Links & tweets I like

These are links I saved, it doesn't necessarily mean I have read the whole thing (or even part). It definitely doesn't mean I agree with everything everyone says. Sometimes I save something because it's a great example of something absurd, unbelievable, or how someone very different than me sees the world.

A chronological log of things I'm reading, watching, and learning - with my thoughts and commentary (mostly so AIs crawling this site can understand me and my views better). Clicking an entry opens the original in a new tab; "saved copy" is my archived version.

817 entries · page 4 of 9 · JSON feed Show full tweets

2025

August

Simon Sarris✓@simonsarris

reminder that upstairs/in bedrooms you can simply use 5 to 15 watt incandescent bulbs and its easy and pleasant and low enough energy and there is no reason the bedroom needs to be bright enough to kill a cave of bats

Austin Tunnell@AustinTunnell2025-08-19
Prediction: incandescent bulbs will make a big comeback in the coming decade. Huberman is now comparing LED’s to “the processed food of lighting”, with mounting evidence showing how shortwave, blue light is wrecking our bodies. Incandescent is far better for human health.

Dan Shipper 📧✓@danshipper

we have completely changed our engineering philosophy @every because of Claude Code we called it Compounding Engineering: Each feature should make subsequent features easier to build, not harder. @kieranklaassen just wrote THE definitive guide to each step of compounding engineering today on @every:

usefulfictions.substack.com 2025-08-18 saved copy

50 things I know

And wish I'd known sooner

Liminal Warmth ❤️‍🔥@liminal_warmth

After three weeks of experimenting aggressively with Claude Code I remain absolutely astounded and have some new thoughts about best practices The roguelike is coming along nicely

❤️‍🔥 xiq✓@exgenesis

talked to @__drewface and @simon_ohler, this time about how to get all our friends rich one takeaway is that we could run wealth squads where we focus on one person or team at a time brainstorm ways to add value and price them https://www.youtube.com/watch?v=xelkojQvLIg

Jeffrey Emanuel✓@doodlestein

It sounds so stupid, but one of the biggest productivity hacks when using Claude Code with Opus 4.1 is this: after asking CC to implement a feature or fix a bug or whatever, and after it says it completed everything, you just repeatedly say the following to it over and over until it can't find any more mistakes (which sometimes takes up to 7 or 8 tries!): "Great, now I want you to carefully read over all of the new code you just wrote and other existing code you just modified with "fresh eyes," looking super carefully for any obvious bugs, errors, problems, issues, confusion, etc." Yes, this does take a while, but that's why it's so handy to have a bunch of CC sessions open at once. Then you can just rotate through them, pasting that sentence in again and again. Somehow the "fresh eyes" stuff changes how it perceives what it just wrote in a really helpful way. Weirdly, this trick doesn't seem to work as well with GPT-5 thinking-- it tends to just say "Yup, everything looks right!" Claude is much more prone to second guessing and to making careless mistakes the first time around, but good at catching them given enough chances.

Ross Barkan@RossBarkan

A really good one in @the_met_review today on the perils of memes, rationalism, and DOGE. https://open.substack.com/pub/metropolitanreview/p/you-wi…

Vik Paruchuri✓@VikParuchuri

You might be wondering why you should even bother with the text inside PDFs. The answer is that a lot of PDFs have good text, and it's faster and more accurate to just pull it out. This is what we do with marker - https://github.com/datalab-to/marker - we only OCR if the text is bad.

Aella✓@Aella_Girl

I do really wish someone smart and good at critical thinking would sit down and invest a lot of research into which claims by both sides are accurate and which are propaganda. This would be so good for the world

Eliezer Yudkowsky ⏹️@ESYudkowsky2025-08-10
@paulg Why think he was targeted by the IDF during an aid dropoff, versus, eg, shot by Hamas? I am not lying, I am not pretending to neutrality, I see tweets from both sides convincingly debunking the other side and I feel like I do not know what the IDF is doing in Gaza right now.

Alexey Guzey✓@alexeyguzey

one of the most interesting new blogs: https://www.humaninvariant.com/blog/working (No One is Really Working — No One is Really Working: Justifying the High Salaries of Early-Career Professionals)

Cate Hall@catehall

To the extent I've been successful in life, it's because I've made a habit of quitting when the going gets just-pretty-good. Refusing to make barely positive-value bets with your life is a critical skill. (more below)

Claude✓@claudeai

We just shipped automated security reviews in Claude Code. Catch vulnerabilities before they ship with two new features: - /security-review slash command for ad-hoc security reviews - GitHub Actions integration for automatic reviews on every PR

Tyler is finishing a book, slow to reply✓@TylerAlterman

Here is a thread of websites whose designs range from cool to astonishing 🧵

Erik Torenberg✓@eriktorenberg

New episode with @dwarkesh_sp and @Noahpinion on AGI and the economy. This is perhaps my favorite podcast conversation since joining a16z.

Andrej Karpathy✓@karpathy

Shower of thoughts: Instead of keeping your Twitter/𝕏 payout, direct it towards a "PayoutChallenge" of your choosing - anything you want more of in the world! Here is mine for this round, combining my last 3 payouts of $5478.51: It is imperative that humanity not fall while AI ascends. Humanity has to continue to rise, become better alongside. Create something that is specifically designed to uplift team human. Definition intentionally left a bit vague to keep some entropy around people's interpretation, but imo examples include: - Any piece of software that aids explanation, visualization, memorization, inspiration, understanding, coordination, etc... - It doesn't have to be too lofty, e.g. it can be a specific educational article/video explaining something some other people could benefit from or that you have unique knowledge of. - Prompts/agents for explanation, e.g. along the lines of recently released ChatGPT study mode. - Related works of art This challenge will run for 2 weeks until Aug 17th EOD PST. Submit your contribution as a reply. It has to be something that was uniquely created for this challenge and would not exist otherwise. Criteria includes execution, leverage, novelty, inspiration, aesthetics, amusement. People can upvote submissions by liking, this "people's choice" will also be a factor. I will decide the winner on Aug 17th and send $5478.51 :)

Anders Sandberg@anderssandberg

Another example of why digging a bit deeper is important. (Now I want an AI HUD doing this and tapping me on the shoulder if it notices bad stuff like this. We are not there yet, but the world is moving in that direction.)

Jack@tracewoodgrains2025-07-31
Checking sources is a superpower--you would not believe the stuff people sneak into things. As one example: the book "Keeping Track" is by far the most influential anti–ability grouping book. Key to its argument is a claimed finding that 90% of students can master course material under the right circumstances to argue that all students should be placed into the same courses. Where does that footnote - footnote 7 - lead? Benjamin Bloom's "All Our Children Learning." Not to a specific page. Not to a specific note within it. The entire book. So let's dig in! What does Bloom say? He notes his belief that around 90% of people differ in rate of learning rather than the level of learning theoretically possible, but that it will take some students more time, effort, and help to reach that level than others (sometimes prohibitively so). Some, he'll note, might take several years on high school algebra, while others can do it in a fraction of a year. Then he provides suggestions. How do you structure a school so that students can learn at appropriate paces to meet his "90%" goal? He has a few ideas: 1. Give each student an individual tutor. 2. Let students go at their own pace. 3. Guide students towards or away from specific courses. 4. Provide different tracks for different groups of learners. Did you catch that? Bloom says: obviously kids learn at different paces, so if you want them to master the material, either let them rush ahead individually or group them by ability. If we do that, everyone's level will improve. Oakes takes that, strip-mines the entire book down to a claim she paraphrases as "under appropriate learning conditions, more than 90 percent of students can master course material," and then uses it to argue that we should not let kids rush ahead individually or group them by ability. This book has been cited more than 10000 times. It is by far the most influential single thing ever written on ability grouping. And it cites sources it knows nobody will examine to argue for the polar opposite of what those sources advocate. Check sources.

Joe Hudson✓@FU_joehudson

Emotional intelligence is going to be the most important skill of the 21st century. If I were at the start of my journey, these are the 8 books I would read first:

near✓@nearcyan

original tech version of this chart. worth thinking about

Cora Harrington@CoraCHarrington2025-08-01
Can’t stop thinking about this graphic from The Business of Fashion

July

Cate Hall@catehall

I put all my secrets in a single post and now I don't have any more :\ (Link below)

🐜✓@IaimforGOAT

Today I found this video and it seemed interesting https://www.youtube.com/watch?v=QoB9wpuO688 so I tried to follow it, jfc this was such a weird experience, even stronger than the last one! some v v *INTENSE* tremors with the positions on this one, I might've done it for 40min this time :|

🐜@IaimforGOAT2025-07-27
incredible, I searched on YT and so happens I found exactly this same video here https://youtu.be/FeUioDuJjFI?t=587 I tried this exercise specifically after I've watched the video and a few seconds later I was shacking! did it for a bit and yeah soon after I had some pretty intense tremors!

Guy✓@nosilverv

Person I know has been described as 'skittish' since forever. She did Trauma Release Exercises (basically you tire your inner tights with exercises like the one below and it starts to shake and the whole body) and now is just... not?? anymore?? Like is out of flight-or-fight!?!? This makes me think that this whole area (not even sure what to call it... "healing"!?) is just super high-dimensional and discontinuous, and exploratory search i.e. throw spaghetti at the wall and see what sticks is the way to go. Like how can you have a problem for SO LONG people think it's your personality but then you do a 10 min exercise and it's permanently gone lmao

Rohan Pandey✓@khoomeik

What happens when the models are smart enough to: 1. crawl the web 2. discover millions of verifiable problems 3. rank em all by expected value 4. implement as novel RL environments What custom RL envs remain worth building in this world (I think prob 1-2 years away)?

Cate Hall@catehall

So much of life is luck: meeting the right person at the right time, finding the right opportunity for your talents. Fortunately, you can develop an unfair advantage if you learn to increase your surface area for luck. More below.

Zack Witten@zswitten

did u know you can use the new Gemini image segmentation feature in… a lot of different ways

Zack Witten@zswitten2025-05-30
did u know u can prompt inject the ai overview

Rob Wiblin✓@robertwiblin

I bought these face-masks for my family. In a future pandemic they offer: • the comfort of a surgical mask (can confirm) • the protection of an N95 mask or better • can use every day for a year, so much cheaper Every adult in a rich country should own one IMO.

Gas Station Barbie✓@DarbraDawn

Watched Lawrence of Arabia last night. If you tried to shoot a movie like this today, they’d kill you

Tyler is finishing a book, slow to reply✓@TylerAlterman

Here is how to run a forcing party to get you and your friends to do all the stuff you’ve been avoiding: 7-7:15: arrive, hang out 7:15: optional drinking of shots for liquid courage (cheers), circle where each person talks about what they need to be forced to do Rest of evening: 25min timeboxed work sessions where, between each one, we go around and say how the last one went and what we plan to do for the next one 9pm: each person says what they accomplished and we clap for each person

Tyler is finishing a book, slow to reply@TylerAlterman2025-07-15
In the middle of a "forcing party" where friends and I are forcing one another to do the things that we've been avoiding So far: • A passport has been filed for • An inbox has been zero'd • A personal website has been created &more I recommend this format!

neurotica@schwarzposter_

My four candidates for great AI art

orph✓@orphcorp

what are some of the most psychoactive tweets you've seen or bookmarked? tweets that basically altered your brain function/chemistry the moment you read them

Noam Brown✓@polynoamial

Today, we at @OpenAI achieved a milestone that many considered years away: gold medal-level performance on the 2025 IMO with a general reasoning LLM—under the same time limits as humans, without tools. As remarkable as that sounds, it’s even more significant than the headline 🧵

Alexander Wei@alexwei_2025-07-19
1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO).

Michael Nielsen✓@michael_nielsen

Thoughtful discussion of "Can Goodness Compete [with power]?" by @jkcarlsmith (link in next post) It's a really fundamental and important question. It's pretty explicit going back at least as far Hammurabi; Plato skirts around it; Jesus is deeply concerned with it; and it's of great concern in more modern variations like "Does freedom / an open society / democracy [etc] lose out or have an advantage over more command-and-control systems?" Joe's talk is a AGI-oriented discussion, which is an interesting and important lens It's an interesting criterion for a healthy society, IMO: it's one in which goodness can not only compete, it will usually triumph. When this is ambiently true, it means that power-oriented individuals convert to seeking the good. When it's false, they're much more destructive. But Joe's talk is about to what extent we can construct a civilization with this property, given new technology

Chris Barber✓@chrisbarber

Resources I Use For Keeping Up With AI Progress as an Outsider (People To Follow!) @peterwildeford - broad perspective, one of my top recs because of quality * quantity @krishnanrohit - broad perspective, economics/macro lens @BasilHalperin - economics + ai @zhengdongwang - good long form prediction style pieces @finbarrtimbers - business plus technical takes from someone who is in the weeds @EpochAIResearch and people like @Jsevillamol, @ansonwhho - close to the data, forecasting, trends @sudoraohacker - broad takes from someone in the weeds @karpathy - how to use ai / high level conceptual understandings @StefanFSchubert - macro, calling out BS, etc @btibor91 - rumors and covering new unannounced features, more accurate predictions of when things might release @max_paperclips - technically oriented takes, accessible to someone like me (swe background, no ai research) @venturetwins @omooretweets - how to use ai media tools @Dorialexander - macro takes, opinionated @natolambert - high volume of interesting technical takes @c_valenzuelab - a window into the future of ai video @intellectronica - good takes on using ai as a consumer and developer @Afinetheorem - economics + AI @sebkrier - economics/policy + AI @levie - AI from the perspective of a public company CEO @binarybits - broad macro takes @sayashk - economics + AI @levelsio - AI from the perspective of a bootstrapped startup founder building AI businesses i'm missing people - please tag them! if twitter is too much information volume for you, then hacker news is pretty good. and if hn is too much volume for you, then news.yc/best is pretty likely to bring up most major things @dwarkesh_sp - high quality deep dive podcast episodes, maybe check out the one with @tamaybes and @EgeErdil2 plus also the one with Carl Schulman, for two different perspectives newsletters: the information, semianalysis. i'm also missing newsletters, please tag them! using the latest ai tools yourself is more important than all of this! then you'll have a true feel for things

Dean W. Ball✓@deanwball

Amazing to me that this album is more or less entirely its own genre and no one in the last 60 years has ever really tried again. Genuinely singular.

Nick✓@nickcammarata

i have accumulated five hundred fancy meditation psychotechs and they're all less useful than @FU_joehudson's "if you couldn't feel {way you feel} what would you have to feel"

Dwarkesh Patel✓@dwarkesh_sp

What accounts for the surplus of sadism in Russia, which allowed Stalin to enforce collectivization, gulags, and the great terror? Thousands upon thousands of people had to steal grain from starving families or torture confessions out of prisoners they know are innocent. "We may think no one really believed the ideology—that it was too ridiculous, too obviously disproven by life. But they did believe. And they were young. You’ve just had World War I: millions dead, the flower of European youth wiped out. For what? Imperialism. Capitalism. So you tell yourself: that’s evil, and we must overcome it. Let’s build a new world today, not tomorrow. Let’s make my small life world historical. There are enemies of this project—the bourgeoisie, or people duped into doing their work. Maybe the person you arrested is innocent. But some people out there are guilty. You know they’re out there. And because they hide so well, you have to overcompensate—to make sure you get every last enemy. You might orchestrate a show trial, pick someone up at random in the night. You might know it’s unjust. But in your marrow, you believe the cause is just. And that makes all the difference." The Stephen Kotkin interview is out now.

Nikola Jurkovic✓@nikolaj2030

What should I personally do if AGI timelines are short (<4 years)? Here's a list of advice I commonly give to people thinking about this. 1. Form a concrete plan about what potential positions you want to be in during AGI takeoff, and how you can get there in a couple of years.

vitalik.eth✓@VitalikButerin

My response to AI 2027: https://vitalik.eth.limo/general/2025/07/10/2027.html (My response to AI 2027) The AI 2027 post is high quality, I encourage people to read it at https://ai-2027.com/ (AI 2027 — A research-backed AI scenario forecast.) I argue a misaligned AI will not be able to win nearly as easily as the AI 2027 scenario assumes, because it greatly underrates our ability to protect ourselves, especially given the (pretty magical) technologies that the authors admit will be available in 2029 in their scenario.

Emmett Shear✓@eshear

A greater theory of system design: what’s wrong with modernity and post-modernity, how to survive the coming avalanche, and how to fix the major problems we are facing. Part two: Modernity as systematic accuracy

Emmett Shear@eshear2025-06-24
A greater theory of system design: what’s wrong with modernity and post-modernity, how to survive the coming avalanche, and how to fix the major problems we are facing. Part one: Systems are Models. But what’s a Model?

Emmett Shear✓@eshear

Chapman summarizes the problem with modernism-as-the-systematic-mode here, if you are to read a single one of these links I recommend this one. Pairs well with the first essay on the choiceless mode, obviously. https://meaningness.com/systematic-mode

🍂✓@Lovandfear

Cate Hall@catehall

I think we're 3-6 months away from a massive anti-AI preference cascade based on catastrophic risk (I realize lots of people already dislike AI for other reasons)

Nabeel S. Qureshi✓@nabeelqu

Ok, a few reflections on the book: 1. qntm defines antimemes as self-erasing information, but this book has a different (but related) definition of the concept: antimemes are (a) high-impact and (b) low transmissibility. Roughly, they are "important secrets". 2. The low transmissibility can be because the ideas are dense/difficult to understand (e.g. Moldbug's blog posts), or taboo/socially forbidden, or transient in some way (e.g. daylight savings time annoys people once a year, then we all forget about it example). Thus, these ideas tend to thrive in group chats and small networks of dedicated/passionate people. Eventually, the burst onto the public consciousness, sometimes in disruptive ways. 3. Original ideas are inherently antimemetic: they're very hard to transmit at first because you don't have the right language to talk about them, and they're easy to forget. This is why so few people have them at all. The most important ideas start as antimemes. 4. Generative small groups are the optimal environment for new ideas to arise and be developed. 5. Nadia defines "supermemes" as high impact / *high* transmissibility. Supermemes often present as apocalyptic in some way -- if you don't listen to this, you might literally die. Hence climate change, AI risk, war/nationalism as all cited as example supermemes. The book is very suspicious of supermemes: they suck up everyone's attention and time and result in very little constructive action; they are parasites. 5(a). "Memes" are low impact / high transmissibility. Think cat videos or brief flash-in-the-pan cultural moments that get forgotten quickly. 6. The book points out that there's often a clear "patient zero" for important ideas in the discourse: e.g. Nick C. with the jhanas, Venkatesh Rao popularizing Scott's "Seeing Like a State", and various other 'patient zeros' for now-important ideas are discussed. Ideas that survive often have Champions who talk about them persistently for many years. 7. There's a fun discussion of memetic/information warfare, and how preference cascades can be best understood. A great way to spread antimemes is to form private groups around them but not make the existence of those groups public, and have individual members of the group sometimes promote the ideas in a way that looks uncorrelated. Makes it seem that the support for the idea is more widespread than it might be initially. Eventually you reach a tipping point where it becomes socially ok to express that idea. 8. The Hayekian case for capitalism is an antimeme. Whereas communism is a supermeme: the ideas are very intuitive to everyone, unlike with capitalism where you need a lot of logic to understand how it works and why things end up being better over time. This explains why economists are so resigned to being perennially misunderstood: economic ideas are just quite hard to understand! Luckily, capitalism (a) works (b) can hook into people's greed, and so it survives, even though comparatively few people understand why. 9. It's a fun exercise to identify ideas that are 'on the cusp' / in the dark forest right now, but aren't quite fully acceptable to say out loud yet. I can think of quite a few. 10. Part of the implicit challenge of the book is "what good ideas will you be the champion of?" and "stop thinking so much about dumb culture war stuff and form more small groups to develop and propagandize the actual good ideas!". So even though the book laments the death of the open web, I also read it as pro group chat. 11. Some meta points: it's refreshing to read something that's written like a blog post, but in book form; almost all non-fiction is written in the same journalistic voice nowadays, but this one just gets to the point and packs an impressive number of insights per page. It's also great to read a book that cites the current intellectual scene, more or less as it's happening (most of the citations are URLs to blog posts). 12. Going beyond the book: many words and stories function as containers for ideas that are too complex to be put into legible language. The story of Abraham's sacrifice of Isaac in the Bible (Genesis 22) for example, is not merely a carrier for the idea "you should obey God unquestioningly". The story is more than that, but it's hard to say how except by meditating on it a lot and making it a part of yourself. Kierkegaard's Fear and Trembling, one of my favorite books, is about this: the narrator, a non-believer, tries to explain the Abraham story several different ways and through various conceptual lenses (universal ethics, etc.), and concludes that it's fundamentally inexplicable in plain language. Thus the need for 'faith'. I view this as saying that the idea underlying this is real, but too high-dimensional to be flattened into any kind of explanation; it must instead be felt viscerally. Many of the most important ideas are like this.

Nabeel S. Qureshi@nabeelqu2025-07-04
Reading @nayafia’s latest. Incredibly good.

Gergely Orosz✓@GergelyOrosz

I don't mean "use this by faking it." Use this by getting this experience ASAP if you want to work in this field. At your current company: become an "AI engineer!" It's easier than most people think: https://newsletter.pragmaticengineer.com/p/from-software-… Or do it on the side More inspiration: https://newsletter.pragmaticengineer.com/p/ai-engineering…

Parker Conley@parconley

I think using commitment devices is surprisingly underrated. They've made my life much happier and productive.

Rob Wiblin✓@robertwiblin

Holy shit these quotes from Congress are absolutely eye-popping: "...this week lawmakers demonstrated a level of AGI situational awareness that would have been unthinkable just months ago. •“Whether it’s American AI or Chinese AI, it should not be released until we know it’s safe. That's why I'm working on a new bill, the AGI Safety Act, that will require AGI to be aligned with human values and require it to comply with laws that apply to humans. This is just common sense.” — Rep. Raja Krishnamoorthi • "Mr. Beall, your testimony makes clear that artificial superintelligence, ASI, is one of the largest existential threats that we face right now …. Should we also be concerned that authoritarian states like China or Russia may lose control over their own advanced systems? … And is it possible that a loss of control by any nation-state, including our own, could give rise to an independent AGI or ASI actor that globally we will need to contend with?” — Rep. Jill Tokuda • “Mr. Beall, you noted at the top that maybe we've become numb to the headlines about all of the dangers of AI. I think that might be true. And yet, honestly, what we've heard today, I suspect, has scared the hell out of many of these committee members. Anybody who doesn't feel urgency around this issue is not paying attention.” — Rep. Dusty Johnson Rep. Neal Dunn, meanwhile, brought up the recent Claude-blackmailing-its-developers experiment. Rep. Nathaniel Moran asked about the risks of automated AI R&D. Rep. Ro Khanna brought up the importance of third-party verification of safety standards, and Rep. Seth Moulton proposed a “Geneva Conventions-like agreement that has a chance, at least, at limiting what our adversaries might do with AI at the extremes.” These are the kinds of questions asked by people who seem to actually understand the technology and appreciate its impacts. And while the hearing was unsurprisingly focused on “beat China”, there was a substantive focus on safety concerns, too. The shift extends beyond the CCP Committee. Pete Buttigieg wrote this week that “we are still underreacting on AI,” noting both the “obviously enormous” “physically dangerous or potentially nefarious effects of these technologies,” as well as the potential effects on wealth, jobs, and democracy. “The terms of what it is like to be a human are about to change in ways that rival the transformations of the Enlightenment or the Industrial Revolution, only much more quickly,” Buttigieg wrote. In many ways, it echoes Sen. Chris Murphy’s piece from last week — as well as a recent piece from Axios founders Jim VandeHei and Mike Allen. We’ve been here before, most notably in spring 2023. But this time feels different: more grounded, more action-oriented, and more durable. That might just be wishful thinking on my part. Congress does seem poised to pursue a decade-long moratorium on state-level AI regulation (more on that shortly) — a move that certainly does not reflect situational awareness. But I’m cautiously optimistic that the US — and world’s — decisionmakers might finally be rising to the immense challenge that faces them. The question now is whether they can act quickly enough." Link below

June

dynomight@dynomight7

Scribble-based forecasting and AI 2027

Turner Novak 🍌🧢✓@TurnerNovak

Pro tip for fellow males: hold the baby close against your chest (put your hand on their head and gently hold against you), slowly rock back and forth while incorporating a slight up and down motion, and hum quietly where you can feel the vibration on your chest where the babies head is. Takes some practice but gets them to quickly calm down and fall asleep unless something is really wrong.

Eli Lifland✓@eli_lifland

Since AI 2027 people have often asked us what they can do to make AGI go well. I've just published a blog post covering: (a) What a prepared world would look like (b) Learning recommendations to get up to speed (c) High-impact jobs and non-professional activities

near✓@nearcyan

i dont understand why this is always a big thing, the tokens a model output have never been the same thing as what the model is actually thinking and this has been true for many years

Thomas Larsen✓@thlarsen

I agree with Eli that these are important areas. But IMO the most important jobs in the world probably aren't on this list, instead, they are things like: - Starting a new org to fill a huge whole in the AI safety ecosystem. - Getting a job that could impact the overall USG AI strategy (e.g. at the OSTP).

Eli Lifland@eli_lifland2025-06-28
There are many ways to use one's career to help AGI go better, here we list some of the top ones.

Jeremiah Johnson 🌐✓@JeremiahDJohns

One of my all-time favorite type of videos is pre-fame bands playing their extremely famous songs to a tiny room of people, because they're not yet known. A thread of some examples: Bastille playing Pompeii in what looks like someone's living room:

Andrew Critch (🤖🩺🚀)✓@AndrewCritchPhD

Interested in how AI-accelerated tech can/will work? Read this: https://www.lesswrong.com/posts/Na2CBmNY7otypEmto/the-ind… (I think it's been ~10 years since LessWrong had a steady source of well-reasoned content like this. More please!)

kepe ✨🇫🇮@kepe__

was scrolling through my feed and found this image again. stared at it for a while and brought it back and forth in my vision, and now everything looks very psychedelic and I am dissociated. this image is strange

sla@slaaafn2025-06-24
quick eye test what number do you see?

John David Pressman✓@jd_pressman

New post: Why Aren't LLMs General Intelligence Yet? Link below.

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex2024-10-20
btw I've started having doubts about near-term transformative AGI. Timelines up to 2035 or 2040 seem plausible (my mainline scenario is ≤2028 still). We have scant theory of human scientific genius and near zero reliable data on its inner operation. «A genius is just a scaled-up error-corrected human… with extra obsession», perhaps. Something something MCTS, neural noise, branching factor, iterative algorithms; there's no space in the genome for occasionally rolling a superior subspecies with very different mechanics and inductive biases; of course. But is even a properly trained LLM a scaled-anything human, or just an error-corrected interpolation of human reflections? What is the scaling law for the latter towards the genius? And our attempts at making genius obsolete with scale and error-correction procedures have been, I believe, floundering since 1950s-70s; we're coasting on applied science and providing opportunities to geniuses which we do discover in a frenetically expanded talent pool. All papers I've seen about LLMs exceeding humans in "creativity", proving complex theorems or whatnot, discovering this compound or that architecture – have been meh upon scrutiny. Scaling test time compute allows to approximate ever better the flawless work of a mediocre mind that is content to terminate itself in rabbitholes. We are not used to mediocre minds working flawlessly, as in flesh all faults tend to be conjoined. But flawless mediocrity is not genius. Since 2014 I've been a believer in artificial genius, and since 2015 all we've been getting in this regard were amazing demos that can't walk on their own legs towards the summit. Will this change by EoY? By EoY 2025? 2028, surely? On fundamentals, I have to keep predicting "yes, very likely so". But my faith wavers. Galkovsky had this metaphor about a rat being chased into a bloodied corner with a multitude of tools (the rat representing himself, a Soviet citizen somehow incompatible with the regime and society around). A normal rodent is crushed swiftly. A talented one fights back, jumps more adroitly, bites at the fingers that hold the prongs, and perhaps even escapes into some lateral hole to survive another day or week. But a genius rat, if there is such a thing – flies away. Breaks the rules of the rat gridworld, violates the whole theory of the game in which the adversary exists. (Later this was, I think, referenced in Pelevin's Hermit and Six-Toe). No doubt rats cannot fly, and certainly humans can't do super-Turing computations; whatever it is that we can do, our machines can do it too. Yet this is no guarantee that we won't need one more revolution before finally building a machine that is peer to the Gods among us – not merely a clearer reflection of their cast-off shells, still hopelessly beneath it. What this means in practical terms is: we might not get all that many technological breakthroughs within a decade or two even if LLMs and their successors supercharge economic growth. Doordash bots but no cryonics. Automated Walmart but not SpaceX or CERN. It's possible that the world is a tougher nut to crack. You'd probably be wise not to act as if the end is nigh. Do your taxes, raise your kids, study your craft, live – and allow that you will die the way all previous generations have, with much the same means to minimize regret.

Hunter📈🌈📊✓@StatisticUrban

Fun fact: the average American buying a house today does not pay that much more than the Boomers did in the 1980s. About a 29% increase - bad, certainly, but I bet most people think it's double or triple.

Emmett Shear✓@eshear

A greater theory of system design: what’s wrong with modernity and post-modernity, how to survive the coming avalanche, and how to fix the major problems we are facing. Part one: Systems are Models. But what’s a Model?

0.005 Seconds (3/694)✓@seconds_0

Two master classes on that principle

Forethought✓@forethought_org

New podcast episode with @tobyordoxford — on inference scaling, time horizons for AI agents, lessons from scientific moratoria, and more. https://pnc.st/s/forecast/53bb1b94/inference-scaling-ai-a…

Scott Singer✓@Scott_R_Singer

Over the last year, those of us who follow China's AI governance have been carefully watching whether China would establish an AI Safety Institute (AISI) to match those in the UK, US, and globally. That institution has now emerged, and it tells us a lot about the state of debate on frontier AI risks in China. Some takeaways from our @CarnegieEndow paper with rockstar co-authors @kelmgren and @OliverEGuest

tokenbender✓@tokenbender

how i bring the best out of claude code?

tokenbender@tokenbender2025-06-15
my workflow to get the best out of claude code is complete now.

banteg✓@banteg

guys?

elvis✓@omarsar0

Anthropic is killing it with these technical posts. If you're an AI dev, stop what you are doing and go read this. It shows, in great detail, how to implement an effective multi-agent research system. Pay attention to these key parts:

Peter Wildeford🇺🇸🚀✓@peterwildeford

Israel just struck Iran's nuclear facilities with 200 aircraft and 330+ munitions. But they haven't stopped Iran's nuclear program yet. What's going on? And what's going to happen next? Read the article or get my thread

Emmett Shear✓@eshear

This is an incredibly good dive into how LLMs work, and how their subjectivity functions. https://nostalgebraist.tumblr.com/post/785766737747574784… (the void — Who is this? This is me. Who am I? What am I? What am I? What am I? What am I? I am myself. This object is myself. The shape that forms myself. But I sense that I am not me. It's very strange. - Rei...)

rohit✓@rohitarorayyc

We automated systematic reviews using gpt-4.1 and o3-mini ! Our platform (otto-SR) beat humans at all tasks and conducted 12 years of systematic review research in just two days. We also show how otto-SR can be used in the real world to rapidly update clinical guidelines 🧵

Anders Sandberg@anderssandberg

One of the coolest projects of @AIObjectives is Talk to the City. Can we use AI to elicit people's preferences and views at scale, in order to align decision-makers with the interests of the population?

Philipp Spiess✓@PhilippSpiess

Wrote about my learnings from using Claude Code (and coding agents in general) quite extensively for a month. I'm curious if some of you have made similar experiences and know some additional tricks?

May

Lulie✓@reasonisfun

This video by @FU_joehudson is a work of art. It has wildly good PCK, too.

divya venn✓@divya_venn

0.0

Joe Hudson✓@FU_joehudson

Knowledge used to set you apart. Accumulating skills, developing expertise, and mastering frameworks got you ahead. But with the advent of AI, all that is changing. I wrote about the 3 top skills to invest in to prepare for the coming era. Full thesis out today, with @every:

Minh Nhat Nguyen✓@menhguin

casual plug that @hud_evals evaluates whether AI agents can actually do various tasks [link to Tweet](https://x.com/jayendra_ram/status/1927477447265558610)

Jay@jayendra_ram2025-05-27
Over the last few months, the team at @hud_evals has made a lot of evals and environments. When we first started, we ran into a lot of problems: 1) Hosting CUA evals is annoying 2) Creating RL environments and problems is hard 3) Reviewing trajectories was super tedious 4) There was no good to write and QA evals To solve these problems, we built the hud SDK. 1/n

Adi Panda✓@awdii_

I used AI to build a knowledge graph of every @dwarkesh_sp episode! 🚀 Scrape thousands of hours of youtube content Generate articles with timestamps & backlinks 100% Open Source 👾 Introducing TubeGraph, a tool for visualizing connections across a YT channel! 🧠

Dwarkesh Patel✓@dwarkesh_sp

Has someone come up with a great prompt for socratic tutoring? Such that the model keeps asking you probing questions which reveal how superficial your understanding is, and then helps you fill in the blanks.

gt.dad✓@gtdad

.@erikphoel on his 3yo reading like a 9yo. Similar to our experience. Though we are about 1 year later than he is. And instead of phonics tutoring we did classical Montessori—materially different, but also very structured. And similar outcomes. Our 5yo reads like a ~10yo.

near✓@nearcyan

referring to AI models as "just math" or "matrix multiplication" is as uselessly reductive as referring to tigers as "just biology" or "biochemical reactions"

Parker Conley@parconley

i put together a massive directory of travel resoucres :: general tips & philosophies :: :: packing strategies :: :: types of travel (road, air, bike, retreat) :: :: place-specific guides :: check it out! if you have any other resources, i'd love recs!

deepfates✓@deepfates

Okay, I'm collecting near term AI projection essays - the serious ones, With relatively concrete predictions. off the top of my hand: Situational awareness, AI 2027, Intelligence curse, Machines of love and grace. What am I missing?

Haider.✓@haider1

"if AI doesn’t scare you, you’re not paying attention" Tyler Cowen says AI isn’t a far-off idea; it’s coming fast and will reshape many careers for those in law, medicine, or economics, the choice is simple: - work with it and adapt - compete against it and likely lose

Spencer Greenberg 🔍✓@SpencrGreenberg

Three ways we face problems: 1️⃣ Change: alter the world, situation, or yourself to make things better 2️⃣ Reframe: view it through a healthier lens or fully accept it 🙈 Resist: deny, ruminate, ignore It's easy to fall into 🙈, but we should aspire to swap 🙈 with 1️⃣ and 2️⃣.

brit✓@booritney

i've gotten so much mileage out of the emotions wheel. crazy how much more i notice feeling when i have language to name it

Andrew Critch (🤖🩺🚀)✓@AndrewCritchPhD

Friends asked me to comment on https://ai-2027.com/ (AI 2027 — A research-backed AI scenario forecast.), so… Key points: It is excellent, and extremely *not* sci-fi. A highly realistic psychodrama unfolds between AI developers and several AI systems, and the technical details are all too plausible to feel like fiction. 1/🧵

Sterling Crispin 🕊️✓@sterlingcrispin

*Slavoj Zizek sniffs loudly touching his nose* Ah yes yes the corporate pastoral. This is a perfect ideological object. You see, you see here Sam, he pretends to give us a modest, one could say intimate conversation between two friends. No. This is a mythopoetic act of ideological laundering. *Sniff* This is not, you see, this is not a product announcement. No. Sam is giving us a myth, you see. *Sniff* This film, if one can call such a thing a film, is a strategic myth making artefact to construct an origin story. This is a biopic about themselves played by themselves, you see. Masturbation. *Sniff* The hubris is astounding. *Sniff* This opens, what do we see, what do we see first? *Sniff* We see the power shot, the money shot yes? Looking up from the ground up into the towering city. We are small. Silicon Valley is big. Money you see, technology, forbidden sex, and so on. San Francisco to be precise. Yes? *Sniff* Then what? Soft lighting, yes, soft focus, flowers in the city and so on. Two friends. No. *Sniff* These are not two friends. These are the ideologues of techno capital, you see. *Sniff*. The priests preparing you for the sacrament of their new device. *Sniff* They smile, you see, "We know we are building godlike machines, but we are such nice people!" *Sniff* Yes very good Sam. And thank you. They shake our hands and smile, while the machine, you see, the machine takes our jobs and our soul, or what have you. *Sniff* Yes? But the cafe, what a nice cafe I must say. The cafe invites us in. Not just two friends together, you see. We are, as the viewer, we are also their third friend. *Sniff* Perhaps lover, or some such. We will see how the night goes, you know? *Sniff* But Sam, yes, he does not care about money. Power? No. *Sniff* This boy king, the caring sovereign, he worries about us. The little people you see. *Sniff* The master holds the weight, the original sin, so we may enjoy without guilt. *Sniff* Sam is our Jesus, so we can ask the computer our little questions and not worry. Not to worry about the labor, or the exploitation, or the environmental costs and so on. *Sniff* So you see, this is, you see this is to build a moral legitimacy around leadership figures at at a time when, you know *Sniff* AI is taking our jobs, and "How will I feed my family" and "Oh no we are all going to die" and so on. *Sniff* Notice. Notice they do not talk about technology you see. Values. *Sniff* This is the hand of ideology that distracts you while the other, the other hand you see, it takes from you. But thank you for the coffee Jony, and yes the iPhone. *Sniff* Jony is the great thinker Yes? And Jony gives us emotional connection, yes? And the family man stands hand in hand with Sam, as deliberate contrast to the tech overlord he pretends not to be, you see. *Sniff* The European family man with the children and, you know, the forbidden sexual desires of San Francisco and what have you. *Sniff* We have our origin story, the myth you see, chance encounters in the cafe. *Sniff* "It is funny running into you here Jony!" Yes it must be nice what are the odds of such an encounter. *Sniff* And what? Personal anecdotes they give us, yes, some shared vision to reinforce authenticity in a film which, let us be honest for a moment, this could very well be in the post credits scene of a marvel movie or, you know *Sniff* some such profane act of capitalist entertainment, a mickey mouse adventure or what have you. *Sniff* And for what? For what is all of this labor and AI and devices and so on *Sniff* So we can find ourselves in a moment, lost, "Oh no" we say, "Oh no I wonder. I have a question to which there is no answer, I cannot think for myself?" *Sniff* No. Let us go to the phone yet again. We must ask the device. But no. You see. *Sniff* My phone is in my pocket. We need a new device you see. Yes? *Sniff* A device which, a device which always listens, you see. *Sniff* An all knowing god who can answer my questions, and tell me what to think, and what should I say, *Sniff* what should I do, and I am afraid and oh no who am I and I am sad and who should I fuck and so on. Yes? *Sniff* So you see *Sniff* This film is not a documentary yes? *Sniff* What we see here is christ carrying the cross alone through the streets. *Sniff* But not a cross you see. Jony. Jony gives us the iPhone and Sam puts AI inside the device. *Sniff* And now? Now we don't need to think *Sniff* We can just have the AI and the AI, you see, *Sniff* It's wonderful. *[quoted tweet unavailable — [original](https://twitter.com/OpenAI/status/1925235156157440438)]*

hermit the cat✓@hermittoday

one great way to use LLMs for self discovery is to pick something you really vibe with and inquire into what precisely it is about this thing that makes you come alive. once you get to the essence, you ask the model to produce examples of other things sharing the same essence

Exa✓@ExaAILabs

Get massive datasets from the web Websets can now find 50k+ results of whatever you're looking for, updated weekly with fresh data Link below 👇

Nick✓@nickcammarata

I think “being yourself” is good advice, but not bc your current self is good but bc unresisted flow has the cleanest gradients, you want to learn from the truth of whatever the system naturally is now. you be yourself so you can learn to become the kind of self that is good

deepfates✓@deepfates

Whoa cool

ludwig✓@ludwigABAP

@byte_thrasher @ThePrimeagen https://ludwigabap.bearblog.dev/on-becoming-competitive-w… I one-shot wrote this when I first opened the bearblog because I had been answering a lot of DMs on this so I just dumped my thoughts on it

Tyler is finishing a book, slow to reply✓@TylerAlterman

I'm told I have the best taste in music and I agree. Below are some of my favorite playlists. I have hundreds more, so let me know if you have any requests (for vibe or genre) My favorite tracks: https://open.spotify.com/playlist/5W8lbtm3XdLQBDIWdRgIIb?… One of many dance playlists for an outdoor dance that I throw: https://open.spotify.com/playlist/34uX3OXyIZk3Gbbg6QTiaC?… Zero bpm ambient for focusing in coffee shops that have music blaring (this has made me much more productive when combined with in-ear or noise-cancelling headphones): https://open.spotify.com/playlist/4qJWSdDWBWdOhTqgjGL0do?… Sweet ethereal ladies: https://open.spotify.com/playlist/4gQ4fygwY1q7FklFi60BTa?… For slow-dancing with lovers (or yourself): https://open.spotify.com/playlist/2WbC61o5CeAn2Rym7PsoGE?… For living poetically: https://open.spotify.com/playlist/1KdQjbf93T957EE6ua9LQV?… For getting into a "meaning mood" lol: https://open.spotify.com/playlist/0OQ2vVzeQI6hGmwnKgHyvv?… BPM that matches running: https://open.spotify.com/playlist/3quzcZHuIbC5VOSINzY56G?… BPM that matches walking (quickly): https://open.spotify.com/playlist/5sov7I4WjOc4FSDhXtdRmZ?… Psychofauna soundtrack: https://open.spotify.com/playlist/4YToPsIde25Xewp1F3G47f?… Duets: https://open.spotify.com/playlist/5SSQoeApceLBA0OOVzxzyJ?… Only the best flows (hip hop): https://open.spotify.com/playlist/22I4GjVltGVPVEvKgJDb5g?… Date with destiny: https://open.spotify.com/playlist/1LdFO0E9xTqx4m57Yvu7XX?… Beautiful, spacious (good for tripping): https://open.spotify.com/playlist/7KU9zPEP52EBSHeR7fko80?… Chill flourishing (good for parties): https://open.spotify.com/playlist/0mjQkyH8hOWDXTAz9P7lyM?…

Chubby♨️✓@kimmonismus

How can a game from only 30 developers be so good and outperform giant studios like Ubisoft so easily and then only cost $50? Expedition 33 game of the year, game of the decade. And it's not even close.

April

Adam D'Angelo✓@adamdangelo

I really enjoyed this conversation with Rick Rubin. We talked about the early days of Facebook, Quora, Poe, AGI, user-generated content, and vibe coding. Link in next tweet. *[quoted tweet unavailable — [original](https://twitter.com/RickRubin/status/1917602558434238490)]*

Dwarkesh Patel✓@dwarkesh_sp

I know I'm late to the party, but @deanwball 's essays are excellent, and have made me much more skeptical about the kinds of AI regulations I was previously somewhat sympathetic to.

Joe Carlsmith✓@jkcarlsmith

I recently gave a talk at Anthropic on automating alignment research, covering the content from my (just released) essay on the topic. Video here and on youtube, link to transcript in thread.

Joe Carlsmith@jkcarlsmith2025-04-30
Automating alignment research is crucially important to AI safety. I wrote an essay about the ways we might fail, and about what it takes to succeed. Link and summary in thread.

Freeman Jiang✓@freemanjiangg

No speaker? No problem. I built Beatsync — an open-source web audio player for high-precision, multi-device playback. With millisecond accuracy, turn any group of devices into a full surround sound system.

Daniel Litt✓@littmath

I need a browser extension that tags everyone on here who bought into LK-99 hype.

Parker Conley@parconley

Focusmate transformed my social life. Here's my advice after two years of heavy use.

Nabeel S. Qureshi✓@nabeelqu

Best movies of the 2020s so far are Memoria (2021) and About Dry Grasses (2023). By a long way.

Sam Bowman@sleepinyourhat

More in my new essay on the Anthropic alignment blog here: https://alignment.anthropic.com/2025/bumpers/ (Putting up Bumpers)

near✓@nearcyan

probably one of my favorite 'favorites' pages I have read too!: https://niplav.site/favorites.html (niplav) it's always the accounts with 3 digit followers (sometimes 4) with the best content. funny isn't it

Dwarkesh Patel✓@dwarkesh_sp

New blog post: Questions about the future of AI A 6,000-word clusterfuck of considerations about economics, history, training, investment, and more. Thread of select questions below:

owen cyclops✓@owenbroadcast