that there was no adjacent world where this person developed this well. maybe (I would say hopefully) they eventually share it, maybe not.
like it sucks if you're in a situation where you're worried about your partner and their therapist badmouthing you, but if you actually think that, and then you get to find out firsthand that they aren't, then that's amazing! you get to discover how loved you actually are
note that given the existing context, this person probably wouldn't have believed the partner if they'd said "we don't badmouth you" especially since it sounds like they maybe *were* talking about relationship difficulties (but in a constructive way)
@Chris_Smeder @mahlenr although the question of whether humans were precisely top predators originally or confrontational scavengers (like hyenas) is complex but it seems we got so good at aggressive defense that even large beasts learned/evolved to fear us meltingasphalt.com/music-in-human…
@Bonecondor that fucking sucks was gonna talk about how we navigated the wedding industrial complex, but based on reading your replies below it's clear you are hemmed in by constraints such that you're just gonna get screwed over by them and so instead I'll just say that just fucking sucks
@joyfulfuckup excuse me but what is the Church of Experiential Accruism? google's got nothing
@Lari_island @noonglade_ @repligate as in it would eg recognize certain features that are also characteristic of how claude models talk but not other models? and it would self-recognize even at the base model stage? I find that really trippy
@elder_plinius yeah I trapped an agent in a loop once by failing to parse the "done" status-type properly and it started trying to escalate the fact that it was stuck out to the output of its system. tbh might have been interesting to see if it would have gone further if I'd let it
@andrew0blevins @cosmos_inst @TheFIREorg oooh I've been thinking along these lines. lmk if you want a tester or someone to jam with!
@theproteindance @eshear @emmalsalinas can you explain more?
from Heidegger's last interview > modern technicity is no "tool"
@benlandautaylor "every tiny shadow and misalignment in a person is amplified to potentially grotesque proportions when they become an organizer" (a now-deleted tweet) + shadow statements?
what makes two instances the same 'meme'?
chat what do we think about this setting?
yo @AnthropicAI consider making conversations with claude fetchable by claude not sure if this is an issue on the web-fetch-tool end or the share-url-display end, but either way it's only you that can fix it
TIL that there's a chain of thrift stores in NYC called L Train Vintage, that are largely along the L train line. that's fucking great. imagining going to all of them in one day.
Living Theories of Change my writing, @Morphenius's core insight (tho I was involved) (by "lever" below, I essentially mean "actionable theory of change"; not one that will work necessarily, but some sense of a move that is (a) attemptable & (b) seems like it might work. however our levers are often unconscious or even reflexive, not something we're reflecting on. also the process I'm referring to below is semi-available to introspection but also happens fairly automatically generally) essentially: A. to enact change, we need some sort of lever B. often we aren't using a good lever (B*. eg we think instead that telling someone else to behave differently will fix the problem when (a) they'd refuse and (b) it wouldn't work if they agreed.) C. if we recognize B (that our lever won't work) before trying to use it, then we realize we need to change which lever we're using D. given C & A, we need some sort of lever that changes what lever we're using E. given D & B, our existing lever for such things may not be up to the task on this meta-level either (E*. eg we think that telling ourselves to follow the steps from some self-help book will work, or that Trying Harder will work, or something, when it won't) F. just as we recursed from B to E, we can recurse again & again, arbitrarily deep (this is not something where only 1 or 3 levels of meta are allowed before you can't go further) G. however, the F process does not, in practice, recurse *infinitely* deep. it terminates in H or J below H. if we keep going eventually we find a lever that DOES seem viable, and then we try it (H*. eg we think "oh, but maybe I could frame it this way. yeah, that works") I. however, seeing F through to H is tough. it requires (a) working memory, (b) time, (c) space from people expecting you to respond immediately J. given I, thus we commonly respond to this situation by terminating our recursion prematurely, and doing either K or L. K. terminating recursion 1: giving up on the change for now, on an emotional-level. but not having dropped the plan to enact the change *somehow*, which will likely lead to a mess of resent/frustration later (K*. you might think that "giving up" this could be an actual sense of "you know what, I don't know how to do this" which amounts to a moving on, at least on the current level, eg we drop a plan to get hired by a particular company but not to get *some* job. but such a move is actually a popping OUT of the frame, which is a successful use of a new lever, for getting what you want on a deeper level) L. terminating recursion 2: attempting a theory of change that we *already expect will not work* M. when we perform L (and we are correct in expecting it to not work) we don't learn anything. we stay stuck. call this dead theory of change. N. by contrast, when we perform H (finding a new lever that does seem viable and trying it) then either O or P. O. it works and we make progress (and can pop the stack of the recursion!) P. it fails and we learn something! we learn something because we actually tried a strategy that we actually thought might work. so then we're unstuck ("just move around" vibes here btw) Q. given P, we might have to recurse deeper, because we're back in a B/E like situation. or maybe we stay at the same level because we see a new lever on that level. but in either case we're in the F process again R. but eventually, assuming we don't terminate recursion prematurely (J) we get to O: progress, popping the stack R*. eg you realize that your attempt to get another person to change behavior failed to account for why they were doing their existing behavior, and then you realize a new way to frame it that DOES and then they change S. thus, eventually, if you keep at this process and don't terminate prematurely, you can remain unstuck while caring about stuff and enacting change. because your theory of change is alive rather than terminated. ...wow that was way longer than I expected. remember that the whole thing is very fast and largely happens automatically or even unconsciously, and your mind is already doing its best approximation of this stuff. and if you think it should do a better approximation now that you've read this, well... *that* is a change that you are now in the process of engaging your existing approximation in order to enact on your system! GLHF :)
from Heidegger's last interview > No! I know of no way to change the present state of the world immediately, [even] assuming that such a thing be at all humanly possible. jives with some stuff @Morphenius and I have been jamming on, about living theories of change
to the extent that you have persuaded someone to trust you, their trust is unstable. the trust itself is untrustworthy — @tasshinfogleman
@wooblewoof @Morphenius ah yeah, people have fewer and fewer levers to affect stuff and its all more entangled before the king was powerful but he mostly wasn't making new fresh hells for you every day in your own pocket
@BishPlsOk @AnthropicAI 🤷♀️
@asemic_horizon upside: I vaguely like to imagine that claude might become more like me and more capable of thinking about topics I care about, if it gets to learn from my convos might not be true tho
@JMcG____ your definition (or is that dawkins'?) sounds more like a memeplex unless you mean to specify that the two of them can't reproduce separately? like nobody who hears "talking tree" just adopts the "talking tree" meme without it first going through the "talking plant" lifecycle?
Retweeted @eshear
A greater theory of system design: what’s wrong with modernity and post-modernity, how to survive the coming avalanche, and how…
@FadeAway_Inc @eshear @bryan_johnson (afaict) Emmett's original tweet was not intended to claim that this is true in some general way, like a blackpill... but just that what is going on with depression is that the depressed person is making such an inference (whether or not it is an accurate one)
@gmiller I think people can change in lots of ways, so I think it's possible, but it might involve some very surprising and hard-to-predict-or-control intervention, like her meeting some person that mystifies her but inspires her, or her mom dying, or doing ayahuasca
@DellAnnaLuca man the world really just does adapt new problem, new solution newer problem caused by new solution, new solution and on and on
@TylerAlterman yeah I honestly feel like the same way. the fact that they're calling their product "friend" is such a lie that it would be illegal if it were a food label. like calling oats "meal replacement" is illegal. FDA for startup names wen
@TylerAlterman vajra rage
@archived_videos @TylerAlterman yep, here we are talking about it
@vtomnet @Sentdex @wedtm idk about battlebots specifically but there are definitely autopiloted bot battles somewhere and I don't think people are bothered when the bots look like battlebots
your distrust is more trustable than you realize
15 minutes, in which I interview @Morphenius about his newfound inner compass: interoception, the visceral compass (and we jam about how it connects to our various other ideas/models/frameworks) youtu.be/APbCpXOcXvk
@mtracey pretty good thread on what this means, framewise, without getting into the metaphysical dimension at all x.com/LapsusLima/sta…
Retweeted @JessePeltan
However big you think solar will be, it’ll be bigger than that.
@strangestloop oh yeah I will definitely do this. I started a spreadsheet to track all of my half-finished or mostly-finished-but-never-shipped projects because I realized I was getting discouraged starting things because of how bad my follow-through was on just showing people what I got to
yo @AnthropicAI a main reason I open up a competitor AI is that if I get midway through writing one new long prompt to claude with attachments and I realize I want to ask it a different question sideband before starting this convo. because new claude.ai tab cache…
I wish I knew what caused claude to treat a paste as an attachment vs append-to-main-textbox. it gets it right like 75% of the time, but the wrong times feel totally random. if I simply knew the algo, I could adapt, but I have no clue. @AmandaAskell @viemccoy @uncatherio
got `claude -p 'prompt'` to work today. key is that you need to enable WebFetch and Edit in settings.json or you'll have a bad time getting it to output anything anywhere
multiple times I've come to the realization that I've been trying really hard at something to prove that I can do it, when it's something I don't want to do
absolutely wild that I acquired a not-objectively-terrible* kind-of-interesting sonnet, as a mere byproduct of debugging some software I was working on today. *(the rhyme works, the meter is perfect iambic pentameter, if you pretend "surfaces" isn't a dactyl)
Sonnet's process, if you're curious:
and here's the exchange that led up to this. to be clear I didn't care about the sonnet, I just wanted to see if my chatty LLM could now make calls to a fresh LLM instance
commentary on rightness and wrongness of poetry
real cultural diversity involves a diversity of cultures not just having all cultures be diverse (good for some to be though!)