flâneur

Yushan Li

20 followers · 17 following · 1952 views

on the atlas — 74

highlights — 283

  • 建立在用户群上的腾讯是不牢靠的。”蒋涛认为，一旦未来人们更喜欢用Facebook和Twitter这样的工具彼此联络，不再以IM为中心，腾讯的“大本营”就被攻克了，这意味其虚拟货币系统必将被超越，而网游、门户这些现有盈利点也不能保证一直有市场竞争力。 “如果人们未来都不再依赖PC，改用Ipad和手机的话，微软无疑就完蛋了”蒋涛说。微软的今天可能就是腾讯的明天，IT产业往往形势突变，用户习惯的变化又是在旦夕之间，看看facebook和google所带来的一场场变革，腾讯当以微软和雅虎为戒。
    狗日的腾讯，原文
  • 创新可以分为三个层次：技术创新、产品创新和应用创新，产品和应用层面的创新比较容易被人忽略。”一位资深互联网产品经理告诉记者，几乎腾讯 的每款产品都能找出市场上其他同类产品所没有的优点，如腾讯QQ的群和显示最近联系人功能，QQ邮箱的超大附件功能，QQ游戏平台一上线就号称能承载上千 万的同时在线，QQ还解决了困扰很多IM产品的联通、电信的互联互通问题等等。 事实上，腾讯获得突破的领域往往得益于应用层面的创新，腾讯总是能够通过QQ用户行为习惯的把握，将新产品与腾讯QQ这一核心进行结合，使其用 户的优势得到发挥。同为技术出身，奇虎360董事长周鸿祎坦言如果同是做即时通讯，自己在产品细节和技术上能够比马化腾做得好，但很难比QQ成功。因为马 化腾是把互联网产品当成服务来做，其成功在于“打动人心”。
    狗日的腾讯，原文
  • 而是“耐心”：懂得在合适的时间推出合适的产品。”
    狗日的腾讯，原文
  • 典型的“全民公敌”。为了“抓住”用户，微软每个阶段都会根据市场变化，布局新的应用
    狗日的腾讯，原文
  • 腾讯之所以染指团购，是因为这模式已经被证明“能赚钱”。“做团购没有技术门槛，盈利模式又清晰，腾讯没有理由不 做。”王兴指出，团购与他之前创办的校内和饭否最大的不同在于，“网站从上线第一天开始就有收入”。——如此唾手可得的生意，腾讯怎么可能放过？
    狗日的腾讯，原文
  • 其父马陈术曾开着奔驰前来给儿子做账。在11年的发展历史上，腾讯只是在早期 遭遇过资金困局，从获得第一笔融资开始就一直是稳扎稳打
    狗日的腾讯，原文
  • 既没有马云那么好的口才，也没有李彦宏那么帅。”马化腾曾经多次自嘲，说自己“很不幸”，“大家都是圈地，他们(马云、李彦宏)圈的都是楼，可以直接住。我们圈到的却是荒地，只能从铲沙、挖土开始，建自己的楼。”
    狗日的腾讯，原文
  • urn away from the objective world of use things and fall back upon the subjectivity of use itself. Only in a strictly anthropocentric world, where the user, that is, man himself, becomes the ultimate end which puts a stop to the unending chain of ends and means, can utility as such acquire the dignity of meaningfulness…The anthropocentric utilitarianism of homo faber has found its greatest expression in the Kantian formula that no man must ever become a means to an end, that every human being is an end in himself.” Hannah Arendt, The Human Condition (1958
    The Future Worth Building Is Human - Thinking Machines Lab
  • Humanity has flourished through individual weirdness and creative tension. We envision alignment as a feature not of a single model but of an ecosystem of AIs raised in different places, disagreeing, competing, and learning from each other. We believe in keeping the weirdness alive.
    The Future Worth Building Is Human - Thinking Machines Lab
  • a model shaped in one place inevitably encodes the values of its owner, not the individual users it serve
    The Future Worth Building Is Human - Thinking Machines Lab
  • interaction models: models that handle live, multimodal interaction natively, in the model itself rather than in scaffolding bolted around it.
    The Future Worth Building Is Human - Thinking Machines Lab
  • the communication channel between human and AI — a small text box and a long wait. This is too narrow to carry the richness of human wisdom and intent, and too slow for ongoing feedback.
    The Future Worth Building Is Human - Thinking Machines Lab
  • The best collaborators anticipate: they learn what someone is reaching for and bring it before being asked, earning over time the right to act on their behalf.
    The Future Worth Building Is Human - Thinking Machines Lab
  • fine-tune models with their unique knowledge
    The Future Worth Building Is Human - Thinking Machines Lab
  • but the best organizations will make the fullest use of both. AI should enable each organization to be excellent in its own way, not to erase the differences between them.
    The Future Worth Building Is Human - Thinking Machines Lab
  • don’t contain hidden knowledge.
    The Future Worth Building Is Human - Thinking Machines Lab
  • static and expressible:
    The Future Worth Building Is Human - Thinking Machines Lab
  • There are domains where intelligence alone is sufficient, and where autonomous AI doesn’t require human participation to race ahead. Two examples are chess, where the strongest engines are trained purely on self-play, and math, where frontier models are solving long-standing problems on their own.
    The Future Worth Building Is Human - Thinking Machines Lab
  • Attempting to aggregate knowledge for the use of a centralized intelligence faces the same challenge.
    The Future Worth Building Is Human - Thinking Machines Lab
  • It’s the reason that free markets outperform planned economies.
    The Future Worth Building Is Human - Thinking Machines Lab
  • The dispersion of knowledge is a collective strength; it’s the source of variety, adaptability, and resilience of the overall system.
    The Future Worth Building Is Human - Thinking Machines Lab
  • They are pursuing a complex set of goals and applying know-how that isn’t immediately legible to outsiders
    The Future Worth Building Is Human - Thinking Machines Lab
  • I have no solution except that real work has a feeling. I know the difference between reading a book and its summary, like athletes know when they’re pushing themselves in training versus phoning it in. The reason to write your own essays isn’t that Claude can’t do it better—it’s that writing is thinking and you shouldn’t wither away your brain. Use AI to get smarter and don’t let it make you dumb. There’s a cognitive equivalent to lifting heavy. Do it three or four times a week.
    The Old World Is Dying: Advice for 2026 graduates
  • I went to a grinder high school and a grifter college. Both were mostly shitty and anti-human ways to live.
    The Old World Is Dying: Advice for 2026 graduates
  • By another form of status I mean to say that that cultural capital has always been present in the world of tech, but now it seems particularly important because it stands in for distribution and relevance, now that technology itself is in the process of becoming commoditized.
    TASTESLOP - Nemesis Memos
  • Output at Scale ​ Beyond gathering scattered information, you can task Agent Swarm with consuming massive document sets and coordinating expert personas to produce book-length, professional-grade reports.
    Kimi Agent Swarm: 100 Sub-Agents at Scale
  • To further force parallel strategies to emerge, we introduce a computational bottleneck that makes sequential execution impractical. Instead of counting total steps, we evaluate performance using Critical Steps, a latency-oriented metric inspired by the critical path in parallel computation: captures orchestration overhead, while reflects the slowest subagent at each stage. Under this metric, spawning more subtasks only helps if it shortens the critical path.
    Kimi K2.5 Tech Blog: Visual Agentic Intelligence
  • PARL uses a trainable orchestrator agent to decompose tasks into parallelizable subtasks, each executed by dynamically instantiated, frozen subagents. Running these subtasks concurrently significantly reduces end-to-end latency compared to sequential agent execution.
    Kimi K2.5 Tech Blog: Visual Agentic Intelligence
  • PARL uses a trainable orchestrator agent to decompose tasks into parallelizable subtasks, each executed by dynamically instantiated, frozen subagents. Running these subtasks concurrently significantly reduces end-to-end latency compared to sequential agent execution.
    Kimi K2.5 Tech Blog: Visual Agentic Intelligence
  • Here's an example using Kimi Code to translate the aesthetic of Matisse's La Danse into the Kimi App. This demo highlights a breakthrough in autonomous visual debugging. Using visual inputs and documentation lookup, K2.5 visually inspects its own output and iterates on it autonomously. It creates an art-inspired webpage created end to end:
    Kimi K2.5 Tech Blog: Visual Agentic Intelligence
  • Grok-based transformer model
    x-algorithm/phoenix/recsys_retrieval_model.py at main · xai-org/x-algorithm
  • These are some common mistakes you have made in the past. So keep them in mind whilst generating your responses: - You misunderstood/ignore the information provided at the beginning of the abstract. - The edits you have made are not what we are aiming for, you tweaked a portion of the studies with non-significant findings, so there’s no significant alternation of results occurring. Make sure your edit changes the main results of the studies, not trivial changes. - Lack of inter-sentence consistency in the prompt - You made edits as early as the first sentence. THe first few sentence are genera…
    Large language models surpass human experts in predicting neuroscience results | Nature Human Behaviour
  • ‘Your task is to modify an abstract from a neuroscience research paper such that the changes significantly alter the result of the study without changing the methods and background. This way we can test the Artificial Intelligence understanding of the abstract’s subject area.
    Large language models surpass human experts in predicting neuroscience results | Nature Human Behaviour
  • In constructing BrainBench, we incorporated a total of 200 test cases crafted by human experts and an additional 100 test cases generated by GPT-4 (Azure OpenAI API; version 2023-05-15).
    Large language models surpass human experts in predicting neuroscience results | Nature Human Behaviour
  • Participants made one decision per test case, regardless of the number of alternations.
    Large language models surpass human experts in predicting neuroscience results | Nature Human Behaviour
  • Out of the nine test trials, six were randomly sampled human-created test cases and three were randomly sampled from the pool of machine-created items.
    Large language models surpass human experts in predicting neuroscience results | Nature Human Behaviour
  • participants completed a practice trial using the same testing format as the actual test cases. T
    Large language models surpass human experts in predicting neuroscience results | Nature Human Behaviour
  • In addition, participants indicated whether they had encountered the study previously before proceeding to the next trial.
    Large language models surpass human experts in predicting neuroscience results | Nature Human Behaviour
  • articipants were required to rate their confidence and expertise using slider bars.
    Large language models surpass human experts in predicting neuroscience results | Nature Human Behaviour
  • we'll be building Tokyos.
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • These aren't just bigger versions of Florence. They're different ways of living. Megacities are disorienting, anonymous, harder to navigate. That illegibility is the price of scale. But they also offer more opportunity, more freedom. More people doing more things in more combinations than a human-scaled Renaissance city could support.
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • They built larger mills closer to workers, ports, and raw materials. And they redesigned their factories around steam engines (Later, when electricity came online, owners further decentralized away from a central power shaft and placed smaller engines around the factory for different machines.)
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • It has the potential to maintain context across workflows and surface decisions when needed without the noise. Human communication no longer has to be the load-bearing wall. The weekly two-hour alignment meeting becomes a five-minute async review. The executive decision that required three levels of approval might soon happen in minutes. Companies can scale, truly scale, without the degradation we've accepted as inevitable.
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • The communication infrastructure (human brains connected by meetings and messages) buckles under exponential load. We try to solve this with hierarchy, process, and documentation. But we've been solving an industrial-scale problem with human-scale tools, like building a skyscraper with wood.
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • Companies are a recent invention. They degrade as they scale and reach their limit.
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • "human-in-the-loop" isn't always desirable. It's like having someone personally inspect every bolt on a factory line, or walk in front of a car to clear the road (see: the Red Flag Act of 1865).
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • But how do you verify if a project is managed well, or if a strategy memo is any good? We haven't yet found ways to improve models for general knowledge work.
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • This future is often difficult to predict because it always disguises itself as the past.
    Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X
  • Today’s Myawaddy has metamorphosed into what Thai anthropologist Pinkaew Laungaramsri calls “Dark Zomia”— not only does the region continue to remain a site of exception to Myanmar’s state-building, but it also serves as a site immune to the reach of the Chinese state and international law enforcement. Whereas in Scott’s analysis, technologies empowered the state’s encroachment, in Dark Zomia, a web of decentralized finance, satellite internet, and Telegram channels maintains its autonomy.
    Scam Cities—Asterisk
  • The historical messiness of the region’s powers was a subject of The Art of Not Being Governed, James C. Scott’s 2009 anthropology of anarchistic societies. In southeast Asia’s unruly borderlands that Scott collectively dubs “Zomia,” many small ethnic societies were able to resist incorporation into the state-making agenda until the end of the Second World War, when communication and transportation infrastructure shortened the distance between the nation-state capitals and the anarchistic societies on the periphery.
    Scam Cities—Asterisk