Newstart retweeted
This is huge news for Siri AI (and one of the last things I was working on when I stepped away)
Any app can now offer dynamic skills from their app to Siri so users can take broader actions across their apps
(App Intents instead of MCP)
Big news for our developers! We are shipping a developer preview that dramatically expands the capabilities of all apps to integrate into Siri. So excited to see what you folks do with this.
developer.apple.com/news/?id…
Newstart retweeted
花了一天写完了，也算是给自己的阶段性成果一个交代。
想一想这一年，还真的有一些魔幻。
一年前，我是一个完全不懂开发的设计师。
一年后的今天，我搭了我自己的Agent自动化系统，做出了百万月活的公益产品，还开源了，拿到了9000 star的好成绩，也看到无数的反馈，说帮到了他们。
也在用Agent的方式，去迭代整个公司。 每一天，我都在开发中、在创作中学习。
一年前，我连CI是什么都不知道，半年前，我甚至连CDN是什么都不知道，但是不妨碍我，现在都知道了。
这可能，就是学习的快乐吧。
希望今天这篇内容，能对大家有用。
也愿我们，永远对世界，保持好奇。
Newstart retweeted
We have lots of younger kids (age 5-7) in our 3rd grade math course, who can barely read.
To help them learn math anyways, we created interactive DI tutorials that teaches out loud to them.
To our delight, the tutorials work just as well for older kids. Watch this lesson.
Newstart retweeted
Following the fantastic reaction to our interactive learning experiments, we’ve decided to create an open source skill for anyone to build their own 💪🏗️
It's called the Marble-Builder. You tell it your child’s age and what they’re curious about and it builds a playable experiment for them.
What’s built into the skill:
- 1,837 concepts kids learn through primary school, and the 3,853 links between them
- 1,049 teaching research records on how kids learn best
- It agrees a plan with you, then builds, tests and fixes the experiment itself
You stay in control and can adjust everything. AI is a tool with huge potential to make teaching even better for both parents and teachers. We want anyone with something to teach to be able to bring it to life in an interactive, personalized way.
It’s a first prototype so there is a lot to improve. One big next step is to make the experiments more beautiful visually.
Parents, teachers, homeschoolers: show us what you make. We’ll gather the best experiences into the first open, interactive encyclopedia for kids.
Incredibly exciting times for kids education.
Links to try it below 👇
Newstart retweeted
"Contemporary Mathematics" is a free textbook that introduces fundamental mathematical concepts over nearly 1,900 pages through clear explanations, numerous worked examples and exercises.
The book covers set theory, logic, Venn diagrams, mathematical statements and proofs, De Morgan’s laws, number systems, elementary number theory, prime numbers, rational and irrational numbers, arithmetic sequences, geometric sequences, numeral systems, calculations in different bases, and many other topics, providing a broad overview of foundational mathematics.
It is particularly useful for high school students, and I recommend keeping it if you have children studying mathematics at that level. The book is well structured, accessible, and enjoyable to browse, making it a valuable resource for both learning and reviewing mathematical concepts.
assets.openstax.org/oscms-pr…
Newstart retweeted
In July we open-sourced a dynamic graph of everything a child learns in primary school. 1,590 concepts, 3,221 dependencies.
Now we’re turning every concept into interactive experiments kids can play and learn from.
First one, age 7: heat and water.
Everything that used to be a boring textbook can now be an engaging experience, potentially tailored to every child.
Kids education just got a massive upgrade.
Newstart retweeted
对 AI 能力的认知正在变成一个陡峭的漏斗！绝大多数人几乎没接触过它，接触过的多数人看到的是一个有点笨的 Chat Bot，少数用前沿模型的专业人士正在经历加速的眩晕，而极少数接触内部系统的人认为：AGI 就在 2027 ✨这几种现实同时存在，彼此几乎无法沟通🤪
Karpathy 说这不是技术鸿沟，而是野心鸿沟！过去，能力是稀缺的，一个人有再大的想法，也受限于时间、人手和技能。现在，执行的门槛在迅速塌陷，真正稀缺的变成了“值得去做的大问题”本身。
那些说 AI 对自己没什么用的人，很多时候并没有说错，他们的生活里确实没有一个需要动用这种能力的问题。这不是他们的问题，而是这个时代悄悄换了一道考题：从你能做什么，变成了你想做什么。
往前看，访问权限会逐渐下沉，免费层和内部系统之间的差距会缩小，参差不齐的能力也会从代码和数学扩散到更多领域。但野心这一层不会自动弥合。将来决定一个人、一家公司乃至一个国家能从 AI 里拿到多少的，可能不是谁先拿到最强的模型，而是谁手里攒着最多值得解决的难题。漏斗的顶端，最终属于那些敢于提出大问题的人🤔
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
Newstart retweeted
For the past several months, my boys have been using a learning app I created. Today I'm making it public!
It's called Unspool.
Here's the idea → learning is most fun when it doesn't feel like "work."
There's plenty of strict curriculum out there. But I wanted a place where kids could follow their curiosity without the pressure of tests. So I built one.
Here's what you get:
1. 1500+ built in lessons (that are quite good, you can try them on the website) across History, Science, and Art
2. Generate a custom lesson on any topic (my boys ask things like "how tall can a tree get?")
3. Math and typing practice (think a good replacement for math flashcards)
Would love to see what other parents think. Link is in the comments.
Newstart retweeted
Searching in my archives for a person I was about to visit ... I found a hit from 40 years ago... and was reminded of lots of interesting (and surprising) history about ruliology, neural nets and more... (oh, and a "lost" abstract) wolfr.am/CA86
Newstart retweeted
Today, Modern Classrooms Project is announcing that we have raised $12 million to build a free Tier 1 math curriculum for middle and high school, grounded in cognitive science.
The more time I spend in math education, the more concerned I get. Outcomes have been falling for more than a decade, and the steepest declines are among the lowest-performing students.
Meanwhile, many of the leading cognitive scientists and learning scientists have been sounding the alarm that their research is not being applied at scale in math classrooms. We think they might be right.
So we are building a curriculum that treats cognitive science as its north star. We plan to launch middle school in summer 2027 and high school in summer 2028.
Three things to know about what we are building:
1. It will be free. All lessons and materials, teacher-facing and student-facing, will be publicly available online, printable, and modifiable.
2. It will be built with the researchers. In the coming weeks, we will announce a research advisory committee of leading learning scientists. We will work as hard as we can to make sure the structure, design, and underlying content align with what cognitive science tells us about how students, especially struggling learners, master math.
3. It will work with or without technology. Modern Classrooms Project has spent years at the intersection of technology and classroom instruction. This curriculum will have two clear implementation pathways. One that is completely paper-based, built around whole-group, teacher-facilitated instruction. The other will be designed to leverage technology to enable flexible pacing, mastery-based progressions, and increased 1-1 and small-group support from educators.
Perhaps most importantly, we want to understand what works. While we have conviction that following cognitive science is the right path, we are deeply committed to trying to move the field forward so that students can achieve their full potential in math, particularly students farthest from opportunity.
To learn more about what we are building and explore ways to get involved, go to modernclassrooms.org/curricu….
Newstart retweeted
education that is not personalized to the individual is not education.
it is schooling.
schooling seeks one output format for all inputs, which are individuals.
every individual is different and should be educated as such, so as to become the greatest, most flourishing version of themselves.
Newstart retweeted
Too fast, too slow...The reality is there is no perfect pace because there is no "typical" child. The sooner we realize that, the faster we solve the education crisis.
“I understand why they’re doing this, don’t get me wrong, and some of it needs to be done. I just don’t know pushing the kids as fast as they’re pushing them is the right way.”
Parents question pace of Lake Worth ISD academic overhaul after state takeover
fortworthreport.org/2026/10/…
Newstart retweeted
The idea of an intelligence explosion caused by recursive self improvement has been around for a long time but until very recently it did not seem imminent. Now many leading researchers think it may happen quite soon. You can read our paper about it here:
casp.ac/reports/intelligence…
Newstart retweeted
PRO TIP: If you've been trying to teach your adolescent something and they aren't really listening, find a TED Talk where someone says the exact same thing in front of a large audience.
I've been trying to teach my shy kid some simple tricks for making friends. She wasn't interested in trying them.
Found a TED talk, and she immediately started putting them into practice.
Sometimes our jobs as parents isn't to teach, but to find teachers.
Newstart retweeted
贝塔朗菲的名字和他的《一般系统论》，注定是大众认知里的绝对盲区。
概念的高阶同构性，超越了大众的感知上限
贝塔朗菲建立一般系统论的终极野心，是寻找一种适用于一切系统的根本模式、原则和规律。他证明了无论是流体动力学、细胞的代谢、流动的平衡、神经元的网络，还是人类社会的经济结构，在数学拓扑上都存在同构性。
普通人的大脑只能感知特异性。
他们只能看到这是一只猫、那是一台电脑、这是一个假期，他们已经习惯在完全不同的具象事物之间筑起高墙。
而贝塔朗菲要求感知普遍结构。
他要求主体剥离所有肉眼可见的物质表象，直接看到跨越物理、生物、社会维度的共同方程与反馈极性。这种将世界彻底符号化、网络化、抽象化的视角，对于依靠具象体验活着的普通人来说，无异于一种无法解码的高频噪声。
很多人哪怕不学科学，他们的价值观也是被机械还原论深度洗脑的。还原论告诉普通人把东西拆碎了就能理解整体，你只要努力做工局部，就能获得幸福的整体，只要把坏人清除，社会就会变好。
想想贝塔朗菲的名言吧：整体大于部分之和（The whole is more than the sum of its parts）。他早已无情指出，系统的核心功能不在于要素本身，而在于要素之间的结构和互动关系。
如果人们接受了贝塔朗菲，他们就不得不承认自己每天为了买房、度假、结婚、在行业内卷而透支的生命局部要素，在脱离了高阶的系统秩序后是毫无涌现价值的低熵碎片。 大众为了保护自己的生活意义不至于幻灭，会本能地排斥这种解构。
也正因为贝塔朗菲的思想太超前，他催生了系统生物学、控制论、信息论，甚至直接成为了后代心理学中家庭系统治疗的 conceptual root。
或者说，他被化整为零了。
现代的量化行业在用他的系统开放性，AI行业在用他的网络涌现性，神经科学家在用他的层次结构。这些行业拿走了他的工具去变现、去制造经济刺激，但为了维持自己的精英神话和局部利益，它们绝口不提这位将一切一统的奥地利生物学家的名字。
Newstart retweeted
我最近深刻地意识到一件事，概念的字形、命名方式、组合方式与 chunks，会在理解一个对象之前，提前制造一种感觉，而人们经常把这种感觉误认为对象本身的性质。
比如，第一次看到概率论，线性代数，实分析，随机过程，泛函分析，非平衡统计力学。
很多人的第一反应不是理解，而是一种近乎生理性的陌生、抽象和畏惧。但有意思的是，把它们换成 Probability Theory、Linear Algebra、Real Analysis，或者干脆继续剥掉自然语言，只留下：
(Ω, F, P)，Ax = b，ε - δ
原来附着在汉字、英文和学科名称上的某些感觉突然就消失了。
对象没有改变，我们只是把它们符号化，改变了representation。
这让我怀疑，我们平时所谓的这个东西很难、很神秘、让人害怕，究竟有多少是在描述对象，又有多少只是在描述大脑面对某种表征时产生的先验反应。
另一个很有意思的例子是，骷髅会。
一个熟悉中文的人第一次看到“骷髅”两个字，脑中几乎不可能是完全中性的。
白骨、死亡、秘密、阴森、禁忌，视觉意象和情绪在理解组织本身之前就已经抵达了。
于是我们以为自己在认识一个现实对象，实际上首先认识到的，是语言替这个对象制造出来的一层心理皮肤。
如果它首先以另一个名字进入你的大脑，比如 The Order，最初形成的心理模型很可能完全不同。
所以我觉得人并不直接认识世界。
更准确地讲，我们首先接触的不是 Object，而是 Representation。而 Representation 并不是透明的。
一个词的字形、发音、翻译、文化背景、既有语义网络，甚至第一次接触它时所处的环境，都可能给这个对象附加一个 prior。
于是P(difficulty | representation)被我们无数次误认为了P(difficulty | actual structure)
这是学习中一个被严重低估的问题。
很多人并没有输给了知识本身，他们输给了知识第一次出现时制造出来的感觉。
一本数学书翻开，满页都是：
∀ ∃ ε δ ∫ Σ ∇
大脑立刻报告，我看不懂。
但“我现在无法流畅处理这种 representation”和“我没有能力理解它”，根本不是同一个命题。
前者更接近representational distance，后者描述的才是cognitive capacity
我们却经常把二者混为一谈。
这也是为什么我越来越喜欢一种接近冯·诺伊曼式的符号化思维。
因为符号化提供了一种强大的认知操作，那就是暂时把对象从它的语言、文化、情绪和神秘感中剥离出来。
我们永远不要问先问概率论是不是很难，
先问(Ω, F, P)，分别是什么。
Ω 是可能结果的集合，而F 是我们允许讨论的事件集合，P则是给这些事件分配概率的规则。
我们不要去害怕随机过程，先看{X(t)}，t ∈ T，它不过是在每个时间或索引 t 上对应一个随机变量。
名字可以非常宏大，结构却经常异常朴素。一个概念被成功重新编码以后，它的心理重量会突然坍缩。这可也是专家和初学者之间最重要的区别之一。
初学者看到的是字符。字符组成单词，单词组成定义，定义再艰难地映射到意义。而专家经过长期训练以后，大量中间步骤已经被 chunk 掉了。
所以 expertise 在某种意义上，就是不断重新编码 representation：
symbol → chunk → structure → intuition
直到有一天，符号本身开始消失，一个原本需要占据大量 working memory 才能处理的复杂符号串，逐渐坍缩成一个可以整体调用的认知对象。
这时候，学习发生了一次质变。
你不再是在读它，你开始在用它思考。
自然语言给你意义，符号给你精确，结构给你迁移，直觉给你速度。
然后，你需要拥有改变 representation 的能力。自然语言让一个问题显得过于复杂的时候，就把它符号化。符号让一个问题失去意义时，就把它重新翻译成自然语言。
公式太长，就寻找 invariant、symmetry 或新的 variable。几十个变量让 working memory 崩溃，就寻找能够把它们压缩成一个整体的 chunk。
于是，学习成为寻找一种更好的方式去表示知识的绝佳方式。这件事，在我看来，甚至比单纯增加知识量更加重要。
Newstart retweeted
The New York Times: No puedes “gamificar” una educación real.
La obsesión por hacer que aprender sea siempre “divertido” ha llevado a llenar las aulas de pantallas, juegos y estímulos rápidos. Pero educar no es entretener: es exigir atención, esfuerzo y pensamiento profundo.
La tecnología debe ser una herramienta complementaria, no el centro del aprendizaje. nytimes.com/2026/04/19/opini…
