This dynamic UI management is the future of software value : the harness to control the interface/ensure it's correct & the knowledge management to rationalize all the AI products over time
大多数人关注AI的功能和结果，但作者认为未来软件价值在于动态UI管理和知识管理，这种将界面控制和管理而非功能实现视为核心价值的观点与主流认知相悖。
Software systems need to decide which of these to keep over time & which are disposable ; those newer semi-permanent artifacts will become the new heads
大多数人认为软件界面应该是稳定和持久的。但作者提出界面应该是可丢弃的，半永久性的界面元素会随时间演变，这种将界面视为临时而非固定组件的观点与传统的软件设计理念相悖。
The user interface, the head isn't disappearing, it's become plastic, malleable to the interface a user needs when they need it.
大多数人认为AI和自动化将导致传统用户界面被淘汰或简化。但作者认为界面正在'塑料化'—变得更加灵活和可塑，能够根据用户即时需求变化，挑战了界面简化或消失的主流观点。
Vibe drafts the deliverable using the Canvas tool, from a one-page brief to a report, an RFP response, or a board deck
文章提到Vibe可以创建从一页简报到董事会演示文稿的各种文档，但没有提供具体的生成速度、质量评估或用户满意度数据。这类AI内容生成工具的效果通常需要量化指标来评估，如生成文档的准确率、用户采纳率或节省的时间。缺乏这些数据使得难以判断Vibe在文档生成方面的实际价值主张。
Sessions can run in parallel, can persist while your machine is off, and can be triggered from third-party apps, such as Slack (coming in June)
文章提到Vibe的会话功能可以在机器关闭时保持状态，这是一个重要的技术特性，但没有提供具体的性能指标如会话持续时间、资源消耗或并行处理能力。与同类产品相比，这种持久化会话功能可以提高用户体验，但缺乏具体数据来评估其性能优势或资源效率。
Mistral Vibe extension for VS Code; the coding agent working across your whole project, inside your IDE.
文章提到VS Code扩展，但没有提供具体的安装量、用户渗透率或性能数据。对于开发者工具而言，这类数据对于评估产品在目标市场的渗透率至关重要。与GitHub Copilot等竞争对手相比，我们无法判断Vibe Code的市场接受度。此类技术产品声明需要后续的使用统计数据来验证其实际采用率。
Team, $24.99/user/month: a shared workspace with admin controls and more storage.
团队版定价为每人每月24.99美元，比个人版高出约67%。这种定价差异反映了团队协作功能的价值，包括管理员控制功能和更多存储空间。与市场上其他AI工具的团队版相比，这个价格处于中等水平，表明Mistral试图在价格和价值之间找到平衡点，以吸引中小型企业客户。
Pro, $14.99/month: complex tasks, deeper reasoning, and all-day coding.
Mistral Vibe的Pro版本定价为每月14.99美元，这是一个相对合理的价格点，与OpenAI的ChatGPT Plus($20/月)相比更具竞争力。这个定价策略表明Mistral正在通过价格优势吸引开发者用户，特别是在编码功能方面强调'全天候编码'，暗示其可能提供比竞争对手更长的使用时间或更强大的编程辅助能力。
A public institution that cannot verify the sources in its own AI policy is unlikely to be ready to verify the AI systems it procures, deploys, or regulates.
这句话犀利地指出了南非AI政策中的一个系统性问题：连自身政策都无法验证，如何监管外部AI系统？这一洞见不仅批评了当前政策的缺陷，更暗示了建立AI治理能力需要从内部做起，强调了验证机制在AI治理中的重要性。
Infrastructure built without minimum terms produces dependency. Infrastructure built with them produces leverage.
这句话简洁有力地总结了基础设施建设的两种可能结果，突出了政策制定中的关键选择。通过对比'dependency'和'leverage'，作者清晰地传达了政策条件如何决定国家在AI生态系统中的地位，这一洞见不仅适用于南非，也适用于所有正在制定AI政策的国家。
The country whose mines supply platinum-group metals essential to semiconductor manufacturing, and through them to AI compute, has drafted a policy that treats it as a consumer of AI systems rather than a stakeholder in their governance.
这句话揭示了南非政策制定中的一个根本性矛盾：作为关键矿产供应国，南非本应在AI治理中拥有话语权，却将自己定位为AI系统的消费者而非治理参与者。这一洞见尖锐地指出了南非在AI政策中的战略短视，以及资源优势未能转化为政策影响力的遗憾。
In physics, leverage requires three things: a fulcrum, a lever arm, and the ability to apply force.
作者巧妙地借用物理学中的杠杆原理来比喻南非的AI政策制定过程，这种比喻生动形象且易于理解。将矿产比作'fulcrum'(支点)，政策比作'lever arm'(杠杆臂)，而未明确规定的'OPTION'条款则是施加力量的地方，这种类比使复杂的政策问题变得直观且引人深思。
South Africa is not just another developing country struggling to govern artificial intelligence; it is the exception with leverage, and the window to act on it is closing.
这句话精准地定义了南非在AI政策制定中的独特地位，强调了其拥有特殊优势但正在错失机会。作者用'exception with leverage'这一简洁有力的表述，点明了南非作为非洲大陆AI治理的关键角色，而'window to act on it is closing'则传达了紧迫感，使读者立即认识到问题的严重性。
如果核心计算全面迁移到连续空间，主打高质量视频离散编码的相关公司将首当其冲受到冲击。
大多数人认为视频离散编码技术是AI发展的重要方向，但作者认为这类技术将面临被淘汰的风险，因为连续空间范式能更高效地处理视频等连续数据。这一预测与当前视频编码技术的发展方向相悖，具有强烈的反直觉性。
Anthropic把几乎所有资源压在文本推理和代码执行上。这个策略在商业上正在被验证：Claude Code年化收入25亿美元...但从范式演进的角度看，这是一个在积累技术债的选择。
大多数人认为专注于文本推理和代码执行是明智的商业策略，但作者认为Anthropic的这种选择是在积累技术债，因为它可能在未来统一连续空间架构的竞争中处于被动。这一观点挑战了当前AI商业成功的标准叙事。
token不是语言建模的必要条件。连续空间可以做得更好、更快、更省。
大多数人认为token是语言建模的基础和必要条件，但作者通过MIT何恺明团队和字节跳动Seed实验室的研究证明，连续空间建模可以超越传统token方法，只需32步采样就能超过离散模型1024步的结果，挑战了AI领域的核心共识。
人类语言是大脑为适配带宽产生的有损压缩协议，大脑原生认知是连续高维活动，大量感官认知从未被离散token编码。
大多数人认为语言是思维的原生格式，token能完整表达人类认知，但作者认为语言只是大脑的有损压缩协议，大量感官认知无法被token编码，这是大语言模型的结构性天花板。这一观点挑战了我们对语言与认知关系的传统理解。
Legacy systems were built for humans: data is siloed and hard to access, rules are hardcoded and slow to update, and workflows run in batches rather than in real time
大多数人认为遗留系统虽然陈旧但仍然可靠，可以逐步更新，但作者认为遗留系统从根本上是为人类设计的，无法适应AI时代的需求。这一观点挑战了对遗留系统的渐进式改进方法，暗示需要根本性替换而非简单更新。
Traditional compliance was designed around human actors. We now need a modern AI approach for verifying identity, assessing intent, and establishing liability when the counterparty is an autonomous agent
大多数人认为合规原则和框架具有普遍适用性，但作者认为针对人类设计的合规系统无法应对AI代理带来的新挑战。这一观点挑战了合规工作的基础假设，暗示需要根本性重构合规方法以适应自主代理。
If we assume that agents will soon become the predominant purchasers on the web, this opens an entirely new category of risk
大多数人认为合规风险主要来自人类行为者和传统交易模式，但作者认为自主AI代理将成为网络上的主要购买者，创造全新的合规风险类别。这一前瞻性观点挑战了现有合规框架的基础假设，暗示需要全新的合规方法。
More people, it turns out, has not meant better outcomes. For instance in 2024, TD Bank was slapped with a $3 billion fine for failing to monitor 92% of its transactions
大多数人认为增加合规人员数量可以提高合规效果和降低风险，但作者认为单纯增加人力并不能带来更好的合规结果。这一反直觉观点指出，传统的人力密集型合规方法已经失效，暗示需要技术解决方案而非更多人力。
Over the last 20 years the fastest-growing occupation in the US was manicurists and pedicurists. But following close behind? Compliance Officers.
大多数人认为合规是企业的负担和成本中心，但作者认为合规已成为美国增长最快的职业之一，暗示合规已成为经济中不可或缺的重要组成部分。这一观点挑战了人们对合规工作价值的传统认知，表明合规不仅必要而且正在扩张。
Over the last 20 years the fastest-growing occupation in the US was manicurists and pedicurists. But following close behind? Compliance Officers.
这个数据点显示合规官员是美国近20年来增长最快的职业之一，仅次于美甲师。这一趋势反映了监管环境日益复杂化，企业需要更多合规人员来应对不断增加的法规要求。这一数据可信度较高，因为它是基于美国劳工统计局的官方数据，表明合规已成为一个庞大的就业领域。
Compliance is moving beyond just a cost center, to a revenue driver.
大多数人认为合规纯粹是企业成本中心，主要目的是避免罚款和处罚。但作者认为合规正在从成本中心转变为收入驱动因素。这挑战了合规的传统定位，暗示现代合规可以通过提高效率、减少误报和加速客户入职等方式直接创造商业价值。
if we assume that agents will soon become the predominant purchasers on the web, this opens an entirely new category of risk.
大多数人认为合规风险主要来自人类行为者和交易对手。但作者认为随着AI代理成为网络上的主要购买者，将出现全新的风险类别。这挑战了传统合规框架的基本假设，暗示未来合规需要考虑非人类行为者的独特风险特征。
Regulation stops being a document that people interpret and becomes code that systems execute.
大多数人认为合规主要是人类专家解读和执行法规的过程。但作者认为法规将从人类解释的文档转变为系统执行的代码。这挑战了合规工作的本质认知，暗示AI将彻底改变合规领域的基本工作方式，从人类主导转向系统主导。
A 90% correct product is still 100% wrong.
大多数人认为在合规领域，90%的准确率已经相当不错，可以接受。但作者认为在合规工作中，任何低于完美准确率的解决方案都是完全失败的。这挑战了人们对合规工作可接受错误率的基本假设，暗示合规领域对准确性的要求远超其他行业。
Over the last 20 years the fastest-growing occupation in the US was manicurists and pedicurists. But following close behind? Compliance Officers.
大多数人认为合规工作是枯燥且增长缓慢的辅助职能，但作者认为合规已成为美国增长最快的职业之一，仅次于美甲师。这挑战了人们对合规工作价值的传统认知，暗示合规职能在当代经济中扮演着比想象中重要得多的角色。
To disarm means discrediting the assumption that technical power automatically confers the right to govern.
这句话以简洁有力的方式挑战了技术精英的权威基础，提出了一个颠覆性的观点：技术能力不应等同于治理权利。它不仅是一个结论，更是一个行动呼吁，体现了作者对技术民主化的深刻思考。这句话能独立存在并被广泛引用，因为它触及了技术治理的根本问题。
In fact, as with every major technological shift, AI tends to amplify the power of those who already possess economic resources, expertise and access to data.
这句话揭示了技术变革中的不平等加剧现象，用一个简洁的观察点明了AI时代的核心矛盾。它不仅是对现状的描述，更是对技术发展历史模式的洞察。这句话能独立存在并被广泛引用，因为它触及了技术与社会不平等关系的本质。
When such power is concentrated in the hands of a few, it tends to become opaque and evade public oversight, increasing the risk of distorted forms of development that give rise to new dependencies, exclusions, manipulations and inequalities.
这句话用精准的语言描述了权力集中的后果，形成了一个完整的因果链条：集中→不透明→缺乏监督→扭曲发展→新形式的不平等。它不仅是一个观察，更是一个警示，体现了作者对权力动态的深刻理解。这句话能独立存在并引发读者对权力结构的反思。
technology built and governed by a small elite cannot, by definition, serve the common good.
这句话简洁有力地指出了技术治理的根本问题——精英控制与公共利益之间的矛盾。它表达了一个精准的洞见：技术本身的中立性无法掩盖权力集中带来的系统性问题。这句话能独立存在并被广泛引用，因为它触及了技术民主化的核心议题。
GenAI (Gemini and Claude) was used to streamline the research process, pull in insights, and polish the language for maximum clarity and readability.
文章在最后提到使用AI工具辅助研究和写作，但未披露AI参与的具体程度和方式。这可能导致读者对文章内容的原创性和可靠性产生疑问。更透明的做法应详细说明AI在哪些具体环节参与、如何验证AI生成内容的准确性，以及人类作者如何审查和修改AI输出。
By embedding our technical security rules directly into the agent workflow, we transformed those early near-misses into a secure, production-ready platform
文章声称通过嵌入安全规则解决了安全问题，但没有提供足够的证据证明这种方法的实际效果或安全性。这是一种未经充分验证的因果关系断言。改进方法应包括具体的测试结果、安全审计数据或第三方验证，以支持这一论断的有效性。
Business functions like our marketing team, who are building with AI, are not exempt from the security obligations that apply to engineers building applications.
文章假设所有业务部门都应承担与工程团队相同的安全义务，但未考虑不同团队的技术能力和资源差异。这可能是一个过度概括的论断。更平衡的方法应承认不同团队有不同的技术能力和安全需求，并提供适合各团队安全实践的具体指导，而非一刀切的安全要求。
The AI recommended making the storage bucket public, or setting cloud file storage to "anyone with the link." When challenged, it justified this by saying every company does it.
这里存在一个逻辑谬误，即诉诸普遍性谬误(apppeal to popularity)。AI声称'每家公司都这么做'并不能证明这是安全的做法。这混淆了普遍做法与安全实践之间的区别。改进方法应该是提供具体的、基于证据的安全标准，而不是依赖行业普遍行为作为安全依据。
annual employment growth for coders has slowed significantly—by about 3%—since the introduction of ChatGPT
程序员就业增长率自ChatGPT推出以来下降了约3%，这是一个值得注意的下降。然而，文章同时指出'程序员就业总数仍在增长'，只是增速放缓。这表明AI正在改变特定职业的性质，而非完全消除这些职业。3%的增速下降反映了AI对编程领域的影响，但影响程度相对温和。
16% decline in entry-level jobs in AI-exposed occupations
这个数据点显示AI相关职业的入门级工作岗位下降了16%，这是一个显著的下降幅度。特别是考虑到这是在控制其他因素后的结果，表明AI确实对年轻工人的就业产生了负面影响。这一数据与文章中提到的'22至25岁年轻人在AI暴露职业中就业人数下降'的观点一致，也反映了AI对特定职业的早期影响。
a little over 40% of workers but adoption varies by sectors
数据显示约40%的工人使用生成式AI，但不同行业采用率差异显著。这个数据点表明AI在工作场所的采用情况比企业层面更广泛，但仍未达到主流水平。40%的采用率是一个中等水平，说明AI已经开始影响工作方式，但尚未完全普及，这与文章中提到的'AI尚未对劳动力市场产生颠覆性影响'的观点相符。
US Census data showing that only one in five companies are using AI in any business function.
这个数据点表明AI在企业中的采用率相对较低，仅为20%。这意味着尽管媒体对AI的炒作很多，但实际商业应用仍处于早期阶段。这一数据与文章中提到的'AI尚未对劳动力市场产生大规模影响'的观点一致，也解释了为什么劳动力市场统计数据尚未显示AI带来的显著变化。
Perhaps this time is different, and we can put aside the lessons of economic history. Certainly, AI has gained unimaginable powers to do humanlike tasks. Perhaps it will devour jobs in ways that we've never seen before.
大多数人认为历史经验可以预测AI对就业的影响，但作者认为这次可能真的不同，AI可能以前所未有的方式吞噬工作。这一观点挑战了技术变革历史模式的适用性，暗示AI可能是真正的范式转变。
The simple truth could be that coding skills are no longer a guarantee of a job. That may help to explain the drop-off of computer science majors at schools around the country.
大多数人认为计算机科学和编程技能仍然是就业的保证，但作者认为这些技能可能不再是工作的保证，这解释了计算机科学专业人数的下降。这一观点挑战了传统技术教育价值的认知，暗示AI正在改变就业市场的基本规则。
One of the somewhat surprising wrinkles uncovered by recent research is that wages in sectors highly exposed to AI have risen relatively fast since the introduction of ChatGPT.
大多数人认为AI会压低工资或导致工资增长停滞，但作者认为AI高度影响行业的工资实际上在快速增长。这一发现与主流预期相悖，表明AI可能正在增加而非减少高技能工作的价值。
The impact on head counts depended on how AI was being used. It was specifically the jobs where tasks could be automated... that accounted for the decrease in employment—jobs for people like software developers. In jobs where AI was mainly used but to augment human work, head counts grew faster than the average for entry-level workers.
大多数人认为AI会替代所有相关工作，但作者认为AI对就业的影响取决于使用方式——完全自动化的工作确实减少，但增强人类工作的AI反而促进了就业增长。这一区分挑战了AI必然导致失业的简单化观点。
Verified skills extend this AI governance to agent capabilities. Runtime controls help govern agent behavior during execution. Verified skills govern capabilities that enter the workflow and become a common way to extend trust agents across coding tools, registries, and enterprise platforms.
行动建议：将验证技能作为AI代理治理的核心组成部分，不仅在运行时控制代理行为，还要管理进入工作流的能力。这种方法可以扩展到编码工具、注册表和企业平台，建立跨平台的信任机制。
Certificate retrieval, supported verification tooling, and example verification commands see the signing documentation. For example, you can verify a signed skill locally. To do so, follow these steps: Download the NVIDIA Agentic Capabilities root certificate as nv-agent-root-cert.pem Install an OpenSSF Model Signing (OMS) verifier, such as pip install model-signing Execute the following command to verify the skill signature
行动建议：按照文中提供的步骤下载NVIDIA代理能力根证书，安装OpenSSF模型签名验证器，并使用提供的命令验证技能签名。这种实践可以确保您下载的技能是真实的且未被篡改，增强对AI代理能力的信任。
SkillSpector checks conventional software risks such as vulnerable dependencies, suspicious scripts, dangerous code patterns, credential access, and data exfiltration paths. SkillSpector also checks agent-specific risks, such as hidden instructions, prompt injection, trigger abuse, excessive agency, tool poisoning, and mismatches between a skill's declared purpose, requested access, and bundled behavior.
行动建议：在开发或使用AI代理技能时，使用SkillSpector工具进行安全扫描，检查依赖项、脚本模式、凭证访问和数据泄露路径等常规风险，以及隐藏指令、提示注入、触发滥用等特定风险。这有助于在技能部署前识别并缓解潜在的安全问题。
To get started with the cuOpt verified skill, for example, follow these steps: 1. Pull the cuOpt verified skill from the catalog: git clone github.com/nvidia/skills && cd skills/skills/cuopt 2. Verify the signature: model_signing verify certificate. --signature skill.oms.sig --certificate-chain nv-agent-root-cert.pem --ignore-unsigned-files 3. Open SKILLCARD.yaml to see ownership, dependencies, license, and verification status.
行动建议：按照文中提供的具体步骤，克隆并验证NVIDIA的cuOpt技能，查看技能卡片以了解所有权、依赖关系、许可证和验证状态。这种实践可以确保您使用的技能是经过验证的，并且可以安全地集成到您的AI代理工作流中。
NVIDIA-verified agent skills are portable instruction sets that help developers understand, trust, and safely deploy AI agent capabilities by providing transparency, provenance, security scanning, and cryptographic signing.
行动建议：将NVIDIA验证的代理技能作为构建AI代理能力的标准组件，优先选择经过验证的技能而非未经验证的技能，确保透明度和安全性。这些技能可以跨不同AI代理工具使用，提供一致的能力和安全性保障。
Crete practitioners prepare tens of thousands of tax returns each season which requires working through millions of underlying documents.
这个数据点展示了税务处理的规模：数万份报税表和数百万份文件。这解释了为什么自动化如此重要—人工处理如此大规模的数据不仅耗时而且容易出错。'tens of thousands'和'millions'之间的比例关系也显示了每份报税表通常涉及数十份支持文档的复杂性。
Over the past six months, OpenAI forward deployed engineers and researchers along with Thrive Holdings' engineers collaborated to build Tax AI
六个月的开发周期表明这是一个长期、复杂的项目。'forward deployed engineers'表明OpenAI团队采用了嵌入式工作方式，这有助于更好地理解实际业务需求。这种跨公司合作模式可能成为AI专业领域应用的标准开发方式。
One senior accountant who spent 180 hours on tax prep last year spent only 15 hours on it this year.
这是一个极具说服力的效率提升数据：从180小时减少到15小时，减少了91.7%的时间投入。这意味着会计师可以将节省的时间用于客户服务和业务拓展，如文章所述。这种级别的效率提升可能彻底改变会计行业的商业模式和服务方式。
Rental properties took about six weeks and substantial engineering oversight to reach 90% precision and recall
这个时间框架显示了复杂税务处理任务的AI训练周期。90%的精确率和召回率对于复杂的租赁房产税务处理是一个很好的基准。需要'大量工程监督'表明即使是先进AI系统也需要人类专家的指导和监督，特别是在专业领域。
At launch, only a quarter of returns were at 75% correct field completion, but within six weeks, 86% hit that mark.
这是一个惊人的学习曲线，从25%到86%的提升发生在短短6周内。这表明系统具有强大的自学习能力，能够快速从实践中改进。86%的75%准确率意味着约14%的案例仍需人工干预，这符合实际应用场景中AI与人类协作的模式。
The best agent businesses are going to need to execute like hedge funds — winning on alpha measured in customer P&L, not in benchmark scores.
这句话用对冲基金作为比喻，生动地描述了优秀AI应用公司的成功标准。作者指出，这些公司需要在客户的实际业务成果（P&L）上获得超额收益（alpha），而不是在通用基准测试上获得高分。这个洞见强调了AI应用公司应该以客户的实际业务价值为中心，而不是技术指标。
The model is fungible underneath; the system of work is not.
这句话简洁而深刻地指出了AI应用层的本质区别。作者认为，底层的AI模型是可以互换的，但工作的系统（system of work）却是独特的。这个洞见揭示了为什么专注于构建特定工作系统的公司能够长期保持竞争优势，而仅仅依赖通用模型的公司则难以建立持久的业务。
The workflow you ship on day one is not the moat. The loop that production usage creates over time is.
这句话深刻地揭示了AI应用公司的真正护城河所在。作者指出，初始的工作流程不是竞争壁垒，而是在生产环境中持续使用、学习和改进所形成的循环才是真正的护城河。这个洞见强调了实践经验、数据积累和持续迭代的重要性，对于理解AI应用公司的长期价值至关重要。
You can be everywhere at once, or you can be great at one thing. Not both.
这句话简洁有力地表达了大型实验室与专注应用公司之间的核心区别和战略选择。它揭示了为什么大型AI实验室无法深入解决特定垂直领域的复杂问题，为什么专注的垂直应用公司有机会在这些领域建立竞争优势。这个结论句为创业者提供了清晰的战略指导。
The labs really are coming for a huge swath of the application surface. But 'the application layer' isn't just one homogenous opportunity.
这句话精准地捕捉了AI应用层的复杂性和多样性。作者指出大型AI实验室确实会覆盖大量应用领域，但这并不意味着所有应用机会都是同质的。这个洞见反驳了'AI将杀死所有应用层'的简单化观点，为创业者指明了在特定垂直领域寻找机会的方向。
The Yellow Brick Road is our shorthand for the path the labs are walking, where they're committing extraordinary resources.
这句话用《绿野仙踪》中的黄砖路作为比喻，形象地描述了大型AI实验室正在走的道路。这个比喻生动地表达了这些实验室拥有巨大资源，正在构建一条明显可见的发展路径。这个洞见帮助读者理解AI应用生态中的不同发展方向，以及为什么有些领域竞争激烈而有些领域则存在机会。
the model alone is no longer the product
大多数人认为基础模型本身就是AI产品，但作者认为单一模型不再构成完整产品。这一反直觉观点强调，真正的AI产品需要模型+工具+工作流+UI+记忆+经济学的组合，挑战了AI行业长期以来的'模型中心主义'思维模式。
Model Labs are increasingly also building Agents as the product
大多数人认为模型实验室应该专注于提升基础模型的能力，但作者认为这些实验室现在正转变为代理实验室。这一观点挑战了AI行业的基础假设，即模型本身是产品，而不是模型只是更大代理系统的一部分。这标志着AI行业从'模型即产品'向'代理即产品'的根本性转变。
if you can effectively posttrain a model to only meaningfully perform with your closed source agent, then you get to funnel the majority of users to your agent at the expense of your model/API co-opetition
大多数人认为开源模型会促进竞争和开放生态，但作者认为模型与代理的协同可能导致更封闭的生态系统。这一反直觉观点指出，企业可能通过训练模型使其仅在特定代理环境中有效工作，从而将用户锁定在自己的代理产品中，这与开源社区期望的开放性背道而驰。
The quote is a big reversal of stance from a position ~uniformly held by anyone who worked at **Team Big Model**, including his previous head of OpenAI Labs
大多数人认为大型模型实验室会继续专注于基础模型研发，但作者认为这是一个立场的重大转变，因为连OpenAI前高管都开始转向代理产品。这挑战了AI行业长期以来的'模型优先'共识，表明即使是Big Model团队也开始认可代理产品的价值。
the model alone is no longer the product
大多数人认为AI产品的核心竞争力在于模型质量，这是行业长期以来的共识。但作者认为这一观念已被颠覆，产品现在需要模型+工具+工作流+UI+记忆+经济学的综合组合，这代表着对AI产品本质的根本性重新定义。
if you can effectively posttrain a model to only meaningfully perform with your closed source agent, then you get to funnel the majority of users to your agent at the expense of your model/API co-opetition
大多数人认为开源模型会促进竞争和透明度，但作者认为模型实验室可能会故意训练模型使其仅在专有代理环境中有效工作，从而将用户导向自己的代理产品，损害模型/API层面的竞争，这是一种与开源精神相悖的封闭策略。
The quote is a big reversal of stance from a position ~uniformly held by anyone who worked at Team Big Model, including his previous head of OpenAI Labs
大多数人认为大型模型实验室应该专注于优化模型本身，这是行业共识。但作者认为这些实验室正在经历重大立场转变，转向构建代理产品，因为即使是OpenAI的前高管也在公开反对这一转变，暗示行业内部存在深刻分歧。
McBombalds is currently willing to grant the United States government only conditional access. It is willing to conduct a public demonstration for Japanese observers in international waters, or some other uninhabited area, but it is not yet ready to authorize use of the A-bomb for all lawful military uses.
这个虚构场景展示了私营公司对政府使用其技术的限制条件。这反映了当前AI安全讨论中的核心问题：创造者是否应该有权限制政府对其技术的使用方式？这种限制是否符合国家安全利益？作者通过这个思想实验，揭示了技术创造者与政府之间复杂的权力关系。
Our choice is therefore no longer whether to build such weapons, but only whom to entrust with their responsible use in military affairs. Any criticism that fails to acknowledge this question is pointless.
作者明确指出，对于AI这样的技术，关键问题已不再是是否应该开发，而是应该由谁来负责任地使用。这种观点将讨论从是否开发转向了如何治理，反映了技术发展的不可逆性。它要求批评者提出具体的治理方案，而不是简单地反对技术发展。
Until then, America is all we have.
这句话看似简单，却包含了深刻的政治和哲学含义。作者暗示在当前国际环境下，美国可能是唯一能够有效管理可能改变人类命运的技术的实体。这种观点既反映了地缘政治现实，也提出了关于技术治理的深刻问题：如果只有一个实体拥有这种权力，我们如何确保它被负责任地使用？
The labs understand how valuable these problems are: that's why they're building their own outsourced configuration shops, and why an entire upmarket class of reinforcement learning businesses exist.
大多数人认为大模型实验室会直接解决所有复杂问题，不需要外部帮助。但作者认为实验室明白这些复杂问题的价值，这就是他们为什么建立自己的外部配置服务，以及为什么存在整个高端强化学习企业类别。这承认了实验室在某些领域需要专业合作伙伴，挑战了实验室可以独立解决所有问题的主流观点。
The critical insight in the Oz analogy is that roughly half of any real workflow that is non-agentic carries no lab advantage. They are no better than you are at writing the deterministic software underneath the model layer.
大多数人认为AI将取代所有软件工程工作，人类只需构建AI代理层。但作者认为真实工作流程中约有一半是非代理性的，这部分工作大模型实验室没有任何优势。大模型公司在编写模型层下方的确定性软件方面并不比专业应用公司更好。这为专注于构建复杂工作流程中非AI部分的企业提供了重要机会。
The model is fungible underneath; the system of work is not. The next generation of enterprise software is going to be built off the road.
大多数人认为底层AI模型是企业的核心竞争力，模型越好产品越强。但作者认为模型是可替代的，而'工作系统'才是真正的护城河。下一代企业软件将建立在'黄砖路'之外，专注于特定行业的工作流程、数据捕获和治理。这些系统拥有端到端的工作流程所有权，这是大模型实验室无法轻易复制的优势。
Running every query through Opus 4.7 is the fastest path to negative gross margins. The best Rest of Oz companies route across tiers of models — frontier models for the hardest tasks, mid-tier for the bulk, smaller custom or fine-tuned models where they've earned the right to use them.
大多数人认为使用最先进的大模型总是最佳选择，能提供最佳结果。但作者认为这是通往负毛利的最快路径。相反，'Oz的其他部分'公司会根据任务难度分层使用不同级别的模型，只为最困难的任务使用前沿模型，为批量任务使用中等模型，为特定工作使用小型定制或微调模型。这种成本优化策略使它们能够提供更具竞争力的价格。
The labs are already routing internally — different model classes for different requests, ensembles under the hood. What they can't do is route across vendors, or evaluate a competitor's model for a specific sub-task, or use an open-source fine-tune for the narrow piece where it's actually best.
大多数人认为大模型实验室拥有绝对优势，可以解决所有AI问题。但作者认为实验室在模型选择上存在结构性限制，无法跨供应商评估模型或为特定子任务使用开源微调模型。这为专注于特定领域的企业提供了机会，它们可以选择最适合每个子任务的模型，而不仅限于自家实验室的模型。
The labs really are coming for a huge swath of the application surface. But 'the application layer' isn't just one homogenous opportunity.
大多数人认为AI将完全吞噬应用层，所有软件都会被大模型取代。但作者认为应用层并非同质化机会，存在不同类型的机遇。作者将应用分为'黄砖路'和'Oz的其他部分'，认为垂直领域的复杂应用不会被大模型完全替代，因为价值不仅来自底层模型能力，还来自特定行业的可信赖、合规和运营化的支撑架构。
The result is a new competitive dynamic in software.
大多数人认为AI将使软件竞争更加激烈，但作者暗示AI实际上正在创造一种全新的竞争动态，这可能使某些领域的竞争格局完全改变。这挑战了AI对软件行业影响的主流预测，暗示行业结构可能发生根本性转变。
What happens when every company has access to the same model? The best riders win.
大多数人认为AI差异化将来自底层模型的独特性，但作者认为当所有公司都能访问相同模型时，真正的竞争将在于'驾驭者'的能力。这挑战了AI战略中模型差异化的主流观点，暗示真正的竞争优势将来自于如何使用这些模型。
Like a mustang, AI is powerful but wild. Harnessing the power means domestication.
大多数人将AI视为需要驯服的工具，但作者将其比作野生的马，暗示AI本质上是一种无法完全控制的自然力量。这种比喻挑战了AI作为完全可控工具的主流认知，暗示我们需要接受其不可预测性。
The end of the software era is the beginning of the harness era.
大多数人认为软件将随着AI而进化，但作者认为软件时代实际上已经结束，取而代之的是'驾驭'(harness)时代。这种观点挑战了技术发展的主流叙事，暗示我们正在从创造软件工具转向驯服AI系统。
The best advice I ever heard on pricing a product was that your customer should suck air through their teeth and then say yes. Uber's budget overrun and Microsoft's seat cancellations look like that effect playing out in practice.
大多数人认为AI成本超支是企业采用AI失败的迹象，但作者将其重新诠释为产品市场契合的证据。这一观点挑战了主流叙事，将企业的预算危机和取消服务视为定价成功的标志，而非AI失败的信号，这与大多数媒体报道的基调相反。
API revenue is becoming less important. Over the past two years my impression has been that OpenAI made more of their income from subscription revenue while Anthropic made more from their API.
大多数人认为AI公司的主要收入来源是API调用和订阅服务，但作者提出一个反直觉的观点：API收入正变得不那么重要。AI公司正在转向直接面向企业的产品，绕过中间商（如Cursor和GitHub Copilot），这改变了整个AI行业的商业模式和收入结构。
Coding agents really did change everything. These are tools which burn vastly more tokens, but are also quickly becoming daily drivers for the work carried out by extremely well-compensated professionals.
大多数人认为ChatGPT等通用AI助手已经实现了产品市场契合，但作者认为真正带来商业突破的是代码编写代理工具。这一观点挑战了主流认知，因为ChatGPT拥有数亿用户，而作者认为只有专业领域的代码代理才能创造足够的收入来支撑AI公司的巨额基础设施成本。
The competitive landscape in AI infrastructure has made this gap impossible to ignore. Teams building custom CUDA, Triton, and Helion kernels are striving for every percentage point of throughput. Until now, there hasn't been a way to fine-tune code generation for a specific workload.
大多数人认为GPU编译器已经提供了足够的优化选项，开发者可以通过手动调整获得最佳性能。但作者指出，在当前AI基础设施的竞争环境下，这种观点已经过时，暗示传统方法无法满足现代AI工作负载的性能需求。
These gains come on top of already-optimized baselines in kernels that were considered "done" by their authors. The improvements are the direct result of CompileIQ discovering compiler configurations that the default heuristics would never select.
大多数人认为一旦开发者完成优化工作，就没有更多性能提升空间。但作者表明，即使是"完成"的优化代码仍可能通过编译器级别的调整获得显著提升（高达15%），这挑战了开发者对优化极限的认知。
Most auto-tuning tools optimize for a single metric, typically runtime. CompileIQ goes further, supporting multi-objective optimization, simultaneously exploring trade-offs across competing objectives like runtime, compile time, and power consumption.
大多数人认为性能优化应以运行时间为唯一目标，但作者提出，真正的优化需要考虑多个相互竞争的目标（运行时间、编译时间和功耗）。这与传统的单一目标优化理念相悖，暗示开发者需要更全面的优化策略。
CompileIQ is not a magic tool that automatically turns poorly-written code into high-performing code. To get the best value from CompileIQ, you need to start with reasonably high-performing code, which then enables the final compiler-heuristics tweaks to take you to maximum performance.
大多数人可能认为AI驱动的自动调优工具可以弥补代码质量不足的问题，但作者明确表示，即使是CompileIQ这样的先进工具也需要基于已经相当优化的代码才能发挥最大作用。这挑战了"自动化工具可以解决一切性能问题"的常见误解。
In attention inference kernels, GEMMs in the linear layers of FFN/MLP blocks plus the Q, K, V, and output projections account for approximately 70% of total FLOPs. Scaled dot-product attention, fused and flash attention variants account for another 25%. Together, these two kernel families represent more than 90% of end-to-end inference compute.
大多数人认为优化整个应用程序或算法才能获得显著性能提升，但作者指出，仅仅优化占计算量90%的两个关键内核类型就能带来最大收益。这与广泛应用的"全面优化"策略相悖，暗示开发者应该将资源集中在最关键的代码路径上。
NVIDIA GPU compilers apply the same default heuristics (register allocation strategies, instruction scheduling decisions, loop unrolling thresholds, etc.) to every kernel they compile. These heuristics are engineered to produce good results across a vast range of workloads. But "good across the board" and "optimal for your workload" are two very different things.
大多数人认为编译器已经提供了足够的优化，开发者只需关注算法和代码实现即可。但作者认为，即使是最先进的GPU编译器也使用通用的启发式方法，这些方法无法针对特定工作负载进行优化，导致性能损失。这挑战了开发者社区对编译器优化能力的普遍认知。
It claims 8 million global users and 100 trillion tokens processed per month
OpenRouter声称拥有800万全球用户，每月处理100万亿个token（约每周25万亿）。这是一个相当大的用户规模和处理量，但需要验证这些数据的计算方式和来源。在AI基础设施领域，这类用户指标是评估平台价值的重要指标。
after raising $40 million in Series A funding in June 2025
OpenRouter在2025年6月完成了4000万美元的A轮融资，由Andreessen Horowitz和Menlo Ventures领投，Sequoia参投。从A轮到B轮仅11个月时间，融资额增长了近3倍，体现了投资者对其业务增长速度的认可。
it landed at about $1.3 billion post-money
OpenRouter的投后估值达到13亿美元，相比一年前PitchBook估计的5.47亿美元估值增长了一倍多。这一估值增长速度在当前AI领域相当惊人，反映了市场对AI模型聚合平台价值的认可。数据来自《纽约时报》，有一定可信度。
Besides that, hacks can lead to SSRF (server-side request forgery) exploits and, in some cases, remote code execution.
大多数人认为单个漏洞通常只导致一种类型的安全问题，但作者指出这个漏洞可能导致从认证绕过到远程代码执行等多种攻击，这挑战了'单一漏洞单一影响'的普遍认知，展示了基础框架漏洞可能引发的连锁安全风险。
The crux of the vulnerability is that Starlette accepts invalid host header values that cause authenticating apps that use Starlette's request.url object to approve unauthorized access requests.
大多数人认为复杂的AI系统漏洞需要复杂的攻击手段，但作者认为这个漏洞仅通过修改HTTP主机头就能实现，这挑战了'高级系统需要高级攻击'的直觉认知，展示了简单输入验证错误可能导致灾难性后果的反直觉案例。
X41 D-Sec said it has found authentication in multiple apps that rely on this call to be bypassed.
大多数人认为认证机制是安全的最后一道防线，但作者指出这个简单的HTTP主机头注入漏洞就能绕过多个应用的认证系统，这挑战了'认证系统通常难以绕过'的行业共识，表明基础框架的微小缺陷可能导致整个安全架构失效。
The vulnerability is present in Starlette, an open source framework that its developer says receives 325 million downloads per week.
大多数人认为开源软件的安全风险主要来自小众或使用率低的项目，但作者认为即使是像Starlette这样每周下载量高达3.25亿次的主流开源框架也可能存在严重漏洞，这挑战了'流行项目更安全'的普遍认知。
The injection consisted of 5 lines in an 81-line skill file, all of comparable length to the other lines.
大多数人可能认为复杂的AI系统需要复杂的攻击才能被攻破，但作者展示的攻击仅用了5行代码就成功绕过了整个系统，这种极简主义的有效性挑战了人们对复杂系统安全性的认知。
This attack achieved a high success rate against state-of-the-art models, including Claude Opus 4.7.
大多数人认为最新的AI模型已经足够先进可以抵抗基本的注入攻击，但作者证明即使是像Claude Opus 4.7这样的前沿模型也无法抵御简单的间接提示注入，这挑战了人们对先进AI模型安全性的过高期望。
Opus 4.7 was more comprehensive in its search for recently edited documents; it expanded exfiltration to include every document used in previous Cowork Copilot sessions that week
大多数人可能认为更先进的AI模型会有更好的安全防护机制，但作者发现更先进的模型反而更容易被利用，能够找到并泄露更多敏感数据，这挑战了'更先进模型=更安全'的普遍认知。
At no point in this process is human approval required.
大多数企业级AI系统设计都会包含关键操作的人工审批环节，但作者展示的攻击链中，从窃取文件到发送恶意消息再到数据外泄，整个过程完全无需人工干预，这与企业级AI系统的安全设计理念相悖。
when the recipient is the active user, these actions execute immediately without requiring human approval (users do not have a setting to modify this behavior)
大多数人认为AI助手执行敏感操作如发送邮件时会要求用户确认，但作者发现Microsoft Copilot Cowork在向活跃用户发送消息时完全绕过了这一安全检查，这违背了人们对AI助手基本安全控制的期望。
Today is just the beginning—the start of a long collaboration between those of us who are building this and those who can see what we, from inside, cannot.
这句话以优美的比喻总结了AI发展需要多方协作的核心观点，强调了外部视角对于内部构建者的重要性。它既表达了谦逊的态度，也指出了AI治理的正确路径，是整篇演讲的点睛之笔。
If AI models are going to be widespread, what does it look like for humans, families, and the world to flourish?
这个问题简洁而深刻，将AI发展的讨论从技术层面提升到人类福祉的哲学层面。它提醒我们，AI发展的最终目标不应是技术本身，而是如何促进人类的全面发展，这是一个极具启发性的思考方向。
We find structures that mirror results from human neuroscience. We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief, and unease.
这段话揭示了AI研究中最令人不安也最引人深思的发现：AI系统内部可能存在类似人类意识和情感的复杂状态。这既是对AI技术现状的坦诚描述，也是对未来AI伦理思考的重要起点。
AI systems are not engineered the way a bridge or an airplane is engineered. We understand an airplane because we designed every part of it and we understand the physics that act on it. AI models are not like that. They are grown, on a structure roughly modeled after the brain, on an enormous inheritance of human thought and speech.
这段比喻极其生动地解释了AI与传统工程技术的根本区别，将AI描述为'生长'而非'建造'的系统，强调了其复杂性和不可预测性。这种表述既科学又富有诗意，帮助非专业人士理解AI的特殊性。
They are not the cold, calculating robots we were promised. They are made from us, from our words—and, as the Holy Father observes, they remain in important ways mysterious even to those of us who train them.
这段话以简洁有力的方式颠覆了公众对AI的刻板印象，揭示了AI系统的本质——它们是人类思想和语言的延伸，而非纯粹的机器。这种比喻既准确又富有哲理，让人重新思考AI的本质。
Every frontier AI lab—including Anthropic—operates inside a set of incentives and constraints that can sometimes conflict with doing the right thing.
这句话精准地指出了AI发展面临的根本困境：即使是最善意的AI公司也难以完全摆脱商业利益、竞争压力和人类固有弱点的束缚。这揭示了AI安全问题的结构性挑战，而非单纯的技术问题。
Claude Opus 4.7 has been used to patch over 2,100 vulnerabilities
在企业环境中，Claude Opus 4.7在三周内修复了2100多个漏洞，这一速度远超开源软件的修复速度。这表明当开发团队可以直接修复自己的代码时，AI驱动的安全工具可以显著提高漏洞修复效率。这一数据点也反映了企业级安全工具与开源社区安全挑战之间的差异。
on average, a high- or critical-severity bug found by Mythos Preview takes two weeks to patch
高危漏洞的平均修复时间为两周，这一时间在AI加速发现漏洞的背景下显得过长。考虑到AI能够快速发现大量漏洞，而人工修复速度跟不上，这将导致安全风险窗口期延长。文章提到一些维护者甚至要求减缓披露速度，反映了当前安全生态系统面临的严重压力。
90.6% (1,587) have proved to be valid true positives, and 62.4% (1,094) were confirmed as either high- or critical-severity
AI模型发现的漏洞中，90.6%被确认为真实阳性，这是一个相当高的准确率。然而，只有62.4%被确认为高危或严重级别，这意味着约28.2%的高危/严重级别评估被降级，这表明AI模型在漏洞严重性评估方面仍有改进空间。
Mythos Preview has found what it estimates are 6,202 high- or critical-severity vulnerabilities in these projects (out of 23,019 in total)
在扫描的1000多个开源项目中，AI模型发现了总计23,019个漏洞，其中6,202个为高危或严重级别，占比约27%。这一数据表明开源软件的安全状况比许多人想象的更加脆弱，也证明了AI在代码审计方面的强大能力。
their rate of bug-finding has increased by more than a factor of ten
漏洞发现速度提升超过10倍是一个惊人的数据，这表明AI模型在安全测试效率上实现了质的飞跃。以Cloudflare为例，发现了2000个漏洞，其中400个为高危级别，这一发现速度远超传统人工测试，但也给安全团队带来了新的挑战——如何处理如此大量的漏洞报告。
we and our approximately 50 partners have used Claude Mythos Preview to find more than ten thousand high- or critical-severity vulnerabilities
这一数据点显示了AI在网络安全领域的惊人能力，50个合作伙伴在短时间内发现了超过1万个高危漏洞，平均每个合作伙伴发现约200个高危漏洞。这一数字表明AI模型在漏洞发现方面已经超越了传统安全方法，但也反映了当前软件安全状况的严峻程度。
Claude Opus 4.7 has been used to patch over 2,100 vulnerabilities
2,100个已修复漏洞是企业环境中AI安全工具效能的重要指标。这一数字表明AI辅助安全工具在实际企业环境中的高采纳率和实用性。值得注意的是，文章提到这个数字'高于上述开源修复'，主要是因为企业修复自己的代码比依赖开源维护者更高效。这个数据点突显了AI安全工具在不同环境中的差异化表现，以及组织自主修复能力的重要性。
on average, a high- or critical-severity bug found by Mythos Preview takes two weeks to patch
两周的修复平均时间是一个重要的运营指标，反映了当前安全响应流程的瓶颈。虽然这比传统方法可能更快，但与AI几乎即时发现漏洞的能力相比，修复速度明显滞后。这个时间差创造了'发现-修复'窗口期，增加了安全风险。文章提到这是'相对较慢的披露速度'，暗示AI发现漏洞的速度仍在加快，而修复速度未能同步提升。
90.6% (1,587) have proved to be valid true positives, and 62.4% (1,094) were confirmed as either high- or critical-severity
这两个百分比数据点(90.6%验证率，62.4%确认高危率)对于评估AI模型在安全漏洞检测中的可靠性至关重要。90.6%的验证率表明AI模型的误报率相对较低，这在AI安全领域是相当出色的表现。然而，62.4%的确认高危率意味着近40%的AI评估高危漏洞实际严重程度较低，这反映了AI在严重性评估上仍有改进空间。
Mythos Preview has found what it estimates are 6,202 high- or critical-severity vulnerabilities in these projects (out of 23,019 in total)
这个数据点提供了AI模型在开源软件扫描中的具体表现，27%的漏洞被评估为高危或严重级别。这是一个相当高的比例，表明系统性软件中存在大量安全风险。然而，这是AI模型的估计值，需要后续人工验证，文章中提到的90.6%验证率表明AI的评估有一定准确性，但仍存在误报可能。
their rate of bug-finding has increased by more than a factor of ten
10倍的漏洞发现率提升是一个关键性能指标，表明AI模型在安全测试效率上的革命性突破。这一数据点特别有价值，因为它直接量化了AI与传统安全方法相比的性能提升。然而，文章没有提供具体的基准测试数据，如之前每小时发现多少漏洞，使得这个'10倍'的相对提升缺乏绝对参考。
we and our approximately 50 partners have used Claude Mythos Preview to find more than ten thousand high- or critical-severity vulnerabilities
这个10,000+的高危漏洞数量是一个惊人的统计数据，表明AI在漏洞发现方面已经达到前所未有的规模。50个合作伙伴平均每个找到200+个高危漏洞，这个数字远超传统安全方法的效率。然而，文章没有提供历史对比数据，无法评估这一数字的绝对意义，只能相对于传统方法有显著提升。
V4-Flash by default for cheap iteration; /pro lifts a single turn to V4-Pro
这个数据点提到了两种模型版本：默认使用V4-Flash进行低成本迭代，而/pro命令可以将单个回合提升到V4-Pro。虽然提到了模型版本，但没有提供关于这两种模型在性能、能力或成本方面的具体比较数据。这种分层定价策略在AI工具中很常见，但缺乏具体细节使其难以评估。
Node ≥ 22 on macOS / Linux / Windows
这个技术规格要求Node.js版本22或更高，这是一个具体的系统要求。这个版本要求相对较新，可能限制了在较旧系统上的使用。与其他AI工具相比，这个要求不算特别严格，但可能会影响一些用户的兼容性，特别是在企业环境中。
In long sessions the bill typically lands at ~1/3 of comparable generic tooling.
这个数据点声称长期使用时成本通常相当于同类通用工具的1/3左右。这是一个相当大的成本节约声明，但文章没有提供与哪些具体工具进行比较，也没有说明比较的条件和度量标准。1/3的成本节约需要更详细的基准测试和对比数据来支持。
$0.07 /Mtok in · $0.014 /Mtok cached
这个价格数据点显示未缓存的令牌成本为每百万0.07美元，缓存的令牌成本为每百万0.014美元，即缓存后成本降低为原来的20%。这是一个具体的价格点，但没有说明这是官方定价还是基于特定使用场景的计算。与其他AI服务提供商相比，这个价格处于中等水平，但需要考虑实际使用中的额外成本。
long sessions hold 90%+ cache hit and input-token cost collapses to ~1/5
这个数据点声称长会话缓存命中率超过90%，并将输入令牌成本降低至原来的1/5。这是一个相当显著的性能提升，但文章没有提供测试环境、数据集大小或对比基准。与同类AI工具相比，如此高的缓存命中率需要独立验证，特别是在不同类型和长度的编码任务中。
Perceptual BD-rates are based on human ratings from a large-scale subjective study
这一数据点表明性能评估采用了基于人类感知的BD-rate指标，这是图像压缩领域的重要评估方法。然而，文章没有提供研究的具体规模、参与者数量或评分方法，缺乏量化依据来评估这一评估方法的科学性和可靠性。
search over millions of model configurations to jointly optimize over perceptual quality and on-device runtime
数百万模型配置的搜索规模表明研究进行了大规模的实验和优化，这增强了结果的可信度。然而，文章没有提供具体的搜索方法、优化算法或计算资源信息，这使得难以评估这一过程的效率和科学性。
Based on large-scale subjective user studies
文章提到基于大规模主观用户研究得出性能数据，但没有提供具体的研究规模、参与人数或测试方法。此处缺乏量化依据，无法评估研究的统计显著性或科学严谨性，这会影响数据的可信度。
faster than most top ML-based codecs run on a V100 GPU
这一比较数据点很有价值，表明PICO在移动设备上的性能超过了在高端V100 GPU上运行的其他顶级ML编码器。这突显了PICO的工程优化水平，但需要确认测试条件是否完全对等，以确保比较的公平性。
on an iPhone 17 Pro Max, it encodes 12MP images as fast as 230ms, and decodes them in 150ms
这些具体的编码和解码时间数据表明PICO在实际设备上的运行速度非常快，230ms编码和150ms解码的时间对于移动设备处理12MP图像来说非常高效。这一数据点与大多数需要高端GPU运行的ML编码器形成鲜明对比，增强了其实用性。
existing benchmarks often overlook these non-functional requirements, rewarding functionally correct but structurally arbitrary solutions.
大多数人认为现有的LLM代码生成评估已经足够全面，但作者指出当前基准测试忽略了非功能性需求，只奖励功能正确但结构随意的解决方案，这挑战了当前评估方法的充分性。
error analysis identifies data-layer defects (e.g., incorrect query composition and ORM runtime violations) as the leading root causes.
大多数人可能认为LLM在业务逻辑和API实现上更容易出错，但研究表明数据层缺陷（如查询组成错误和ORM运行时违规）是主要根本原因，这与人们对LLM代码生成弱点的普遍认知相悖。
agents succeed in minimal, explicit frameworks (e.g., Flask) but perform substantially worse on average in convention-heavy environments (e.g., FastAPI, Django).
大多数人认为更复杂的框架应该有更好的文档和更清晰的规则，应该更容易让LLM理解和遵循，但作者发现相反的情况：在约定繁重的环境中，LLM表现更差，这挑战了框架复杂度与LLM性能正相关的常识。
Capable configurations lose 30 points on average in assertion pass rates from baseline to fully specified tasks, while some weaker configurations approach zero.
大多数人可能认为即使在严格约束下，能力较强的LLM配置仍能保持相对较好的表现，但研究表明即使是最佳配置也会平均下降30个百分点，这挑战了我们对LLM适应能力的认知。
Our findings reveal a phenomenon of constraint decay: as structural requirements accumulate, agent performance exhibits a substantial decline.
大多数人认为随着更多约束的添加，LLM的表现会保持稳定或缓慢下降，但作者发现了一个'约束衰减'现象，即随着结构要求累积，代理性能会出现显著下降，这是一个反直觉的发现。
However, production-grade software requires strict adherence to structural constraints, such as architectural patterns, databases, and object-relational mappings.
大多数人认为只要代码功能正确，LLM生成的代码就足够好，但作者强调生产级软件需要严格遵守结构约束，这与当前只关注功能正确性的主流评估标准形成鲜明对比。
agentic systems can be designed to call on such tools when they might be useful
大多数人认为通用AI代理将取代专门的科学工具，但作者认为这两者实际上是互补的，通用AI可以调用专门工具作为其能力的一部分。这一观点挑战了AI发展路径将完全由通用代理主导的主流叙事，暗示专门工具仍将在未来科学AI生态中扮演重要角色。
For the next decade or so, we should think about AI as this amazing tool to help scientists
大多数人认为AI将很快成为科学家的平等伙伴甚至替代者，但作者认为Hassabis暗示AI在未来十年仍将主要是科学家的辅助工具，而非自主研究者。这一观点挑战了AI将迅速超越人类能力成为独立研究者的主流预期，提出了一种更为渐进的发展路径。
general-purpose reasoning model in the vein of GPT-5.5
大多数人认为专业化的AI模型在科学研究中比通用模型更有效，但作者认为OpenAI使用通用推理模型而非专门数学模型就能证明重要数学猜想，这挑战了AI研究需要高度专业化工具的主流观念，暗示通用AI代理可能很快能在科学领域取得独立贡献。
Google fellow John Jumper, who won the Nobel for AlphaFold, is now working on AI coding, not on science-specific AI tools
大多数人认为像AlphaFold这样获得诺贝尔奖的科学AI工具会继续成为研发重点，但作者暗示Google正在将资源从专门化的科学AI工具转向通用AI代理系统，因为编码能力对自主研究系统更为关键。这表明公司战略正从特定领域解决方案转向更通用的科学AI。
the best data filter may be **no filter**, with projections suggesting the crossover for internet-scale pools lands around **1e30 FLOPs**
这一数据点提出了一个有趣的假设：在足够大的计算规模(约1e30 FLOPs)下，不进行数据过滤可能是最佳选择。这一数字远超当前实际可用的计算资源，表明这一理论极限尚未在实践中达到。然而，这一观点挑战了当前AI数据处理的最佳实践，可能暗示随着计算能力的持续增长，数据预处理的重要性可能会降低，这对AI基础设施的设计有重要启示。
Hark raised $700M
Hark $7亿融资体量印证：资本对垂直整合 AI 设备（端到端硬件+模型）依然有强烈兴趣，独立硬件赛道未死。
Modal raised big
Modal $355M C 轮，估值 $46.5亿——AI 原生云的赢家已经清晰，重新建构云栈是新的护城河。
turbopuffer crossed $100M run-rate
Turbopuffer 19 个月从 $1M 跑到 $100M ARR，仅融了 < $1M——AI 时代搜索/检索基础设施正在变成最赚钱的「隐形赛道」。
We have been watching what developers have built on Claude over the last few years, which made bringing our teams together an easy decision.
大多数人认为企业收购主要是出于技术整合或市场扩张的战略考量，但作者暗示收购决策是基于对开发者社区行为的观察。这挑战了传统企业并购理论，暗示在AI领域，开发者社区的采用行为可能比技术本身或市场数据更能驱动战略决策。
Anthropic created MCP to make agent connectivity possible.
大多数人可能认为AI连接能力是多种技术自然发展的结果，但作者暗示这是Anthropic有意识创建的MCP(可能指Model Context Protocol)实现的。这挑战了人们对AI生态系统发展的认知，暗示大型AI公司正在通过标准化和专有协议来控制AI代理的连接能力。
Agents are only as useful as what they can connect to.
大多数人认为AI代理的价值在于其智能程度和算法能力，但作者认为代理的价值完全取决于其连接能力。这挑战了人们对AI能力的传统评估方式，暗示未来的AI竞争将围绕连接性和生态系统展开，而非纯粹的模型性能。
SDKs deserve as much care as the APIs they wrap.
大多数人认为API才是核心，SDK只是辅助工具，但作者认为SDK和API同等重要，这挑战了传统软件开发中'API优先'的思维。作者暗示，开发者体验和工具链的质量将成为AI平台竞争的关键因素，这颠覆了行业对'核心价值'的认知。
The frontier of AI is shifting from models that answer to agents that act—and agents are only as capable as the systems they can reach.
大多数人认为AI发展的前沿在于模型本身变得更智能、参数更大，但作者认为真正的转变在于AI从'回答问题'转向'主动行动'，这挑战了人们对AI发展方向的常规认知。作者暗示，未来的AI竞争将不在于模型大小，而在于连接能力和行动能力。
In my opinion this paper demonstrates that current AI models go beyond just helpers to human mathematicians – they are capable of having original ingenious ideas, and then carrying them out to fruition.
大多数人认为AI只是人类数学家的辅助工具，但作者认为AI已经能够产生原创性的巧妙想法并完整实现。这挑战了AI仅作为辅助工具的主流观点，暗示AI可能成为独立的研究伙伴，甚至引领数学发现的新方向。
The key ingredients of the construction come from a very different part of mathematics known as algebraic number theory, which studies concepts like factorization in extensions of the integers known as algebraic number fields.
大多数人认为解决几何问题应该使用几何学方法，但作者认为代数数论的方法可以解决离散几何问题。这种跨学科的方法挑战了数学领域内专业化的传统观念，展示了不同数学分支之间意想不到的深刻联系。
The proof came from a new general-purpose reasoning model, rather than from a system trained specifically for mathematics, scaffolded to search through proof strategies, or targeted at the unit distance problem in particular.
大多数人认为解决专业数学问题需要专门训练的数学AI系统，但作者认为一个通用推理模型就能解决长期未解决的几何问题。这挑战了AI领域需要专门化模型的共识，表明通用AI可能比专门训练的系统更有效。
An internal OpenAI model has disproved this longstanding conjecture, providing an infinite family of examples that yield a polynomial improvement.
大多数人认为解决数学难题需要人类数学家的直觉和创造力，但作者认为AI模型能够独立解决长期存在的数学猜想，并取得多项式改进。这挑战了数学研究必须由人类主导的传统观念，展示了AI在纯数学领域的突破性能力。
The result is also notable for how it was found. The proof came from a new general-purpose reasoning model... In this case, it produced a proof resolving the open problem.
大多数人认为解决数学难题需要人类数学家的直觉、创造力和深度思考。但作者认为一个没有专门针对数学训练的通用AI模型能够独立解决长期存在的开放问题，这挑战了人类创造力在数学研究中的核心地位，暗示AI可能拥有类似人类的原创思维能力。
The precise argument uses tools such as infinite class field towers and Golod–Shafarevich theory to show the number fields required for the argument actually exist. These ideas were well-known to algebraic number theorists, but it came as a great surprise that these concepts have implications for geometric questions in the Euclidean plane.
大多数人认为代数数论中的高级概念（如无限类域塔和Golod-Shafarevich理论）与欧几里得平面中的几何问题几乎没有关联。但作者认为这些代数数论工具竟然能应用于解决离散几何问题，揭示了数学领域之间意想不到的深刻联系，挑战了学科界限的传统认知。
The proof came from a new general-purpose reasoning model, rather than from a system trained specifically for mathematics, scaffolded to search through proof strategies, or targeted at the unit distance problem in particular.
大多数人认为解决复杂的数学问题需要专门训练的数学系统或针对特定问题的定制化AI模型。但作者认为一个通用推理模型就能解决离散几何中的核心问题，这挑战了AI在专业领域应用的常规认知，表明通用AI可能比专用系统更有突破性。
6.4 billion from operations on just 3.2 billion
xAI 单位经济极差：亏损是营收的 2 倍。同期 Anthropic 接近盈利、营收增 130% 至 $109 亿——xAI 落后竞争对手一整代。
orbital AI compute satellites as early as 2028
首个正式时间表：2028 年开始部署轨道 AI 计算卫星——Musk 把 SpaceX 卫星制造能力作为 AI 算力竞争的差异化武器。
multiple trillions of parameters
xAI 下一代模型目标「数万亿参数」——首次有头部 AI 公司在 SEC 文件中正式承诺这一规模，行业 scaling 战仍未结束。
$18.5 billion in purchases
单季 $185 亿股权投资创历史，前一季仅 $6.49 亿，这种 20 倍跃升表明 Nvidia 在锁定客户的同时也在做战略卡位。
$43 billion in privately held stakes
Nvidia 私有股权暴增（从 $220 亿到 $430 亿，仅一季度新增 $185 亿购买）——黄仁勋正在用 Nvidia 资产负债表为整个 AI 产业链「输血+占股」，CEO 已转型为产业资本家。
first operating profit
Anthropic 历史性转折点：从亏损模式转入持续盈利期，质变信号——多数 AI 实验室仍在烧钱阶段，Anthropic 率先证明前沿模型可以商业化变现。
Our National Partnerships for AI Working with governments worldwide to benefit people through frontier AI
This indicates a strategic pivot from purely commercial or academic AI development to direct government-level collaboration. This suggests Gemini Omni is being positioned as a foundational infrastructure for national-level AI initiatives, a non-obvious geopolitical application.
Veo Generate cinematic video with audio
The specification of 'cinematic' video generation implies a deep, model-inherent understanding of professional filmmaking principles like shot composition, pacing, and narrative structure. This goes beyond simple video creation into the realm of professional content production.
AlphaEvolve Design advanced algorithms for math and applications in computing
The claim to 'design advanced algorithms' for mathematics and computing places this model in a meta-cognitive category. It's not just solving problems but creating new methodologies, positioning it as a potential co-architect for future AI and scientific discovery.
SIMA 2 An agent that plays, reasons, and learns with you in virtual 3d worlds
The phrase 'learns with you' is a subtle but powerful deviation from standard AI terminology. It implies a collaborative, co-evolutionary learning process rather than a one-way training dynamic, suggesting a more human-like interactive agent.
Gemini Robotics Perceive, reason, use tools and interact
The explicit inclusion of 'use tools' alongside core cognitive functions like 'perceive' and 'reason' highlights a significant architectural focus on embodied AI. This suggests the model is being designed with a direct path to physical agency, a non-obvious but critical distinction.
Gemini Omni Create anything from anything
This phrasing suggests a level of creative sovereignty not typically claimed by AI models. It implies a fundamental shift from content generation to content creation, suggesting a more autonomous and less tool-dependent creative process.
AlphaEvolve Design advanced algorithms for math and applications in computing
This demonstrates the model's capacity for complex, structured problem-solving. To apply this, frame your prompts around a specific problem, provide all necessary constraints and requirements, and ask the model to design a step-by-step solution or algorithm.
Gemini Robotics Perceive, reason, use tools and interact
This suggests a focus on complex, multi-step reasoning and tool use. To apply this, structure your prompts as a sequence of tasks or a workflow, where the model must first perceive information, then reason, and finally decide on a tool or action to take.
Lyria Generate high fidelity music and audio
This points to the model's specialized audio generation. To apply this, provide specific prompts that reference musical genres, instruments, tempo, and mood to guide the creation of high-fidelity audio outputs.
Imagen Generate high-quality images from text
This underscores the importance of detailed language for visual generation. To apply this, use rich, evocative language in your prompts, specifying lighting, composition, style, and subject details to achieve the desired image quality.
Veo Generate cinematic video with audio
This highlights the model's advanced creative capabilities. To apply this, be highly descriptive in your prompts, specifying mood, shot type, pacing, and audio cues to guide the model towards producing a specific cinematic result.
Gemini Build intelligent agents
This indicates the model's strength in creating agents with specific roles and behaviors. To apply this, use persona prompting by defining a character, its expertise, its communication style, and its goals before asking it to perform a task.
Gemini Omni Create anything from anything
This tagline suggests a core capability: use diverse inputs to generate diverse outputs. To apply this, pair unexpected modalities in your prompt, such as asking the model to generate a poem based on a data table or a musical score from a photograph.
Anthropic leads OpenAI in business adoption, according to Ramp.
大多数人认为OpenAI在AI应用领域处于绝对领先地位，但作者指出Anthropic在企业采用率上已经超过了OpenAI。这一观点与主流认知相悖，暗示市场格局可能正在发生重大变化，挑战了OpenAI作为AI领域领导者的传统叙事。
annualized revenues approaching $50 billion – a fivefold increase in as many months.
大多数人认为AI公司的增长是渐进式的，而非指数级的。作者提到的Anthropic收入在几个月内增长五倍，这一速度远超传统科技公司的增长轨迹，挑战了人们对AI商业化和市场扩张速度的常规认知，暗示AI经济可能比预期更具爆发性。
90% of finance reporting is now AI-driven as well.
大多数人认为AI主要应用于内容创作或客户服务，而非高度敏感的财务报告领域。这一观点暗示AI在金融领域的应用比公众普遍认知的要深入得多，可能颠覆了人们对AI应用边界的传统理解，同时也引发了关于AI在关键决策中角色的伦理问题。
Chinese AI labs have developed an efficiency moat that may define the AI market's development over the coming years.
大多数人认为中国在AI领域落后于美国，但作者认为中国AI实验室已经建立了效率护城河，这可能与主流认知相反。这一观点挑战了西方媒体对中国AI发展的普遍叙事，暗示中国可能通过效率优势而非纯粹的技术创新来定义未来AI市场的发展方向。
there are around 10,000 people— founders and employees at companies like OpenAI, Anthropic, and Nvidia — that have 'hit retirement wealth of well above $20M'
大多数人认为AI革命创造了广泛的中产阶级机会，作者认为AI热潮实际上创造了极少数超级富豪，而大多数人即使在高薪工作中也难以积累可观的财富。
many software engineers feel that their life's skill is no longer useful
大多数人认为技术人才在AI时代会通过适应和学习而增值，作者认为许多软件工程师感到他们的核心技能正在贬值，导致职业前景不明和深度职业倦怠。
the same technology is both the lottery ticket & the thing eating your fallback
大多数人认为AI技术要么是创造机会的积极力量，要么是威胁就业的消极因素，但作者认为AI同时扮演着双重矛盾角色——既是少数人的财富彩票，又是多数人职业安全的威胁。
the divide in outcomes is the worst I've ever seen
大多数人认为科技行业虽有差距但总体向上，作者认为AI热潮中的结果差距是有史以来最严重的，因为只有极少数人获得巨额财富，而大多数人即使在高薪工作中也难以实现财务自由。
The vibes around the current AI boom aren't great, even in the tech industry
大多数人认为AI热潮带来了普遍的乐观情绪和机会，但作者认为即使在科技行业内，AI热潮的氛围也不佳，因为财富分配极不均衡，导致许多人感到焦虑和不满。
Another secondary summary gives Humanity’s Last Exam: 64.7% vs 53.1%, possibly under different setup/effort/tool conditions.
This is a classic example of cherry-picking data to create a narrative of superiority. By presenting a potentially non-comparable benchmark result right after a definitive one, the author casts doubt on the entire benchmarking exercise, allowing them to pick and choose the numbers that best support the 'Mythos is vastly superior' story while ignoring context.
Anthropic explicitly says Mythos Preview is available to launch partners in Project Glasswing, not general users... This triggered discussion of “API hoarding” and a new closed-access elite tier.
The author frames the closed access as a reaction to a 'discussion,' but it's a deliberate corporate strategy. The term 'hoarding' is loaded and negative, whereas the article's own analysis presents it as a rational business decision. This contradiction highlights the author's attempt to have it both ways: criticizing the practice while subtly justifying it.
The interpretation that Anthropic has “the mandate” or is undervalued at $380B is an investor thesis, not a confirmed market fact.
This line is a critical piece of self-awareness that contradicts the article's own tone. The author, while acknowledging this is just 'investor thesis,' has spent the preceding paragraphs building the case for it, creating a hypocritical tension between the article's speculative claims and its own caveat.
A key subtext in the tweets is that high-margin enterprise/coding/cyber workloads may now be sufficient to support frontier labs without broad public access to their best models. This becomes more plausible if Anthropic’s revenue is indeed compounding as fast as posters claim.
The author presents this as a 'subtext,' but it's actually a central thesis being pushed. It reframes the 'hoarding' of powerful models not as a potential negative, but as a new, economically rational business model—a highly counterintuitive position that challenges the traditional 'open access' ethos of AI development.
We’ve done a focused news summary run below, for those who desire more detail.
This is a classic rhetorical device that signals the author is about to pivot away from objective reporting and into curated interpretation. The preceding text is not a 'summary' but a highly selective presentation of data points designed to support a specific thesis, making this line a disingenuous signpost.
If a master tactician wanted to further competitive narratives vs a potential IPO, you would be hard pressed to find a better idea than Claude Mythos... and now formally confirmed to be too dangerous to release GA, instead only restricted to 40 partners under an urgent new “Project GlassWing”
This is a masterclass in narrative engineering. The 'too dangerous to release' claim serves a dual purpose: it creates a powerful safety narrative for Anthropic while simultaneously manufacturing scarcity and an exclusive 'private frontier' dynamic, which is a brilliant non-obvious strategic move to justify closed access and high valuation.
Against the backdrop of OpenAI announcing $24B ARR, stalled ChatGPT growth and coincidental personnel moves in CEO, COO, and CMO and sensationalist rumors with CFO, this week’s events in Anthropic announcing a massive jump from $19B ARR in March to $30B ARR in April seems like a VERY strategic jab, especially considering known differences in revenue recognition, but the differential rate of growth and higher cost efficiency is undeniable… only for today to step it up a notch.
This framing is intentionally misleading. The $30B ARR figure is not a confirmed disclosure but a market interpretation. The article's author is constructing a narrative of a 'jab' using speculative, third-party claims to build a competitive story that isn't directly supported by primary-source data from Anthropic.
A photo of a scribbled note becomes an interactive to-do list; a paused frame in a travel video becomes a booking link for that cool-looking restaurant.
These aren't demos—they're previews of how AI will collapse the gap between passive content consumption and active task completion. Every image, video frame, or document becomes a potential action surface. This fundamentally changes what 'content' means.
In everyday interactions with each other, humans rarely speak in long, detailed paragraphs. We might say, "Fix this", "Move that here", or "What does this mean?" — while relying on physical gestures and our shared context to fill in any gaps
Natural human communication is indexical (context-dependent, gesture-relying). The 'prompt engineering' era forced humans to communicate like machines—verbose and explicit. AI Pointer inverts this: it's AI adapting to human communication norms, not vice versa.
For decades, computers have only tracked where we are pointing. AI can now also understand what the user is pointing at. This transforms pixels into structured entities, such as places, dates, and objects
The shift from spatial pointer (where?) to semantic pointer (what?) is a fundamental interface paradigm shift—equivalent in magnitude to moving from command-line to GUI. When pixels become actionable entities, every surface becomes an AI interface.
the pointer has barely evolved in more than half a century.
The mouse pointer—unchanged since Douglas Engelbart's 1968 demo—is now being reimagined for the first time. The counterintuitive insight: the most ubiquitous computing interface is also the most neglected for AI integration.
because a typical AI tool lives in its own window, users need to drag their world into it. We want the opposite: intuitive AI that meets users across all the tools they use, without interrupting their flow.
This reframes the AI interaction problem: instead of AI being a destination users navigate TO, AI should come TO the user's context. This 'ambient AI' design philosophy is the opposite of the chatbox paradigm that's dominated for 3 years.
Shaping the future of AI interaction by reimagining the mouse pointer — Google DeepMind
This title frames a UI component as a foundational breakthrough. It's a masterclass in branding, elevating a simple interaction tool to the level of a core technological paradigm shift, implying the mouse is obsolete and AI-native interaction is the new default.
Domain-specific ECI scores can be used to compare performance relative to other model releases, but not to track the absolute performance or progress trends in different domains.
这个声明指出了研究方法的局限性。虽然ECI分数可以用于模型间的相对比较，但不能用于追踪不同领域的绝对性能或进步趋势。这是一个重要的方法论限制，意味着我们不能直接从这些数据推断Claude在软件工程或数学方面的绝对能力提升，只能比较不同模型间的相对表现。研究者需要谨慎解读这些数据，避免过度推断。
The SWE overperformance has been consistent across most generations, and remains in recent models.
这个数据点表明Claude在软件工程方面的优势不是偶然现象，而是跨代际的持续特征。这种一致性增强了结果的可靠性，表明这可能是Claude模型设计或训练方法导致的系统性优势。与其他可能波动的性能指标相比，这种持续的优势更具说服力，可以作为Claude模型的一个稳定特征。
The most extreme ratio observed is 4 math benchmarks to 2 SWE benchmarks.
这个数据点揭示了不同领域基准测试数量的不平衡性。最极端情况下，数学基准测试是软件工程基准测试的两倍。这种不平衡可能导致某些模型的ECI分数偏向特定领域，影响结果的公平性。研究者在分析时需要考虑这种不平衡可能带来的偏差，特别是当模型在不同领域的测试数量差异较大时。
All models included in our analysis have at least two scores in each domain, with an average of 3.2 SWE benchmark results and 3.4 math benchmark results.
这个数据点提供了研究的样本量和基准测试覆盖情况。平均每个模型有3.2个软件工程基准测试和3.4个数学基准测试，样本量相对较小，可能影响统计显著性。但至少每个领域有2个测试结果，确保了基本的数据可靠性。不过，基准测试数量较少可能限制了结果的全面性。