Freeman Jiang
95 followers · 56 following · 4285 views
on the atlas — 308
- Manifesto - TypeSafe AI1 savers
- Introducing System One Models and Jev - TypeSafe AI Blog16 savers
- Welcome to Effect1 savers
- TLDR AI1 savers
- Cash Management Account | Investing | Fidelity1 savers
- Open Yap 1K: 1,000 hours of full-duplex natural conversation, free for commercial use2 savers
- Jig3 savers
- Intuition on ISA Pricing1 savers
- GPU World5 savers
- laurie, play chords from your laptop5 savers
- Mini Blog Post 3: Become a person who Actually Does Things — Neel Nanda45 savers
- Yishan on X: "I really didn't expect this to get as much circulation as it did, but then again looking back, it's always the least-legible things that people get crazy about. I'll explain in more detail, and then maybe we can have some more informed arguments. First, the quoted post is not" / X1 savers
- Rich Friend, Poor Friend – Jenneral HQ 🌠4 savers
- Idiolect2 savers
- Built in Shenzhen2 savers
- Visting Shenzhen - Jake Robert Read1 savers
- The /grill-me Skill1 savers
- Optimal Highway Policy1 savers
- Prime Agent: A self-improving RLM agent1 savers
- Guardian Angels: LLM Personalization for Productivity and Security · Gwern.net25 savers
- Xenia_(Greek)3 savers
- Pangram 4 Technical Report2 savers
- Silicon-based fluid on X: "Liang Wenfeng Investor Conference (Full Text)" / X1 savers
- VoidZero | The Javascript Tooling company1 savers
- How Shipping Container Disasters Became a Scientific Goldmine | The MIT Press Reader2 savers
- Command Line Interface Guidelines14 savers
- Why China got rich and India didn't - David Oks17 savers
- How to Think for Yourself10 savers
- Rewriting Bun in Rust | Bun Blog12 savers
- Childhoods of exceptional people - by Henrik Karlsson69 savers
- Delphi4 savers
- Declaration of Independence: A Transcription | National Archives2 savers
- Shiki3 savers
- Daytona - Secure Infrastructure for Running AI-Generated Code1 savers
- How to be More Agentic - by Cate Hall - Useful Fictions24 savers
- Red Pill/Blue Pill — The most important Poll in the History of the internet2 savers
- Reflexivity (social theory)3 savers
- Reflexivity — Part 1 - guy1 savers
- God help me am I about to change my view on AI-assisted writing?3 savers
- Ideas for Startups4 savers
- Government in the Sunshine Act1 savers
- Averageness2 savers
- 14. What Motivates Us: Sex - YouTube1 savers
- How to Earn a Billion Dollars11 savers
- How funerals keep Africa poor - David Oks7 savers
- Looking for Alice - by Henrik Karlsson - Escaping Flatland90 savers
- SMPL1 savers
- Life is a Picture, But You Live in a Pixel — Wait But Why19 savers
- Qian Xuesen5 savers
- School Is Not Enough25 savers
- Some interesting open problems in technology | Locklin on science6 savers
- The Old World Is Dying: Advice for 2026 graduates42 savers
- How China's Unitree Will Dominate Global Robotics11 savers
- Claude Fable 5 and Claude Mythos 5 \ Anthropic4 savers
- Alberto Fujimori1 savers
- alphaXiv8 savers
- Migadu Email1 savers
- When AI builds itself \ Anthropic34 savers
- In defense of simple architectures1 savers
- what game are you playing? - by Aadil2 savers
- The mistake of conflating intelligence and power4 savers
- Fidget Camp - June 21-29, 20256 savers
- What is Technology? — Letters to a Young Technologist9 savers
- Lecture 3: The Western Zhou - YouTube1 savers
- I Found the Source of the Amazon River - YouTube1 savers
- Because China4 savers
- AI Populism's Warning Shots - by Jasmine Sun5 savers
- Chien-Shiung Wu2 savers
- Project Glasswing: Securing critical software for the AI era \ Anthropic13 savers
- PostTrainBench4 savers
- Let's reproduce GPT-2 (124M) - YouTube3 savers
- Vibe Island - Dynamic Island for Your AI Agents1 savers
- Gemma 4 — Google DeepMind1 savers
- Illenium - Fractures Guitar Tabs - YouTube1 savers
- What's going on here, with this human? - Graham Duncan Blog49 savers
- Cancer - Sytse.com6 savers
- Ship types, not docs1 savers
- Network School2 savers
- How Long Contexts Fail | Drew Breunig2 savers
- Frederick Douglass - Wikipedia2 savers
- List of proverbial phrases - Wikipedia2 savers
- Measuring the performance of our models on real-world tasks | OpenAI5 savers
- AmpMe plans to kill Bluetooth speakers by syncing music between smartphones | VentureBeat2 savers
- Claude Code Best Practices \ Anthropic7 savers
- Fuck willpower - by Cate Hall - Useful Fictions6 savers
- Earnestness10 savers
- Confucius Lives Next Door4 savers
- mcp.run - the App Store for MCP Servlets: portable & secure code for AI Apps and Agents.3 savers
- Three words of advice every student should hear4 savers
- Beware the Metagame5 savers
- Pivoting is an insecurity - Nader Khalil4 savers
- Curius / Onboarding2621 savers
- How To Be Successful96 savers
- Staring into the abyss as a core life skill85 savers
- How to Do Great Work80 savers
- Child’s Play, by Sam Kriss75 savers
- the friendship theory of everything69 savers
- How to Do Great Work67 savers
- Everything that turned out well in my life followed the same design process64 savers
- Thirty Observations at Thirty62 savers
highlights — 370
Do you hope that ChatGPT and Claude will just “quietly” take over your life for you? Honestly—that seems like a bad idea to me. The chatbot personas are deeply misaligned with you, and aligned with their owners; and the economic incentives are to farm you with ads and subscriptions, while racing not to amplify you but to replace you.
Guardian Angels: LLM Personalization for Productivity and Security · Gwern.netIf we try to monopolize the benefits, we will inevitably be abandoned by history. I think this is an objective law, a historical perspective.
Silicon-based fluid on X: "Liang Wenfeng Investor Conference (Full Text)" / XOr, said differently, villains are merely rational while superheroes are, appropriately, superrational. That’s why villains “low-key make sense”. Heroes act in a way that is, from the outside, indistinguishable from the way that people who don’t understand that Red is the dominant option act. But they do so to save them.
Red Pill/Blue Pill — The most important Poll in the History of the internetThe second trial tests a UBI of $1000 a month over three years with a sample size of 1,000 treated participants and 2,000 controls in the United States. They find precise null effects on health, career prospects, and investments in education.
The Causal Effect of Income is (Often) ZeroThe codes for all containers worldwide are distributed by the Bureau International des Containers (BIC)
How Shipping Container Disasters Became a Scientific Goldmine | The MIT Press ReaderAll this meant that between 1949 and 1976, the Chinese state effectively destroyed traditional Chinese society: the social landscape of the old China, with all its complexity and custom, was simplified and smoothed away.
Why China got rich and India didn't - David OksChinese women went from “virtual complete illiteracy” to a literacy rate of about 50 percent during the same period. The progress in health was similarly rapid: child mortality fell by 80 percent between the early 1950s and the late 1970s.
Why China got rich and India didn't - David OksWhatever cultural advantage that China might enjoy over India, it wasn’t operative a hundred years ago.
Why China got rich and India didn't - David OksIn most kinds of work — to be an administrator, for example — all you need is the first half. All you need is to be right. It's not essential that everyone else be wrong.
How to Think for YourselfA language-independent test suite with a million assertions, adversarial code review and when something does go wrong, fixing the process that generates the code instead of hand-fixing the code.
Rewriting Bun in Rust | Bun BlogHow do you review a PR with +1 million lines added? How do you start to build the confidence needed to responsibly merge large quantities of LLM-authored code?
Rewriting Bun in Rust | Bun BlogNever attribute to malice what can be explained by math.
Ideas for StartupsExponential growth is like magic. It generates outcomes that seem impossible. And that's why some politicians distrust it. They don't understand the math of exponential growth, so when they see people becoming what seems to them impossibly rich, they assume they must have cheated.
How to Earn a Billion Dollarsyou want to communicate as much as possible, too, so you can filter people who wouldn’t fit you anyway
Looking for Alice - by Henrik Karlsson - Escaping Flatlandhow major life traumas affect people suggests that if it happened over three months ago, with only a few exceptions, it has no impact whatsoever on your happiness
Life is a Picture, But You Live in a Pixel — Wait But WhyIt is difficult to blame young adults for thinking that work is fake and meaningless if we prescribe fake and meaningless work for the first two decades of their existence.
School Is Not Enoughwe see perspiration becoming increasingly automated
When AI builds itself \ AnthropicFrom now until forever, most people regulating, protesting, and talking about AI will not be interested in AI per se, but rather how it impacts their preexisting belief systems and political agendas
AI Populism's Warning Shots - by Jasmine SunWu nearly visited in 1956, but decided to stay in the US to finish her famous experiment while her husband visited China. By the time she returned, her uncle and brother had perished during the Cultural Revolution, and the tombs of her parents had been destroyed. She was greeted by Zhou Enlai, who personally apologized for the destruction of the tombs. After this, she returned to mainland China and Taiwan several times
Chien-Shiung WuThere is no way that Joe Biden, or any other American political leader, could extract that kind of dedication and loyalty from the American people today. It simply isn’t possible. Our political disagreements are so fundamental and the fabric of our communal life so tattered that such an expenditure of effort isn’t comprehensible.
World War III is Not Possible › American GreatnessRemember the "doorknob phenomenon": people say the most interesting stuff when your hand is on the doorknob — so be ready to listen to (or receive) their “one more thing”
Interviewing tipsThe Goal Is Not to Have a Job
A Motorcycle for the Mindthe harder the ideas you're talking about, the less you can afford to let language get in the way.
Write Like You TalkBut just imagine calling Picasso "the mercurial Spaniard" when talking to a friend. Even one sentence of this would raise eyebrows in conversation. And yet people write whole books of it.
Write Like You Talk"Science concerns itself with what is, whereas technology concerns itself with what is to be" - Henryk Skolimowski, 1966
What is Technology? — Letters to a Young TechnologistThe most impactful thing I did that first year was to multiply the compute of an experiment a thousandfold.
2025 letter | Zhengdongwe present ample numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions -- and that sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation
[1609.04836] On Large-Batch Training for Deep Learning: Generalization Gap and Sharp MinimaThat’s a $167B+ increase in valuation over 90 days, or roughly $1.85B per day.
🧭 The Brief: Anthropic’s $10B Raise + Zhipu’s Historic IPOWe’re all playing centaur chess now.
The Case Against Doomerism - by Nikunj KothariThere is no safe lower limit for alcohol intake, and even small amounts increase the risk of premature mortality.
Running from Death: Can Fitness Outpace Alcohol’s Harm? Changes in Alcohol Intake, Fitness and All-Cause Mortality in the HUNT Study, Norway | Sports MedicineYou can think of a spectrogram as a bunch of FFTs stacked on top of each other.
Understanding the Mel Spectrogram | by Leland Roberts | Analytics Vidhya | MediumLeague of Legends rating
Career Advice That Doesn’t Suck“If you want me to get on my hands and knees, I will,” said one local resident, addressing city staff from the audience. “I am begging you, do not allow this development to happen. It will destroy our neighbourhood. It will destroy everything,” she said.
High Park locals upset over proposed 41-storey development - TorontoToday.camission-driven founders take note of this and seriously discount that firms reputation
“Build What’s Fundable” - by Kyle Harrison - Investing 101build something worth building
“Build What’s Fundable” - by Kyle Harrison - Investing 101Despite allocating 5x the capital, we don’t have 5x the outcomes. But we’re trying desperately to get more through that funnel
“Build What’s Fundable” - by Kyle Harrison - Investing 101For example, George W. Bush administration increased federal funding for abstinence-only programs (the only type funded even under the previous Clinton Administration), while the Obama administration reversed the emphasis and provided more federal dollars for comprehensive sex education.[10] President Donald Trump cut grants to comprehensive sex education[11] and proposed increased funding for abstinence-only education.[12] Increased funding continued through the Joe Biden administration
Abstinence-only sex education - Wikipedia. The FL2 Rust code that makes the check and was the source of the unhandled error is shown below:
Cloudflare outage on November 18, 2025and to the Internet in general
Cloudflare outage on November 18, 2025Four empires collapsed due to the war
Aftermath of World War I - WikipediaHarnessing collective intellect, facilitated by interactive computers, became his life's mission at a time when computers were viewed as number crunching tools
Douglas EngelbartEngelbart's career was inspired in December 1950 when he was engaged to be married and realized he had no career goals other than "a steady job, getting married and living happily ever after".[15] Over several months he reasoned that: he would focus his career on making the world a better place[16] any serious effort to make the world better would require some kind of organized effort that harnessed the collective human intellect of all people to contribute to effective solutions.[17] if you could dramatically improve how we do that, you'd be boosting every effort on the planet to solve import…
Douglas Engelbartpilots had already ingrained the "standard" layout. Thus, the conceptually good idea had unintended results
Human–computer interactionContact between CNSA and NASA is generally limited by the so-called Wolf Amendment, which prevents most bilateral interaction between NASA and Chinese state entities.
China reached out to NASA to avoid a potential satellite collision in 1st-of-its-kind space cooperation | SpaceTo our knowledge, this is the first time that model ossification8 has been observed in robotics
Generalist - Blogthe goal is a nice future, not just a technically advanced future.
Nintil - Cozy futurismNEO is designed for gentle and safe interactions, making it perfect for people and the home.
NEO Home RobotAlignment. There is a super interesting result that in order to have aligned models, you must pretrain on both aligned and unaligned data (“When Bad Data Leads to Good Models”; by Kenneth Li et al. 2025), because pretraining then learns a linearly separable direction between the two. If you drop all the unaligned data, this leads to the model not having a strong understanding of what unaligned data is, and why it is bad
The Only Important Technology Is The Internet — Kevin LuI think a lot of people want to be but they don’t want to do
effort - by Ava - bookbear expressAll you need to do is do things intentionally, get got on purpose, verify that the ladder is good for you, and thus reap the benefits of actual choice.
What's stopping you?