Thu Than

on the atlas · 88

• Persimmon | humans&8 savers
• Learning Tracks — Technically10 savers
• Animations on the web12 savers
• MAKING SOFTWARE62 savers
• the intelligence of desire - by maja - velvet noise3 savers
• Planning for Preservation in the Age of AI — LessWrong2 savers
• Introducing FrontierCode | Cognition2 savers
• The Old World Is Dying: Advice for 2026 graduates43 savers
• E.W. Dijkstra Archive: On the cruelty of really teaching computing science (EWD 1036)2 savers
• Better MoE model inference with warp decode · Cursor7 savers
• Measuring AI Ability to Complete Long Tasks - METR7 savers
• Stay inconvenienced - sémaphore1 savers
• Developer quickstart - Notion Docs1 savers
• Build with Notion’s Developer Platform – Notion2 savers
• Agent Evaluation: A Detailed Guide2 savers
• escaping flatland: career advice for CS undergrads43 savers
• Jason Spielman1 savers
• You and Your Big Heart Will Win | Rosanne Liu1 savers
• I Moved Back to a Second-Tier City and Finally Started Living22 savers
• Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab23 savers
• For a Student Who Used AI to Write a Paper by Joseph Fasano - Poems | Academy of American Poets2 savers
• MIT Visualization Group1 savers
• MIT HCI1 savers
• OpenCode | The open source AI coding agent1 savers
• Opinion | Marshall McLuhan Was Right About Claude, Too - The New York Times4 savers
• Our operating values1 savers
• Contra Labs - Powered by Contra5 savers
• Greg - Sam Altman16 savers
• Attention is all we have - David Bessis6 savers
• METR3 savers
• Readings in Database Systems, 5th Edition3 savers
• Your Name in Landsat 🛰️9 savers
• About—Asterisk1 savers
• Pete Docter - Inside the Creative Community: The Power and Process of Animated Film - 09/28/15 - YouTube1 savers
• Alan Kays Reading List1 savers
• Before Growth - Sam Altman2 savers
• The days are long but the decades are short - Sam Altman73 savers
• Productivity - Sam Altman58 savers
• Do Thing, Do One Thing49 savers
• Andy Matuschak45 savers
• Make Something Wonderful | Steve Jobs39 savers
• Ink & Switch30 savers
• On-Policy Distillation - Thinking Machines Lab30 savers
• Works in Progress29 savers
• Training the Idea Muscle | Alexandria25 savers
• How to make something great23 savers
• Butterick’s Practical Typography21 savers
• Circuit Tracing: Revealing Computational Graphs in Language Models20 savers
• neal.fun20 savers
• James Somers19 savers
• Developing creative identity18 savers
• How to build meaningful relationships after college18 savers
• How to Work Hard18 savers
• The Friendship That Made Google Huge17 savers
• Not a Number17 savers
• The Anthropic Hive Mind. As you’ve probably noticed, something… | by Steve Yegge | Feb, 2026 | Medium17 savers
• First Principles: The Building Blocks of True Knowledge - Farnam Street14 savers
• research as leisure activity - by Celine Nguyen13 savers
• Laws of UX13 savers
• Claude Mythos Preview \ red.anthropic.com13 savers
• AI Design Field Guide11 savers
• The Law of Leaky Abstractions – Joel on Software11 savers
• Inside Notion - Colossus10 savers
• howard rheingold's | tools for thought10 savers
• Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X10 savers
• How Might We Learn?9 savers
• Malleable software: Restoring user agency in a world of locked-down apps8 savers
• Ray Tracing in One Weekend8 savers
• home | p5.js7 savers
• Awwwards - Website Awards - Best Web Design Trends7 savers
• The data model behind Notion's flexibility6 savers
• Don't fall into the anti-AI hype - <antirez>6 savers
• Communicating with Interactive Articles6 savers
• Building complex reasoning skills online through open-ended activities5 savers
• Joel on Software4 savers
• Against vibes - by Andrew Wu - fanciful filigree flights4 savers
• Notes on Roadtrips - The Browser Company4 savers
• Typography in ten minutes | Butterick’s Practical Typography4 savers
• Why React Re-Renders3 savers
• The Brand Age3 savers
• The Tenacious Brain: How the Anterior Mid-Cingulate Contributes to Achieving Goals - PMC2 savers
• The Lessons That Come Out of Tech’s Boom-and-Bust Cycles2 savers
• Making Internet Things, part 3: Storytelling2 savers
• The world is poorly designed. But copying nature helps. - YouTube2 savers
• Rosetta Code2 savers
• Stanford University CS231n: Deep Learning for Computer Vision2 savers
• You Still Need to Think2 savers
• Tag: Engineering | Figma Blog2 savers

highlights · 280

• The act of loving, even without outcome, still expands the soul and reveals what we’re capable of carrying.

  the intelligence of desire - by maja - velvet noise

• To want is to orient yourself toward it, to shape your perception around it. Wanting becomes a way of knowing, a way of becoming.

  the intelligence of desire - by maja - velvet noise

• Because loving, even secretly or stupidly, can’t help but crack something open in us. It shakes us from mundanity, adds colour to the edges of things. Even if it doesn’t lead to love, or leads straight into heartbreak, it’s still proof that the heart is pulsating and that something in you still wants.

  the intelligence of desire - by maja - velvet noise

• Often, they reflect the qualities we long to cultivate in ourselves - confidence, charisma, creativity, spiritual clarity. When we fall for someone, we’re drawn to the version of ourselves that wakes up in their light as much as we are drawn to them as individuals.

  the intelligence of desire - by maja - velvet noise

• People often talk about crushes as aspirational mirrors.

  the intelligence of desire - by maja - velvet noise

• Yes, he explained, improving the government would help with AI, but his reasoning went deeper than that. “If we succeed at slowing AI down, we need to actually solve climate change, the housing crisis, and everything else. I’m looking ahead towards the paths where we survive the short term to make sure we survive the long term.”

  Planning for Preservation in the Age of AI — LessWrong

• We instead take in audio signals as dMel (Bai, et al. 2024) and transform it via a light-weighted embedding layer. Images are split into 40x40 patches which are encoded by an hMLP (Touvron et al. 2022). For the audio decoder we use a flow head (Lipman at al. 2022). All components are co-trained from scratch together with the transformer.

  Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab

• a system with minimal pre-processing

  Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab

• The interaction model works with micro-turns continuously interleaving the processing of 200ms worth of input and generation of 200ms worth of output. Rather than consuming a complete user-turn and generating a complete response, both input and output tokens are treated as streams. Working with 200ms chunks of these streams enables near real-time concurrency of multiple input and output modalities.

  Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab

• When a task requires deeper reasoning than can be produced instantaneously, the interaction model delegates to a background model that runs asynchronously.

  Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab

• concurrently search, browse the web, or generate UI—weaving back results into the conversation as needed.

  Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab

• Picture trying to resolve a crucial disagreement over email rather than in person.

  Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab

• However, humans increasingly get pushed out not because the work doesn’t need them, but because the interface has no room for them.

  Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab

• Only humans can carry the cost of loving and being loved, and that is what will keep us, in the end, irreplaceable.

  Stay inconvenienced - sémaphore

• It shows that the need to be heard can be met by anything that listens, but the need to matter can only be met by someone who risks being changed

  Stay inconvenienced - sémaphore

• And that willingness to be inconvenienced, reshaped, or burdened is what makes human love possible.

  Stay inconvenienced - sémaphore

• all I want is for people to at least make an informed decision about what they want to do with their lives. to step out of the plane they’re operating in ‒ to leave flatland ‒ and to realise how the things they do could look very different in different planes.

  escaping flatland: career advice for CS undergrads

• makes it seem like doing an internship is the only socially-acceptable way to spend one’s summer

  escaping flatland: career advice for CS undergrads

• I’m always pleasantly surprised by how willing they are to engage with this. but more often than not, they say something like “of course you need to do X, that’s just … the thing we’re meant to do at this stage, right?” and when I poke at this a bit more, they often realise they don’t actually have an answer for ‘what motivates them’ that they themselves feel satisfied with.

  escaping flatland: career advice for CS undergrads

• And I eagerly welcome this world where the ones with the bigger hearts, not the ones with slicker tongues, will rise to the forefront.

  You and Your Big Heart Will Win | Rosanne Liu

• This realization is rather consoling to someone like me. Because the muscle skills are more correlated with privilege (whether you happened to be raised and trained in an environment that emphasizes those skills), than the heart ones. And I would love to live in a world where the guiding principle is: “If you want to get better at something, you will get better. Because the very ‘want’ is all that matters.”

  You and Your Big Heart Will Win | Rosanne Liu

• The endearing type is one that while their content, language and delivery may not be perfect, you can clearly see their heart behind it — that they really take this chance of communication seriously, in that they put the audience first, before themselves. Like any good science, a good talk is where you take yourself and your ego out of it.

  You and Your Big Heart Will Win | Rosanne Liu

• All that the note says is: at some specific snapshot in some people’s lives, some were really good at this specific thing. It’s a time bound label, which means after passing that labeling timestamp, it shouldn’t mean anything. I see everyone, me included, to be on a path of trying to get better; where we eventually end on that path is merely a function of the time remaining on our hands, and the amount of energy we put behind it.

  You and Your Big Heart Will Win | Rosanne Liu

• But here’s what nobody tells you about going home: everyone knows everything. In Shenzhen, I was anonymous. I could wear whatever I wanted, eat alone at a restaurant without anyone wondering if something was wrong, live without performing.

  I Moved Back to a Second-Tier City and Finally Started Living

• On weekends, my parents take her for a few hours and I actually read a book. A whole book. I’d forgotten what that felt like.

  I Moved Back to a Second-Tier City and Finally Started Living

• But what are you trying to be free of? The living? The miraculous task of it?

  For a Student Who Used AI to Write a Paper by Joseph Fasano - Poems | Academy of American Poets

• Being kind is not the same as being nice

  Our operating values

• Is this what our customers say they want, or what they actually want?

  Our operating values

• Build a habit of urgency. In a world of unpredictable outcomes, what we can control is executing with speed in mind. As a company, we always ask ourselves: how can we move faster?

  Our operating values

• No theory of cognitive inequality would be complete without a discussion of fear. Why did we evolve cognitive inhibition, this bizarre response to intellectual difficulty, this visceral panic, this brain fog that prevents us—among other things—from even engaging with mathematical abstractions?

  Attention is all we have - David Bessis

• For me, the hardest part wasn’t to identify the right mental habits—I had a clear enough idea from my early 20s—but to cobble together the courage and resilience to practice them with more discipline.

  Attention is all we have - David Bessis

• ability to suppress intellectual fear.

  Attention is all we have - David Bessis

• Because cognitive activity takes place under a veil of invisibility, people underestimate the magnitude of the practice gap.

  Attention is all we have - David Bessis

• the method was really about attention control

  Attention is all we have - David Bessis

• progress wasn’t only about mathematics itself, but also metacognition and emotion control.

  Attention is all we have - David Bessis

• Leonardo da Vinci, William Shakespeare, Nikola Tesla, or John von Neumann were extraordinarily intelligent.

  Attention is all we have - David Bessis

• Should I do this absolutely-crazy spontaneous (but non-risky) trip? (1). Heck Yes. Otherwise you'd always be left wondering. The beauty is that the work fills the time that you grant it, and the opportunity cost is therefore never high.

  Do Thing, Do One Thing

• Should I go to this social event, even though I'm feeling tired? (1) says yes. I don't trust my tiredness — I'm wary it is a low set point of sadness that my mind is overcorrecting towards — so I overrule it by biasing towards action.

  Do Thing, Do One Thing

• A product capable of turning a decade of data and context into a competitive strength. A default to throw away finished work, not as a painful exception, but as a practice to get better. A culture of trust deep enough that a vocal dissenter could become the face of the launch.

  Inside Notion - Colossus

• An identity so strong and specific that it could absorb enormous change without losing coherence.

  Inside Notion - Colossus

• Founders who felt religious conviction in their bones and had the credibility to inspire it in others.

  Inside Notion - Colossus

• “We don’t punish skepticism,” an engineer explains. “We punish pessimism.”

  Inside Notion - Colossus

• As we were writing this, we kept referring to Notinos as the happy pointy people—whistling while they work, but piercingly intense.

  Inside Notion - Colossus

• The whole thing came together in under two weeks. A small handful of us were jamming on the things that make Notion special and someone, in passing, said the line: “think together.”

  Inside Notion - Colossus

• the sciences which do not aim at giving pleasure or at the necessities of life were discovered, and first in the places where men first began to have leisure. This is why the mathematical arts were founded in Egypt; for there the priestly caste was allowed to be at leisure.

  The Internet Classics Archive | Metaphysics by Aristotle

• Again, we do not regard any of the senses as Wisdom; yet surely these give the most authoritative knowledge of particulars. But they do not tell us the 'why' of anything-e.g. why fire is hot; they only say that it is hot.

  The Internet Classics Archive | Metaphysics by Aristotle

• And in general it is a sign of the man who knows and of the man who does not know, that the former can teach, and therefore we think art more truly knowledge than experience is; for artists can teach, and men of mere experience cannot.

  The Internet Classics Archive | Metaphysics by Aristotle

• And experience seems pretty much like science and art, but really science and art come to men through experience

  The Internet Classics Archive | Metaphysics by Aristotle

• but the human race lives also by art and reasonings

  The Internet Classics Archive | Metaphysics by Aristotle

• ALL men by nature desire to know.

  The Internet Classics Archive | Metaphysics by Aristotle