bio,phy,che retweeted
8/20 "The Early Universe" by Kolb and Turner 🌌
A detailed introduction to the physics of the early universe, exploring topics like cosmic inflation, dark matter, and dark energy. Ideal for advanced undergrads and grad students.
amzn.to/447rM2Q
bio,phy,che retweeted
This is a fantastic guide to Kant’s Critique of Pure Reason. It also teaches you some of the basic philosophical concepts Kant relied on when writing the Critique, which new readers may not be familiar with. It has diagrams that make the more abstract ideas easier to follow.
bio,phy,che retweeted
Percy Bridgman's 1927 work appreciated the constrictions on imaginative thought characterizing the strictly empiricist "operational point of view," which precludes use of concepts for which "we cannot give an adequate account in terms of operations."
bio,phy,che retweeted
I worked on electronics for years before I went to college. In school, I had a hard time with transistors. They just didn't work intuitively until I found an old book from 1959 on eBay.
The ABCs of Transistors. If you have trouble understanding transistors, this book will fix that. It's great. I worked on electronics for years before college but struggled with transistors until I found an old book from 1959, The ABCs of Transistors, on eBay. It clarifies transistors and you may find it helpful.
worldradiohistory.com/BOOKSH…
bio,phy,che retweeted
1次元粒子の量子力学は、理学部物理学科では2,3年生で習う基礎的なテーマです。しかし細かいところを誤解したままで量子力学を修了してしまった方が、一般人の中にも、そして物理教員の中にもいらっしゃるようです。前世紀の古い教科書には、「正準量子化」を中心にして説明がなされているために起きている誤解も多いです。古典力学では、点粒子の位置xと運動量pは、同時にいくらでも精密に測れるとされてきました。そして点粒子の状態は、その値(x,p)で決まると定義されています。位置を横軸にとり、運動量を縦軸にとった「相空間」が、古典力学における状態空間でした。そして昔の教育では、この古典力学を学んだ後に、位置と運動量を演算子化して、正準交換関係を満たすとするのが、正準量子化です。この演算子が作用するのが波動関数であり、その波動関数が点粒子の量子的な状態を定めます。波動関数は、数学の関数論でいうところの「ヒルベルト空間」の元になっていて、波動関数の絶対値の2乗が可積分であるという、規格化条件を満たします。
このような正準量子化の説明だと、波動関数よりも、演算子となった物理量のほうが重要だという印象が、初学者に植えつけられることも多いようです。しかし量子力学では、物理量よりも、実験で観測される確率分布、そしてその確率分布が定める波動関数のほうが、本質的なのです。この点を理解しないと、いろいろおかしな間違いをしてしまいます。たとえば「規格化可能だけど、位置の分散(またはその平方根としての標準偏差が発散する波動関数は、物理的ではない」という主張をする人も出てきます。しかし実際には、この主張は間違いです。🧵
bio,phy,che retweeted
I am listing in sequence-
1. Calculus on Manifold (I listed books earlier)
2. ODE by Arnold
3. Manifold, Tensor Analysis & Applications —Jerrold Marsden.
4. Mathematical Methods of classical mechanics - V.I. Arnold (finest ever)
5. Foundation of Mechanics - Abraham & Marsden
Replying to @_hannahrae
Dynamics on manifolds but not interested in specializing to an application. Looking to study like a Math Grad student.
bio,phy,che retweeted
Complexity dissolves when we map the hidden architecture of state variables.
By chaining partial derivatives, abstract thermodynamic chaos transforms into an orderly tree of measurable quantities like volume and temperature.
Every macroscopic state is an interconnected web.
bio,phy,che retweeted
GPT-6 아스트라(Astra) 뜨자마자 로보틱스 판 연구자들 멘붕 와서 프로젝트 10개 넘게 터졌다는 소식 ㅋㅋㅋ
오픈AI가 아스트라 내놓자마자 로봇 스타트업 바닥에서 "우리 이제 다 굶어 죽는 거냐"며 프로젝트 접고 빤스런 치는 연구자들이 속출함. 근데 감정적인 공포 걷어내고 뜯어보면 대가리 깨진 애들과 새로 꿀빨 기회가 명확히 갈린다는 분석임.
로보틱스 3단계 계층 구조로 보면 이해가 빠름:
• S2 (Task Planning): "뭘 할지" 계획 짜는 뇌 (기존 LLM 영역)
• S1 (Policy): "어떻게 움직일지" 정하는 신경계 (로봇 전용 학습 모델)
• S0 (Physical Control): 모터 돌리고 토크/힘 주는 물리 제어
원래는 "LLM 따위가 계획이나 짜지 물리 액션을 어떻게 하냐"며 로봇쟁이들이 코웃음 쳤는데, 아스트라 같은 파운데이션 모델이 카메라 영상 직관하고 바로 액션 때려버리니까 S1 영역까지 침범당해서 집단 공황 온 거임.
소프트웨어 먹던 놈들이 피지컬 AI까지 숟가락 얹으면서 마이크로소프트 $MSFT 같은 빅테크 파운데이션 성장 곡선에 다 갈려 나갈 거라는 공포가 핵심임.
팩트로 짚어보는 LLM의 현실:
• 잘하는 것: 물건 집어서 옮기기(Pick & Place) 같은 기본 조작. 심지어 웬만한 전문 로봇 모델보다 벤치마크 더 잘 뽑음 ㅋㅋㅋ
• 여전히 젬병인 것: 정밀 삽입, 손가락 꼼지락거리는 정교한 조작(Dexterous manipulation), 힘 조절하고 마찰력 버티는 접촉 작업(Contact-rich).
• 한마디로 계획 짜고 대충 팔 휘두르는 건 빅테크가 먹겠지만, 정밀하게 피지컬 구현하는 건 아직 물리적 한계에 부딪혀 있음.
로봇 스타트업이 살아남을 해자(Moats)와 기회:
1. 가성비 특화 모델: 아스트라 같은 거 굴리면 토큰값에 허리 휨. 실제로 Dream-1 같은 특화 모델은 아스트라 대비 60배 저렴하게 굴러감.
2. Agentic System 1: 파운데이션 모델(S2)이 명령 내리면 바닥에서 기가 막히게 실행하는 전용 S1 정책 개발.
3. 촉각(Tactile) 및 접촉 데이터: 인터넷에 널린 게 영상/이미지지만 '손맛'이나 마찰력 데이터는 오픈소스에 없음. 먼저 모으는 놈이 임자임.
4. 역운동학(IK solver)과 제로샷: 하드웨어에 최적화된 기구학 솔버 잘 짜놓은 하드웨어 기업이 유리함.
5. 배포 및 평가 인프라: 연구실에서 한두 번 성공하는 건 의미 없고, 현장에서 수천 번 돌려도 안 터지는 안전성 검증 툴 수요가 폭증 중임.
결론:
아스트라가 어설픈 로봇 프로젝트들 대가리 깬 건 팩트인데, 정밀 물리 조작이랑 인프라 쪽은 이제 막 장터 열린 거라 아직 승자 안 정해짐.
매수/매도 추천아님
bio,phy,che retweeted
量子力学の本で、調和振動子とか角運動量の話をする際特殊関数とか解析の話ばっかりで代数的な話が乏しいやつあんま好きじゃない。正直、具体的な微分方程式がどうこうよりも代数的な構造に着目するほうが後々大事な気がしてる
bio,phy,che retweeted
타고나기를 도무지 관용이라곤 없는 히틀러 였지만, 이상하게도 인간으로서 한 가지 조건
- 타인의 도덕성-에 관해서는 관용을 보였다.
제3제국사.
한국에서 현재 이재명이 하는짓.
bio,phy,che retweeted
Most people learning AI in 2026 are drowning in tutorials and finishing nothing.
10 books will take you further than 1,000 hours of YouTube.
The exact order they should be read to level up in AI Engineering:
bio,phy,che retweeted
Incredible piece of engineering.
Boston Dynamics has optimized for simulatability and ease of manufacturing over exact human-likeness.
New four-finger hand unlock a huge range of dexterous behavior. Now it's AI's turn to explore the hardware limits.
bio,phy,che retweeted
A twist is a rotation whose angle grows along its own axis: Tₓ(s) turns the point (x, y, z) about the x axis through the angle sx.
The face x = 0 stays put, each slice x = c turns rigidly through sc, and edges along the axis bend into helices. With s = cos θ and side 2, the far face swings through ±2 radians, about ±115°.
The map is not linear, yet its Jacobian has determinant 1 everywhere, so the twist keeps volume.
bio,phy,che retweeted
Explore the geometric structures behind mechanics, fluids and dynamical systems. From Lie groups and momentum maps to Euler–Poincaré theory and MHD, this advanced text bridges mathematics and physics with clarity and depth. Get your copy today.
bio,phy,che retweeted
1) principles of mathematical analysis by walter rudin
2) real analysis by gerald b. folland
3) a course in functional analysis by john b. conway
bio,phy,che retweeted
PHYSICS ALL IN ONE:
A little bit of everything, from projectile motion and optics to electricity and radioactivity. So many different things to understand, all under one subject.
The Advanced Calculus (2nd edition) textbook by Gerald B. Folland (2023) is now in @ChapterPal's collection.
The textbook provides a rigorous, analysis-based treatment of multivariable calculus and classical analysis for advanced undergraduate students in mathematics, physical sciences, and engineering. It assumes that readers are already familiar with basic single-variable calculus, standard computational techniques, and elementary linear algebra. The text bridges introductory calculus and higher-level mathematical analysis by establishing a firm theoretical groundwork while maintaining a strong connection to geometric intuition and physical applications.
The material opens with the topology and geometry of Euclidean spaces, introducing limits, continuity, completeness, compactness, and connectedness. From these foundations, the text develops multivariable differential calculus, using linear approximations and the Fréchet derivative as the unifying framework for directional derivatives, the chain rule, Taylor expansions, and extrema classification. A central focus is the local solvability of nonlinear systems through the implicit and inverse mapping theorems, which are applied to analyze coordinate transformations, functional dependence, and the representation of smooth curves and higher-dimensional surfaces embedded in Euclidean space.
Read the book with an AI tutor: chapterpal.com/ebook/22cb5ea…
This is a great continuation to "Active Prelude to Calculus" chapterpal.com/book/f732cfe2… and Active Calculus chapterpal.com/book/9f5af026…
(All books on ChapterPal are free for anyone to read with a free account.)
同感です。非数学科の人が線形代数を学習するのであれば、長谷川浩司先生の「線型代数[改訂版]」をお勧めします。紹介記事はこちらです。
線型代数[改訂版]: 長谷川浩司
ktonegaw.hatenadiary.org/ent…

































