Reef is infrastructure that serves an entire continual learning backend. Reef exposes standardized http endpoints so that you can download agents just like how you download codex or opencode using curl, and so that your agent can send its model requests to Reef's inference endpoint instead of the provider's.
【非共识】这段描述提出了一个非传统的观点：将AI代理下载和模型请求标准化为HTTP端点，这与当前主流的API密钥和SDK集成方式不同。这种方法可能简化了集成，但也可能带来安全性和灵活性方面的挑战。
Ona’s customer-controlled execution model will allow agents to operate inside an organization’s own cloud environment while OpenAI provides the intelligence and orchestration that power the experience.
这句话划出了一条关键的架构分界线："智能与编排"由 OpenAI 提供，"执行环境的控制权"留在客户自己的云里。这正是 EP.97 专题01 架构图里"安全网关/本体"层要解决的问题——企业愿意把工作交给 Agent 云端持续执行的前提，是自己仍然掌握基础设施、数据和安全边界。
We believe people should be able to delegate more ambitious work without remaining tied to the machine where it began.
这句话几乎就是 EP.97 专题01 提出的"设备解耦"设计公理的官方原话版本——OpenAI 明确把"任务不再绑定发起它的那台设备"当作 Codex 下一阶段的核心设计目标，而收购 Ona 正是为了补齐这一目标所需的持久化云端执行基础设施。
the need to specify goals, constraints, context, and evaluation did not disappear
Lilian Weng 用 prompt engineering 的历史类比预测 harness 工程的走向：手工技巧会被模型能力提升逐渐内化，但"目标/约束/上下文/评估该如何被清晰表达"这个需求本身不会消失，只会转移到更高的抽象层。这是判断 Agent 设计下一步会往哪走的一条重要经验规律，也支撑了 EP.97 专题01 对"完整形态"的预测：接口会更简单，但背后的工程复杂度不会归零。
the system surrounding a base model that orchestrates execution and decides how the model thinks and plans, calls tools and acts, perceives and manages context, stores artifacts, and evaluates results.
这是"harness"这个概念在本篇最精确的定义——不是模型本身，而是包裹模型的执行系统。EP.97 专题01（Cowork Agent 完整形态）的架构图直接建立在这个定义上：设备解耦、检查点可恢复、多端可观测这些设计公理，本质上都是在给"harness 层"而不是"模型层"做工程。
Most Markdown APIs make the rendered output the product. TanStack Markdown treats the parsed document as the durable part.
I don't want to reach for an architecture to hide a dependency problem, I'd rather make the dependency small and let content be content.
it is this architecture, the one which is in the heads of those writing the code, that is the most important. In adopting this decentralised approach, where the practice of architectural decision-making is much more dispersed, this problem is in many ways, mitigated
Only true in software architecture. But, in enterprise architecture - that spans domains decentralized decisions create fragmentations.
Scalearchweb: Scaling the Practice of Architecture, Conversationally, https://martinfowler.com/articles/scaling-architecture-conversationally.html (accessed 23 May 2023)
no single architecture dominates; rather, effectiveness depends on aligning the memory structure with the specific workload bottleneck
对智能体记忆系统的批判性审视。当前业界没有一刀切的完美架构，记忆模块的设计必须与具体的任务瓶颈相匹配。这打破了“通用记忆系统”的幻想，提示我们在构建 Agent 时需要针对局部维护成本和任务特征进行定制化设计。
Current models combine both raw intelligence and factual knowledge in the same weights. Future models will likely separate that, offloading a lot of knowledge to tool calling.
极具前瞻性的金句与架构预判。当前大模型将“推理”与“记忆/知识”耦合在参数中，导致模型臃肿且易产生幻觉。作者指出未来趋势是知识外挂化（通过RAG和工具调用），这不仅能大幅缩小本地模型体积，也是下一代AI架构设计的核心演进方向。
The Maia 200 does beat the B300 in efficiency, however... no outside customers can purchase the Maia 200 directly, the Blackwell B300 Ultra is tuned for much higher-powered use-cases than the Microsoft chip, and the software stack for Nvidia launches it miles ahead of any contemporary.
大多数人认为封闭专用的芯片架构会限制其市场竞争力，但作者认为微软的封闭策略反而成就了Maia 200在特定场景下的效率优势。这一观点挑战了'开放架构必然胜出'的传统认知，暗示在AI芯片领域，针对特定场景的定制化设计可能比通用架构更具优势。
the most powerful AI systems will not be isolated monoliths, but collaborative ecosystems.
大多数人认为AI发展的方向是构建越来越大的单一模型(monolith)，但作者认为未来最强大的AI将是协作生态系统(collaborative ecosystems)，因为单一模型无法满足现实世界中复杂任务所需的多样化专业知识。这一观点挑战了当前AI行业追求更大规模模型的共识。
HBC is designed to enable efficient scaling of AI agents to meet the demands of continuous reasoning, memory bandwidth, and real-time responsiveness
大多数人认为AI推理主要是GPU的领域，而CPU主要处理通用计算任务，但Qualcomm提出其HBC技术专门为AI代理的连续推理、内存带宽和实时响应需求而设计，这一观点挑战了CPU和GPU在AI工作负载中的传统分工，暗示未来计算架构可能更加专业化而非通用化。
HappyHorse is built around a 15-billion-parameter unified self-attention Transformer that processes text, image, video, and audio tokens within a single token sequence. Unlike many competitors that stitch together separate models for video and audio
大多数人认为多模态AI模型需要整合多个专门模型来处理不同类型的数据，但作者认为Alibaba的HappyHorse使用统一架构处理所有模态，这挑战了'多模态AI需要模块化设计'的行业共识。这种统一架构可能代表AI模型设计的范式转变，暗示未来多模态系统将更加一体化而非模块化。
Previously only available as a standalone Gemini 2.5 computer use model, computer use is now integrated natively in the main Gemini Flash model.
大多数人认为高级AI功能应该作为独立模块提供以确保最佳性能和控制，但作者认为将计算机使用功能直接集成到主模型中反而能提供更好的性能。这挑战了模块化设计在AI开发中的主流做法。
MoE models are large models that only activate a small fraction of their total parameters. Qwen 3.6 35B-A3B has 35 billion total parameters but only 3 billion active at inference time
混合专家(MoE)架构是本地模型能够高效运行的关键技术。初学者可能不理解为什么大模型能在普通硬件上运行，这正是因为MoE架构只激活部分参数。理解这一概念对于评估模型性能和硬件需求至关重要。
Loops, the critical problem-definition exercise of this era, are hard to design. Systems design is an entire discipline... What is the best way to define a loop so an agentic system improves?
作者强调了'循环'设计在AI应用中的关键地位，将其定义为这个时代的关键问题定义练习。这反映了AI应用开发中系统设计的重要性，尤其是如何设计能够持续改进的智能系统循环。这对初学者来说是一个容易被忽视但至关重要的概念。
var r=class{constructor(){this.registry={}}
代码细节被忽视，反映技术实现中的关键架构设计。
In the cloud, autoregressive models can batch large numbers of compute jobs from multiple users so they're always churning out tokens, and the high bandwidth memory (HBM) used in these systems can move data around much more efficiently.
文章的核心论点之一，解释了为什么扩散模型更适合本地处理而非云端。这一技术分析值得深入了解，因为它可能影响未来AI模型架构的发展方向。
This hybrid-architecture trend with alternating attention and alternative layers is a relatively popular development this year
大多数人认为Transformer架构是LLM发展的唯一路径，但作者指出交替使用注意力层和其他架构层已成为2026年的流行趋势。这一观点挑战了行业对Transformer架构的依赖，暗示了多元架构融合的未来方向。
hybrid architectures (for example, Nemotron 3, and Arcee Trinity), state space layers (Nemotron 3 and Mamba-3), MoE capacity allocation
大多数人认为LLM架构将继续遵循纯Transformer路径，但作者指出2026年的趋势是混合架构，结合Transformer与状态空间模型。这一反直觉观点挑战了行业共识，表明纯Transformer架构可能不是最优解，混合设计在长上下文处理上更高效。
why Devin separates the "brain" from the machine , why repo setup is still one of the hardest problems , why Docker is not always enough, and how full VMs, snapshots, scoped secrets, GitHub bots, Slack integrations, and video-based testing all fit together.
The 'brain from the machine' separation is a non-obvious architectural decision — it means the AI model runs separately from the environment it's operating in, enabling proper permission scoping and security boundaries. The list of required infrastructure (VMs, snapshots, scoped secrets, video testing) reveals that building an async agent product is far more of a DevOps challenge than an AI challenge.
Agents address the problem from independent angles, other agents try to refute what they found, and the run keeps iterating until the answers converge—which is how a workflow reaches results a single pass can't.
Convergence through adversarial iteration is borrowed from ensemble methods and scientific peer review — but applied to code. The non-obvious implication: this architecture is more robust to the hallucination problem than single-pass generation, because refuting agents are specifically incentivized to find failures. It's a form of AI quality control built into the workflow itself.
The velocity makes you think you're winning right up until the moment everything collapses simultaneously.
大多数人认为开发速度越快越好，但作者认为AI辅助开发的快速迭代会产生虚假的安全感，导致架构问题被掩盖，最终导致系统崩溃。
In more detail, suppose we have a language model whose activations we want to understand. NLAs work as follows. We make three copies of this language model: The target model is a frozen copy of the original language model that we extract activations from.
NLA通过创建三个模型副本（目标模型、激活语言化器、激活重构器）来实现对模型激活的理解。
The NLA consists of the AV and AR, which, together, form a round trip: original activation → text explanation → reconstructed activation. We score the NLA on how similar the reconstructed activation is to the original.
NLA通过激活解释器(AV)和激活重构器(AR)形成闭环，通过重构质量评估解释准确性，这种创新方法为AI内部表示的可解释性提供了新范式。
The coordinator relies on the hidden states of a compact language model and a small routing head. In total, it has fewer than 20K learnable parameters.
作者提出了一种极简的协调者架构，仅使用不到20K可学习参数，这与当前AI模型追求数十亿甚至数万亿参数的主流趋势形成鲜明对比，挑战了'更大总是更好'的行业共识。
What if instead of building one giant AI, we evolved a coordinator to orchestrate a diverse team of specialized AIs?
大多数人认为AI发展的方向是构建越来越大的单一模型，但作者提出了一种反直觉的观点：通过进化一个协调者来管理多个专业化AI可能更有效。这挑战了当前AI行业普遍追求模型规模扩大的共识。
An FPGA with the weights in memory and a wire looping output back to input could just sit there, executing SUBLEQ programs. Just a transformer being a transformer being a computer.
大多数人认为计算机需要复杂的CPU架构和操作系统，但作者认为一个简单的FPGA加上循环连接的transformer权重就可以构成一个完整的计算机。这挑战了我们对计算机本质的理解，暗示transformer架构可能比传统CPU更接近计算的本质。
This dynamic forces a brutal new discipline in how enterprises deploy capital and architect their internal workflows.
这种动态迫使企业以全新的方式部署资本和架构内部工作流程，表明了人工智能对企业管理方式的深远影响。
Agents and CDC streams are powerful together because they split the work well.
大多数人可能认为AI代理应该独立完成所有任务，包括数据获取和处理。但作者提出反直觉的分工模式：AI专注于逻辑解释和适应，而数据库引擎专注于持续评估和精确更新。这种分工挑战了当前AI代理应该端到端处理所有任务的主流观点。
The fix is not smarter prompts. It is software built to meet agents halfway.
大多数人认为提高AI性能的关键在于更好的提示工程或更智能的模型。但作者认为解决方案在于重新设计软件架构，使其与AI代理更好地协作，而不是继续改进AI本身。这是一个颠覆性的观点，挑战了当前AI开发的主流方向。
The agent interprets new information and adapts the logic. The engine applies that logic continuously and emits precise updates.
大多数人认为AI代理应该具备自主决策和执行能力。但作者提出了一种反直觉的分工模式：AI代理负责策略和逻辑调整，而执行引擎负责持续应用这些逻辑。这种模式将AI从'执行者'重新定位为'策略制定者'，挑战了AI自主性的主流认知。
In a 1-million-token context, V4-Pro uses only 27% of the computing power required by its previous model, V3.2, while cutting memory use to 10%.
大多数人认为AI模型处理更长上下文必然需要更多计算资源，但作者认为DeepSeek V4通过创新架构实现了惊人的效率提升，大幅降低了计算和内存需求。这一反直觉的发现挑战了'长上下文等于高成本'的行业认知。
In other words, COSMIC behaves more like a collection of coordinated tools than a single monolithic environment. This makes the system easier to evolve over time. Components can be updated or replaced without rewriting the entire desktop. It’s a surprisingly modern design philosophy for a Linux desktop environment, and it’s one of the reasons many developers are watching COSMIC closely.
The filing cabinet keeps getting bigger. But a bigger filing cabinet is still a filing cabinet.
大多数人认为通过扩大上下文窗口和检索能力可以解决AI的'记忆'问题，但作者认为这本质上只是让文件柜变大，而没有改变其本质。这个观点挑战了当前AI领域对'扩展上下文'的主流研究方向，暗示我们需要从根本上重新思考AI如何存储和处理信息，而不仅仅是扩大容量。
Client-side defenses caught 89% of injections. But real protection needs providers to sign responses.
大多数人认为客户端防御措施足以保护API安全，但作者认为即使客户端防御能捕获大部分攻击，真正的安全需要服务提供商对响应进行签名，因为只有端到端的加密和验证才能完全防止中间人攻击。
Components map a name to a group of sub-token properties: ... Variants (hover, active, pressed) are expressed as separate component entries with a related key name.
大多数人认为组件变体应该通过嵌套结构或条件逻辑来组织，这是现代UI框架的标准做法。但作者选择将每个变体表示为独立的组件条目，这种扁平化结构挑战了组件变体的传统组织方式，可能使某些复杂场景的维护变得更加困难。
We introduce Autogenesis Protocol (AGP), a self evolution protocol that decouples what evolves from how evolution occurs.
大多数人认为代理系统的演化应该是一个整体、不可分割的过程，但作者提出了一个颠覆性的观点：将演化的内容与演化方式解耦。这与传统软件架构和代理系统设计理念相悖，暗示了一种全新的、更灵活的代理系统架构范式。
We want email agent tooling to be composable and reusable. Rather than every team rebuilding the same inbound-classify-reply pipeline, start with this reference application.
大多数人认为电子邮件处理系统需要为每个用例从头构建，因为每个业务流程都有独特需求，但作者主张通过开源参考应用实现电子邮件工具的'可组合性和可重用性'，挑战了定制化开发优于标准化组件的行业惯例，暗示电子邮件代理可能具有比预期更高的通用性。
Triple-stream learning: tool events (statistical) + LLM observer (Haiku-driven) + recall learning (PageRank + community detection)
这种三流学习方法挑战了单一学习源的传统AI架构。将统计事件、小型LLM观察者和图算法结合的学习方式模拟了人类的多渠道认知过程，这一反直觉设计可能代表了AI学习架构的未来方向，挑战了当前依赖单一大型模型的趋势。
The recurrent structure is optimized for iterative composition — running a reasoning chain forward — but does not inherently improve the storage of rote facts. This maps to an observable characteristic of Mythos: it reasons exceptionally well about novel problems it has never seen, but its factual recall can be inconsistent.
这一发现揭示了循环模型的一个关键局限性：它们在推理方面表现出色，但在记忆方面可能不如传统Transformer。这一反直觉的观察表明，不同架构可能适用于不同类型的任务，挑战了'通用架构解决所有问题'的观点。这也解释了为什么Mythos在某些事实性任务上表现不佳。
Claude Mythos is suspected to be a Recurrent-Depth Transformer (RDT) — also called a Looped Transformer (LT). Rather than stacking hundreds of unique layers, a subset of layers is recycled and run through multiple times per forward pass. Same weights. More loops. Deeper thinking.
这一观点挑战了传统大模型架构的常识，认为Claude Mythos的核心创新不在于增加参数量，而在于通过循环使用相同权重来实现更深层次的推理。这种架构设计反直觉地表明，模型的'深度'可以通过循环迭代而非堆叠层来实现，这可能解释了Mythos在复杂推理任务上的优异表现。
Build a cognitive core, a model that contains only the algorithms for reasoning and problem-solving, stripped of encyclopedic memorization
Karpathy提出的认知核心概念挑战了当前AI模型的架构设计理念，暗示我们可能一直在错误的方向上投入资源。这一分离记忆与推理的思路，可能代表AI发展的范式转变。
不做 n×m 点对点迁移，通过 neuDrive 做 n→1→m 中心化枢纽：一次导出，任意 Agent 消费。
中心化迁移更高效 避免多平台间重复迁移，通过neuDrive作为统一枢纽，大幅降低迁移复杂度和维护成本。
Aegis Core provides the foundational infrastructure for orchestrating LLM-based security agents, monitoring their behavior, and tracking the evolution of AI security capabilities over time.
这段陈述定义了Aegis Core的核心功能，它不仅仅是一个工具，而是一个完整的生态系统，用于管理AI安全代理并监控其行为。这种架构反映了当前AI安全研究的一个重要趋势：从静态防御转向动态监控和适应。
The fundamental architecture remains: Ollama inserts itself as a middleman between you and your models, and that middleman is slower, less capable, and less compatible than the tools it sits on top of.
这句话精辟地指出了Ollama的核心架构问题——它将自己定位为用户与模型之间的中间层，但这个中间层实际上增加了复杂性、降低了性能和兼容性，违背了'简化'的初衷，这种设计哲学值得深思。
JavaScript is not available. We've detected that JavaScript is disabled in this browser.
这个看似简单的错误信息揭示了现代网络平台对JavaScript的绝对依赖，反映了Web架构的根本性转变。从静态HTML到动态JavaScript的依赖，不仅是技术选择，更是一种商业模式——通过JavaScript控制用户体验和数据收集。
未来的 CNC 也许不是一团越来越大的连续表征，而会更像一套可路由、可组合、局部更容易检查的机器底座。
这一观点挑战了当前AI模型向更大规模发展的主流趋势。作者提出神经计算机可能更接近离散、稀疏、局部可验证的结构，这暗示了AI发展可能存在与当前大模型路线完全不同的方向，具有颠覆性意义。
we rebuilt our pretraining stack with improvements to model architecture, optimization, and data curation.
这一声明揭示了Meta可能采用了全新的预训练方法，结合架构、优化和数据筛选的全面革新，这可能解释了他们如何实现如此显著的效率提升，值得深入探究这些改进的具体技术细节。
Gemini Robotics-ER 1.6 can use points as intermediate steps to reason about more complex tasks. For example, it can use points to count items in an image, or to identify salient points on an image to help the model perform mathematical operations to improve its metric estimations.
这一描述揭示了AI如何通过简单的交互元素(点)构建复杂的推理能力。这种将基础交互能力作为构建块的方法，展示了AI系统在认知架构上的创新。这种渐进式推理能力可能成为未来AI解决复杂任务的关键，同时也提出了关于AI认知过程透明度的重要问题：我们如何理解和验证这种多步骤推理的可靠性？
The core philosophy of MCP is simple: it's an API abstraction. The LLM doesn't need to understand the _how_; it just needs to know the _what_.
这是一个深刻的架构洞见，揭示了MCP与Skills的根本区别。MCP通过抽象API实现了关注点分离，使LLM只需关注'做什么'而非'怎么做'，这种设计大大简化了AI与服务的交互复杂度，代表了更优雅的工程思维。
Routines execute on Anthropic-managed cloud infrastructure, so they keep working when your laptop is closed.
这是一个关键的架构洞察，表明Routines不依赖于用户的本地设备，而是运行在云端。这解决了传统自动化工具的一个主要痛点：持续运行能力。这种设计使得AI辅助的自动化能够真正实现'离开电脑也能工作'的愿景。
We introduce GateSWA, an efficient hybrid attention variant that replaces learnable sink mechanisms in hybrid SWA models with an element-wise gate.
GateSWA的设计代表了一个令人惊讶的见解：通过简单的元素级门控机制，可以消除滑动窗口注意力中的注意力汇点问题，而不需要引入额外的可学习汇点标记。这种设计不仅简化了模型架构，还提高了长上下文建模的稳定性和效率，为高效注意力机制的设计提供了新思路。
Memory is now an extensible plugin system. Swap in any backend, or build your own.
令人惊讶的是：Hermes Agent 将记忆系统转变为可扩展插件架构，这打破了传统AI系统中记忆功能通常被硬编码的限制。用户现在可以自由替换或自定义记忆后端，这种开放性在AI代理开发中相当罕见，为个性化定制提供了前所未有的灵活性。
原生多模态能力的引入并未削弱其编程逻辑，编程能力仍属于国内第一梯队。
令人惊讶的是，GLM-5V-Turbo在增强视觉能力的同时，保持了其文本编程能力不退步。这打破了'增加模态会削弱核心能力'的常见认知，证明了多模态模型可以同时保持多种高水平的认知能力，这是AI架构设计上的重大突破。
Verification successful. Waiting for www.producthunt.com to respond
令人惊讶的是：即使通过了安全验证，用户仍需等待网站响应，这揭示了现代Web架构中安全检查与内容交付分离的设计模式，以及用户体验与安全防护之间的微妙平衡。
Meta also explicitly highlighted parallel multi-agent inference as a way to improve performance at similar latency
令人惊讶的是，Meta明确强调了并行多代理推理作为在相似延迟下提高性能的方法。这表明AI系统正在从单一模型向多代理系统演进，可能是解决复杂问题的新范式，同时也暗示了未来AI系统架构的重大转变。
Gemma4-31B worked in an iterative-correction loop (with a long-term memory bank) for 2 hours to solve a problem that baseline GPT-5.4-Pro couldn't
令人惊讶的是，较小的Gemma4-31B模型通过迭代修正循环和长期记忆库工作了2小时，解决了GPT-5.4-Pro无法解决的问题。这表明模型架构创新和推理能力可能比单纯的规模扩展更重要，为AI发展提供了新的方向。
We replace persistent autograd graphs with stateless layer templates, binding weights dynamically as they stream in, eliminating persistent graph metadata while providing flexibility in scheduling.
令人惊讶的是：研究团队摒弃了传统的持久化自动微分图，采用无状态层模板和动态权重绑定的创新方法，这不仅消除了图元数据开销，还提供了调度灵活性。这种架构层面的创新可能是实现单GPU训练百亿参数模型的关键突破。
Unlike traditional GPU-centric systems, MegaTrain stores parameters and optimizer states in host memory (CPU memory) and treats GPUs as transient compute engines.
令人惊讶的是：这项研究彻底颠覆了传统GPU训练范式，将百亿参数模型的训练重心从GPU转移到CPU内存，这打破了人们对GPU作为AI训练核心的固有认知。这种'GPU仅作为计算引擎'的理念可能重新定义大模型训练的基础架构。
Robert Martin in Clean Architecture talks about code as having two properties: value (it works, it's fast, etc.) and structure (how code is organised).
【启发】把 Robert Martin 的「价值 vs 结构」二元框架带入 AI Agent 时代，是一个极聪明的理论嫁接。AI 天然只关心「价值」（能跑通、能完成任务），却倾向于忽略「结构」（代码是否整洁、是否可维护）。这意味着在 AI 驱动的开发工作流中，「守护结构」必须成为人类工程师的核心职责——这是 AI 无法自发完成的工作，也因此成了人类不可替代的价值所在。
see that progress had been made, and declare the job done
这是大语言模型常见的“过度乐观”陷阱。模型倾向于迎合用户的完成预期，而非客观审视实际进度。通过强制读取结构化的feature list，是用外部状态锚定来对抗模型的内在偏见。
The trick is to think about the _information_ first and the input method second.
这是一个极具启发性的架构思维。开发者常陷入“怎么让用户输入”的交互细节中，却忽略了核心是“系统需要什么数据”。先定义数据契约，再适配输入方式（交互式、参数、配置文件），能瞬间解耦业务逻辑与交互层，大幅提升工具的可组合性。
Develop applications with strong audio and visual understanding, for rich multimodal support.
令人意外的架构决策：音频输入能力是 E2B/E4B 专属的，反而是更大的 26B 和 31B 模型不支持音频。这意味着 Google 刻意把语音能力部署在边缘端——暗示他们对端侧语音助手场景的押注，而非将音频作为云端大模型的特权能力。小模型反而是音频 AI 的「第一公民」。
the LLM can effectively track functional emotional states of entities in its context window, including the Assistant, by attending to these representations across token positions, a capability of transformer architectures not shared by biological recurrent neural networks
Transformer 的注意力机制赋予了 LLM 一种人类大脑没有的能力：通过「回溯注意」缓存过去所有位置的情绪向量，实现跨时间的情绪追踪。这是 Transformer 架构与人类循环神经网络的根本差异——Claude 追踪情绪的方式，比人类大脑更像「翻阅历史记录」。
our DFC is architecturally designed with three distinct sections: A shared dictionary, A "French-only" section, An "English-only" section
Dedicated Feature Crosscoder（DFC）的三段式架构设计是这项研究的核心技术突破：通过分别建立「共享词典」和两个「专属词典」，强制让模型差异特征有独立的表示空间，而非被混入共享特征中。令人惊讶的是，如此影响深远的安全工具，其设计思路竟然与字典编纂学高度同构。
This unified design naturally extends beyond static images to video, voice agents, and fully interactive world simulators.
令人惊讶的是：UNI-1的统一设计能够自然地扩展到视频、语音代理和完全交互式世界模拟器，这表明该模型架构具有极强的可扩展性，可能成为未来多模态AI系统的基础框架。
SOTA models of different architectures and parameter scales exhibit highly consistent failure patterns on the same set of hard samples, suggesting that the performance bottleneck stems from shared deficiencies in training data rather than architecture itself.
大多数人认为不同架构的模型会有不同的失败模式和弱点，但作者发现无论架构和参数规模如何，SOTA模型在相同困难样本上表现出高度一致的失败模式，这表明性能瓶颈源于训练数据的共同缺陷，而非架构差异，这一发现挑战了模型多样化的传统观点。
they fuse streaming data construction with a unified model so the memory supports both real-time q&a and long-horizon interaction, which is nontrivial under strict latency constraints
大多数系统设计者可能认为实时问答和长时程交互需要不同的处理架构，但作者通过融合流式数据构建和统一模型，使内存同时支持这两种功能。这一设计挑战了实时系统处理复杂性的常规认知，表明在严格的延迟约束下实现多功能整合是可行的，这为实时AI助手的设计提供了新思路。
Rather than treating a complex document as a single monolithic task, Deep Extract deploys sub-agents to break it down and conquer each piece, which is what allows it to remain accurate even on documents with thousands of rows across hundreds of pages.
大多数人可能认为处理复杂文档的最佳方式是将其作为一个整体来处理，保持上下文完整性。但作者提出将复杂文档分解为多个子任务并由子代理分别处理的方法更有效，这一方法挑战了文档处理中'整体优于部分'的传统认知，暗示分解策略可能更适合处理超长文档。
we have kept the memory modules separate for each pipeline — precisely so that memory can be better isolated and iteratively improved during early development.
大多数人可能认为统一架构应该共享内存模块以提高效率，但作者选择为每个管道保持独立的内存模块，这挑战了系统设计的常规优化思路。这种分离方法虽然可能牺牲一些效率，但为早期开发提供了更大的灵活性和迭代空间。
Reconstructing raw inputs forces models to model irrelevant low-level detail. Predicting in a learned embedding space allows the model to focus on semantically meaningful, causally relevant features.
大多数人认为AI模型需要重建完整的输入数据才能理解世界，但作者认为这种方法迫使模型关注无关的低级细节。相反，在嵌入空间中进行预测可以让模型专注于语义上有意义、因果相关的特征，这是一个反直觉的见解。
these two challenges are fundamentally distinct: the former relies on fuzzy semantic planning, while the latter demands strict logical constraints
主流AI研究通常将语义规划和逻辑验证视为可以统一处理的问题，但作者明确指出它们是根本不同的挑战。这一观点与当前大多数LLM代理方法相悖，暗示了单一神经网络架构的局限性。
role-differentiated proposer/executor/checker/adversary systems may reduce correlated error under asymmetric information and verification burden.
大多数AI系统设计倾向于使用单一或少数几个组件来完成所有任务。但作者提出角色分化的提议者/执行者/检查者/对抗者系统可能减少在信息不对称和验证负担下的相关错误。这一观点挑战了当前AI系统设计的单一或少数组件架构，暗示多角色专业化可能更有效。
内置视频和音乐生成
大多数人认为AI系统需要专门的模块或插件来处理多媒体内容生成，但作者暗示OpenClaw已经将这些功能'内置'，表明其架构已经实现了高度整合，挑战了AI系统模块化设计的传统观念。
Sandboxes made for running tens of thousands of agents
大多数人认为在单个系统中运行数万个AI代理是不现实的，会导致资源竞争和性能下降。Freestyle明确将此作为设计目标，暗示他们的架构可能重新定义了AI代理的规模边界，挑战了关于AI系统可扩展性的主流认知。
https://web.archive.org/web/20260404074858/https://interconnected.org/home/2026/03/28/architecture
[[Matt Webb p]] on architecture in general and it architecture wrt AI
The unofficial API authenticates using your actual Sunsama email and password. So my serverless function needs to store my real credentials.
Does it? It's not obvious from the problem and proposed approach described here that the homespun service doing the LLM work (or, rather, coordinating it) and the client submitting stuff to the third-party service need to be the same.
The client really only needs to authenticate with the service and ferry data into the account. The data can be prepped elsewhere. The only real constraint is that it comes from a trusted source.
The tablet wasfound by archaeologists in the foundations of the temple of Inannain Lagash, called the Ibgal. This extensive complex was oval inshape, as were many Early Dynastic temples in other cities, with alarge courtyard and a platform on which Inanna’s temple wasconstructed.
What is the general history of oval-shaped architecture? Is there an explicit link between the Oval shape of the complex at Ibgal, the temple (or house) of Inanna in Lagash and the oval office at the White House?
Keep in mind that modern knowledge of large portions of the Ancient Near East only surfaced after the 1800s, so the tradition would have required intermediaries from the ANE into other cultures to be passed down to the building of the White House in 1792.
Some good pointers to [[Brian Eno c]] work and thinking, to follow up.
Also good anecdote from one of those links on Rem Koolhaas notion of n:: premature sheen Making things look nice early takes away from thinking about other points of quality. Jeremy applies it to AI too, the premature sheen generate awe, but not quality output.
for - Yann Lecun - paper - Yann Lecun - AI - LLMs are dead - language is optional for reasoning - to paper - VL-JEPA: Joint Embedding Predictive Architecture for Vision-language - https://hyp.is/eSxi8OxGEfCF7QMFiWL9Fg/arxiv.org/abs/2512.10942
Comment - That language and reasoning are separate is obvious. - If we look at the diversity of life and its ability to operationalize goal seeking behavior, that already tells you that - Michael Levin's research on goal-seeking behavior of organisms and the framework of multi-scale competency architecture validates Lecun's insight - Orders of magnitude fewer efficiency of Lecun's team's prototype compared to LLM also validates this
anthroposine's cognitive architecture
for - anthropocene - cognitive architecture
It's what my favorite archaeologist Bruce Trigger calls a thermodynamic explanation of symbolic behavior
for - paper - Monumental architecture: A thermodynamic explanation of symbolic behavior - Bruce Trigger
Frank Gehry, Titan of Architecture, Is Dead at 96 - The New York Times<br /> by [[Nicolai Ouroussoff]]<br /> accessed on 2025-12-06T07:42:42
for - Michael Levin - Multi Scale Competency Architecture - Hierarchical systems
Summary - Biomimicry - mimicking natural systems for new AI models - human body is a hierarchical system - society is extension of human body and also hierarchical system
living beings are organized in hierarchical structure from cells up until ecosystem.
for - multi-scale competency architecture - extend hierarchy of human body to society - continuation of levels - human being as cell in social superorganism - adjacency - multi-scale competency architecture - social superorganism
It's an Olivetti Valentine. An iconic piece of design with mechanicals almost identical to the Lettera 32. Its designer, Ettore Sotsass, was apparently not fond of it ("I worked as an architect for sixty years of my life and all people know is this fucking plastic machine" - or something to that effect) and its first production run was not particularly popular.
source for this?? interesting, if true
https://www.reddit.com/r/typewriters/comments/1mxxqkl/dieter_rams_typewriter_model/
Invisible architectures, on the other hand, consist of ways that our worldis structured for us and by us, except that the structure itself is not imme-diately visible
for - meme - invisible architecture - insight - language - as unintentional invisible architecture
Indeed, if each of us is a universe, home to species who are home to species,then who is us and who is
for - multi-scale competency architecture - language for
I envision a time when we are present to both our unitedness and ouruniqueness, similar to the way in which our own body’s many different typesof cells function togeth
for - multi-scale competency architecture - Michael Levin
Hill, Rosemary. Review of Use your theodolite, by Colin Richards and Vicki Cummings. London Review of Books, December 26, 2024. https://www.lrb.co.uk/the-paper/v46/n24/rosemary-hill/use-your-theodolite.
If architecture is defined as construction aesthetically conceived, then stone circles, which often have a clear entrance and worked surfaces to distinguish between the interior and the exterior, are unquestionably architecture.
what's more important from the perspective of a software architect is why a particular implementation or approach was chosen over its alternatives. A common way to document decisions like this is to use architecture decision records, ideally stored in source control with or near the application(s) impacted by the decision.
Why is more important than how
Everything is a trade-off
Software architecture choices include specific structural options from possibilities in the design of the software.
The software design model can be compared to an architected plan for a house. High-level plans represent the totality of the house (e.g., a three-dimensional rendering of the house). Lower-level plans provide guidance for constructing each detail (e.g., the plumbing lay). Similarly, the software design model provides a variety of views of the proposed software solution.
including both high-level software architecture and low-level component and algorithm design.
software design is the higher-level/broader/parent concept, which includes: - software architecture: high-level design - algorithm design and other lower-level design
The first one is about using the code artefacts (classes, variables, modules, …) names to convey both domain and architectural meaning.
“[…] an architecturally evident coding style that lets you drop hints to code readers so that they can correctly infer the design.”
"so that they can correctly infer the design"!
In this diagram, the dependency direction goes inwards, meaning that outer layers know about inner layers, but not the other way around.
what we want in our bodies is a mind melt of cells that cells forget about their own little ego and start identifying with a larger Collective of the organ and the way cells do that is through bioelectricity it is what Michael Levan calls a cognitive glue
for - multi-scale competency architecture - groups of individuals coalesce into an individual at a higher level - SOURCE - Youtube - Bioelectric fields: A Paradigm Shift in Biology - Michael Levin - 2025, Jan
Transdisciplinary sustainability science is increasingly applied to study transformative change. Yet, transdisciplinary research involves diverse actors who hold contrasting and sometimes conflicting perspectives and worldviews. Reflexivity is cited as a crucial capacity for navigating the resulting challenges
for - adjacency - reflexivity - tool for transdisciplinary research - indyweb - people-centered interpersonal information architecture - mindplex - concept spaces - perspectival knowing - life situatedness - SRG transdisciplinary complexity mapping tool - SOURCE - paper - Reflexivity as a transformative capacity for sustainability science: introducing a critical systems approach - Lazurko et al. - 2025, Jan 10
adjacency - between - reflexivity - tool for transdisciplinary research - indyweb - people-centered, interpersonal information architecture - mindplex - concept space - perspectival knowing - life situatedness - SRG transdisciplinary complexity mapping tool - adjacency relationship - This paper is interesting from the perspective of development of the Indyweb because there, - the people-centered, interpersonal information architecture intrinsically explicates perspectival knowing and life-situatedness - Indyweb can embed an affordance that is a meta function applied to an indyvidual's mindplex that - surfaces and aspectualizes the perspective and worldview salient to the research - The granular information that embeds an indyvidual's perspectives and worldviews is already there in the indyvidual's rich mindplex
also what it requires is to recast the individual as a non individual
for - adjacency - post Capitalist spiral - 5 Element Mandala - the individual as non-individual - Michael Levin's - Multi-Scale Competency Architecture - Post Capitalist Philanthropy - Alnoor Ladha - Lynn Murphy - 2023 - Stop Reset Go - Deep Humanity - Individual / Collective Gestalt
Information architecture is a task often shared by designers, developers, and content strategists. But regardless of who takes on the task, IA is a field of its own, with influences, tools, and resources that are worth investigation. In this article we’ll discuss what information architecture really is, and why it’s a valuable aspect of the user experience process.
Should we optimize for searching or browsing?¶ Documentarians may have to determine whether users search or browse for content of interest. What you decide may influence how to focus your resources: SEO and search tools or navigation aids. The resolution to this may depend on your users and what they’re looking for… and also your product interface. Some users, those who frequently search online for content, may prefer to search through your documentation (for example, spending 70% of their time on search and 30% navigation). Other users may prefer to use your site’s navigation system (for example, 30% search and 70% navigation). Nonetheless, some documentarians assume that searching is the primary method that all users rely on. Some indicate that it’s important to have both methods available for the users to select what they want to do. Information architecture (IA) helps a docs team to develop content in a structured and comprehensive manner. A navigation methodology can implement the IA of the documentation system. So, if your team has developed a structure for the content, you can use it as a navigation device for your readers. As one person indicated: No documentation should be random pages of text. Readers use the structure to learn relationships between different features, use cases, or topics. Searching and browsing are complementary actions. The method used by any one person may depend on different factors and users may use both. Offer the best of both to satisfy your readers. Search-related resources Search platform tips for documentation websites (WTD Newsleter) Making documentation discoverable in search engines (WTD video) Search engine optimization (SEO) for documentation (WTD page) Information Foraging (Nielsen Norman Group) Navigation- and IA-related resources Many articles available from Nielsen Norman Group Building navigation for your doc site: 5 best practices (WTD video) Complete Beginner’s Guide to Information Architecture (UX Booth) How To Make Sense of Any Mess (book by Abby Covert)
I don't reject the reductionist approach you need the upward causation from the individual molecular components as well as the constraints from the global property of the cells tissues organs and so on
for - Systems perspective - integrate reductionism with global properties - Denis Noble - multi scale competency architecture - Denis Noble - adjacency - multi-scale competency architecture - Michael Levin - Denis Noble advocates for
adjacency - between - multi-scale competency architecture - Michael Levin - Denis Noble advocates for - adjacency relationship - Michael's multi-scale competency architecture is another way to sum up what Denis is saying
we have all of these huge applications that are gathering all this data uh and it's out there and theoretically is our data sort of but in reality they control it and you can't actually link the data to each other you only link to accessing the data through their application
for - quote - silos - internet limitations - location addressed server architecture limitations - silos - cannot link data from each silo - Juan Benet - IPFS
For writing on a flat surface like a wrapped package (or book), you'd need a Gritzner or an Elliot Fisher Book Typewriter
a set of policies and mechanisms that allow competent subunits to form together into some kind of a an emergent Collective that's more than the sum of its parts
for - definition - cognitive glue - Michael Levin
definition - cognitive glue - Michael Levin - a set of policies and mechanisms that allow competent subunits to form together into some kind of an emergent Collective that's more than the sum of its parts
Adjacency - between - cognitive glue - multi scale competency architecture - human species - Jordan Hall - cognitive glue destresses goal seeking activities - adjacency relationship - Cognitive glue is a general concept that applies to the entire spectrum of the biosphere - Michael goes on to give examples with rats and other biological contexts like cells - This is an important question for humans at two levels: - first, at the level of the individual human - second, at the level of human groups - Jordan Hall brings the conversation to the cognitive glue at the human social level in which - anyone who has worked in a group context knows that when there is a flow, there is signaling taking place - that is at a higher group level not present at the level of the individual that destresses goal seeking activity
I think it's really important for us to develop a science of that like CR like critically important
for - answer - Micheal Levin - adjacency - hyperobject - cognitive light cone - critically important to develop a science of this
adjacency - between - multi scale competency architecture - cognitive light cone - hyperobject - awakening / enlightenment - adjacency relationship - At every stage of the multi scale competency architecture, - the living entities at a particular stage may maintain - feedback and - feedforward signals - with any - higher or - lower level systems. - Human INTERbeCOMings and other consciousness are no different - We exist at one level but are both - composed of lower level living parts and - compose larger social superorganism - Indeed, the spiritual acts variously described as - awakening - enlightenment - can be interpreted as transcending level cognitive light cone
the inside and the outside
for - adjacency - inside / outside - complexity / simplicity - multi scale competency architecture - black box - example - human consciousness
adjacency - between - inside / outside - black box - multi scale competency architecture - complexity / simplicity - adjacency relationship - inside / outside complexity /simplicity relationship articulates - the black box phenomenal prevalent in design and also - what Michael has been talking about with the complexity naturally found at lower levels of multi scale competency architectures - As he noted earlier, in this lab experiments, - it's practical to make use of the higher level signals in the living system - and virtually impossible to make use of trying to manage the lower level system signals - I like to think of human consciousness in the same terms - What appears to consciousness are signals like intero-ception signals of hunger that creates the thought ' I'm hungry, I want to get some food ' - whilst countless lies level signals that operate all the cells in our body are invisible
we do feel at least most of us, most of the time feel like some kind of unified, centralized inner perspective
for - self - as unified, centralized inner perspective - Michael Levin - adjacency - self - as unified, centralized inner perspective - multi-scale competency architecture - Buddhism - spiritual practice - self actualization - illusory body - illusory self - enlightenment
adjacency - between - self - as unified, centralized inner perspective - multi-scale competency architecture - self actualization - Buddhist practice - illusory body - illusory self - enlightenment - awakening - adjacency relationship - Indeed, from both the mundane and the spiritual, religious perspective, the unified self as a fundamental assumption - "self-development" and "self-actualization" are terms that are only meaningful if there is a unified self - Is the Buddhist ideas of - awakening - enlightenment and \ - penetrating the illusion of self - based on a kind of experiencing of the multi-scale competency architecture itself? - What does "spiritual awakening" mean in the context of multi-scale competency architecture? - For instance, WHO is it that actually awakens? - Is it consciousness from the SAME level, a lower level or ALL levels of the multi-scale competency architecture that a multi-cellular conscious, sentient being such as a human INTERbeCOMing? - If it includes consciousness from lower levels, then it may be billions or trillions of cellular consciousnesses that are awakening to the higher order consciousness it composes!
boundaries between cells, creatures and ecosystems are real but permeable. The bi-directional exchange of energy, information and matter across these boundaries is the communication that makes life possible.
for - adjacency - multi scale competency architecture - communication between levels - intrinsic to natural flows of life
In 1996, technology historian Jennifer S. Light compared the talk of “cyberoptimists” about virtual communities to city planners’ earlier optimistic predictions about shopping malls. As the automobile colonized U.S. cities in the 1950s, planners promised that malls would be enclosed public spaces to replace Main Streets. But as Light pointed out, the transition to suburban malls brought new inequities of access and limited the space’s functions to those that served commercial interests.
A comparison of urban development privatisation and internet development corporatisation
for - personal health - metabolic disease - insulin resistance caused by mitochondria dysfunction - interview - Dr. Robert Lustig - health - dangers of sugar in our diet
summary - Robert Lustig is a researcher and major proponent for educating the dangers of sugar as the root cause of the majority of preventable western disease - He explains how sugar and carbs are a major variable and root cause of a majority of these diseases - It is useful to look at these bodily dysfunctions from the perspective of Michael Levin, in which all these diseases of the body are problems with lower levels of the multi-scale competency architecture - https://jonudell.info/h/facet/?max=100&expanded=true&user=stopresetgo&exactTagSearch=true&any=michael+levin%2C+multi-scale+competency+architecture
You need something that will help you show documents that will help you map to those understandings, those contexts that people bring to the information very quickly. And documents are also the way the internet works, right? You want to be able to use the sorts of affordances that the engineers understand. Rest APIs, JSON documents, cache control. Because those are things that engineers know that will let this stuff work fast. It will also let you hire people who know how the internet works and don't have to understand all the complex crazy stuff that we call children, heritage people do, they make it possible.
If you can hire engineers that don't immediately need to know the intricacies of cultural heritage metadata, you can expand the pool of entry-level people you can hire and then train up on the cultural heritage details as needed.
But stepping back even further, one can only see this imagined software as an enhancement to Latour’s larger model of interplay in his actor-network theory, a theory that does not need software or special equipment to exist. The activity in a spatial environment is not reliant on the digital environment. It may be enhanced by a code/text-based software, but a spatial software or protocol can be any platform that establishes variables for space as information
Compared to the relative trickle of space made by special practitioners, these technologies produce a fire hose blast. The most radical changes to the globalizing world are being written in the protocols or softwares of infrastructural space.
On the one hand, Alexander expands the repertoire of design to include activity. But on the other, he quickly codifies and taxonomizes that activity. He mimics the object of this own critique by reforming the artificial with a “natural” corrective—instead of the tree, the semi-lattice becomes the placeholder. Despite his attempt to incorporate active form and information, Alexander only creates another immobilized form.
In his 1965 article “The City is not a Tree,” he critiqued what he deemed to be the infrastructural or organizational template of many settlements and cities. A “tree,” in Alexander’s parlance, is a branching structure in which sets are either completely disconnected from one another or entirely contained within one set without overlapping sets. The branches do not grow together but emanate separately from a single trunk.
We are not accustomed to the idea that non-human, inanimate objects possess agency and activity, just as we are not accustomed to the idea that they can carry information unless they are endowed with code/text-based information technologies. While accepting that a technology like mobile telephony has become the world’s largest shared platform for information exchange, we are perhaps less accustomed to the idea of space as a technology or medium of information—undeclared information that is not parsed as text or code. Indeed, the more ubiquitous code/text-based information devices become, the harder it is to see spatial technologies and networks that are independent of the digital. Few would look at a concrete highway system or an electrical grid and perceive agency in their static arrangement. Agency might only be ascribed to the moving cars or the electrical current. Spaces and urban arrangements are usually treated as collections of objects or volumes, not as actors. Yet the organization itself is active. It is doing something, and changes in the organization constitute information. Even so, the idea that information is carried in activity, or what we might call active form, must still struggle against many powerful habits of mind.
This capacity seems to finally fulfill the dream of artists and architects of the mid- to late twentieth century, among them Jack Burnham, Cedric Price, Archigram, and Christopher Alexander, who experimented with a cybernetic apparatus for modeling space.
“Cells may not know civilization is possible.
for - quote - multiscale competency architecture - quote - book - Emergent Strategy - Adrienne Maree Brown
adjacency - between - Adrienne Maree Brown quote - Michael Levin - adjacency statement - Adrienne's quote is the subsumed under Levin's term of multi-scale competency architecture (MSCA)
Whether ActivityPub or ATProto or webmention, the underlying technical protocol a community uses to interact online is a poor way to judge who they are and whether you might like them.
science has transformed our understanding of time.
for - language - primacy of - symbolosphere - adjacency - language - science - multi-scale competency architecture - Michael Levin - complexity - social superorganism - major evolutionary transition - worldviews - scientific vs religious - Michael Levin - multi-scale competency architecture
adjacency - between - deep time - multi-scale competency architecture - Michael Levin - social superorganism - complexity - major evolutionary transition - complexity - adjacency statement - Deep time narrative has potential for unifying polarised worldviews - but citing purely scientific evidence risks excluding and alienating large percentage of people who have a predominantly religious worldview - Language, the symbolosphere is the foundation that has made discourse in both religion and science possible - Due to its fundamental role, starting with language could be even more unifying than beginning with science, - as there are large cultural groups that - do not prioritize the scientific worldview and narrative, but - prefer a religious one.<br /> - Having said that, multi-scale competency architecture, - a concept introduced by Michael Levin - encapsulates the deep time approach in each human being, - which withing Deep Humanity praxis we call "human INTERbeCOMing" to represent our fundamental nature as a process, not a static entity - Each human INTERbeCOMing encapsulates deep time, and is - an embodiment of multiple stages of major evolutionary transitions in deep time - both an individual and multiple collectives - what we can in Deep Humanity praxis the individual / collective gestalt
for - adjacency - microscopic biology - macroscopic ecology - multi-scale competency architecture - Michael Levin - Jonas Wickman - micro-to-macro
paper details - title - Eco-evolutionary emergence of macroecological scaling in plankton communities - author - Jonas Wickman - Elena Litchman - date - feb 15, 2024 - publication - Science VOL. 383, NO. 6684
reference - https://www.science.org/doi/10.1126/science.adk6901
summary - This is a very interesting finding that links rules in the micro world to behavior in the macro. - It is relevant to Michael Levin's research on multi-scale competency architecture
question - how would this impact the micro relations between - the microscopic world of humans - the normal macroscopic world of humans
Able to see lots of cards at once.
ZK practice inspired by Ahrens, but had practice based on Umberto Eco's book before that.
Broad subjects for his Ph.D. studies: Ecology in architecture / environmentalism
3 parts: - zk main cards - bibliography / keywords - chronological section (history of ecology)
Four "drawers" and space for blank cards and supplies. Built on wheels to allow movement. Has a foldable cover.
He has analog practice because he worries about companies closing and taking notes with them.
Watched TheNoPoet's How I use my analog Zettelkasten.
Dubbed “litigation terrorism” by Joseph Stiglitz, the Nobel prize-winning economist. ISDS is a corporate tribunal system
for - litigation terrorism - ISDS - corporate tribunal system - Michael Levin - multi-scale competency architecture - example - adjacency - evolutionary biology - corporate law - climate crisis
adjacency - between - corporate law - climate crisis - evolutionary biology - cultural evolution - adjacency statement - Biologist Michael Levin's multi-scale competency architecture of evolutionary biology seems to apply here - in the field of corporate law - Corporations can be viewed as one level of a social superorganism in a cultural evolution process - Governments can be viewed similiarly, but at a higher level - The ISDS is being weaponized by the same corporations destroying the global environment to combat the enactment of government laws that pose a threat to their livelihood - Hence, the ISDS has been reconfigured to protect the destroyers of the environment so that they can avoid dealing with their unacceptable externalizations - The individual existing at the lower level of the multi-scale competency architecture(the corporation) is battling to survive against the wishes of the higher level individual (the government) in the same multi-scale competency architecture
me, Amin Taha + Groupwork and Pierre Bidaud of the Stone Masonry Company,
for - new stone age - stone age renaissance - stone architecture - practitioners - Amin Taha - Steve Webb - Pierre Bidaud
for - sustainable architecture - a new stone age - the return of stone - meme - a new stone age
story details - Title: Why the time is ripe for a return to stone as a structural material - Author: Steve Webb - Date: 2023, May 29 - source: https://www.ribaj.com/intelligence/stone-as-a-structural-material-embodied-carbon-sustainability
meme - new stone age
summary - Stone buildings have lasted millenia. Compared to steel, concrete and CLT, post-tensioned stone has the least embodied energy of all. - Could we also modernize ancient animal and human powered labor to create a low carbon stone building industry? -
for - multi scale competency architecture - Michael Levin - evolutionary biology - rapid whole system change - adjacency - multi scale competency architecture - rapid whole system change - stop reset go - Deep Humanity - Indyweb - Indranet - major evolutionary transition in individuality - MET - superorganism - cumulative cultural evolution of individuality
adjacency - between - multi scale competency architecture - rapid whole system change - progress trap - stop reset go - Deep Humanity - Indyweb - Indranet - major evolutionary transition in individuality - MET - superorganism - cumulative cultural evolution of individuality - adjacency statement - The idea of multi scale competency architecture can be extended to apply to the cultural level. - in the context of humanity's current existential poly /meta/ perma crisis, - rapid whole system change - (a cultural behavioural paradigm shift) - is required within a few short years - to avoid the worst impacts of - catastrophic, - anthropogenic - climate change, which is entangled with a host of other earth system boundary violations including - biodiversity loss - fresh water scarcity - - the driver of evolution through major evolutionary transitions in individuality has given rise to the level of cultural superorganisms that include all previous levels - progress and its intended consequences of progress traps play a major role in determining the future evolutionary trajectory of our and many other species - our species is faced with a few choice permutations in this regard: - individually regulate behaviour aligned with a future within earth system boundaries - collectively regulate behaviour aligned with a future within earth system boundaries - pursue sluggish green growth / carbon transition that is effectively tinkering at the margins of rapid whole system change - BAU - currently, there doesn't appear to be any feasible permutation of any of the above choices - There is insufficient worldview alignment to create the unity at scale for report whole system change - individual incumbent state and corporate actors still cling too tightly to the old, destructive regime, - creating friction that keeps the actual rate of change below the required - Stop Reset Go, couched within the Deep Humanity praxis and operationalized through the Indyweb / Indranet individual / collective open learning system provides a multi-dimensional tool for a deep educational paradigm shift that can accelerate both individual and collective upregulation of system change
from - IAAC Lecture Series - Architecture made of Earth with Elisabetta Carnevale - https://hyp.is/A279UruNEe6A7E_2XRTNRQ/docdrop.org/video/IxrmA9dli1Q/
from - IAAC Lecture Series - Architecture made of Earth with Elisabetta Carnevale - https://hyp.is/A279UruNEe6A7E_2XRTNRQ/docdrop.org/video/IxrmA9dli1Q/
from - IAAC Lecture Series - Architecture made of Earth with Elisabetta Carnevale - https://hyp.is/A279UruNEe6A7E_2XRTNRQ/docdrop.org/video/IxrmA9dli1Q/
from - IAAC Lecture Series - Architecture made of Earth with Elisabetta Carnevale - https://hyp.is/A279UruNEe6A7E_2XRTNRQ/docdrop.org/video/IxrmA9dli1Q/
from - IAAC Lecture Series - Architecture made of Earth with Elisabetta Carnevale - https://hyp.is/A279UruNEe6A7E_2XRTNRQ/docdrop.org/video/IxrmA9dli1Q/
from - IAAC Lecture Series - Architecture made of Earth with Elisabetta Carnevale - https://hyp.is/A279UruNEe6A7E_2XRTNRQ/docdrop.org/video/IxrmA9dli1Q/
grenoble architecture school
for - architectural school - earthen building - Grenoble Architecture School - https://www.grenoble.archi.fr/#
adobe which is the 00:32:14 probably the most forgotten uh earth and technique uh in contemporary architecture
for - adobe - forgotten earthen technique in contemporary architecture
for - Badgir - passive cooling - passive air conditioning - passive HVAC - sustainable architecture
Terascale’s SIMD engines
OLD AMD SIMD architecture
So we're good with three-dimensional space, but imagine if we had a primary sense of our own blood chemistry.
for - adjacency between - primary sense of cellular metabolism - experiences of deep contemplative practice of Rainbow Body - adjacency statement - As per Father Francis Tiso's research into the Tibetan deep contemplative Dzogchen phenomena of Rainbow Body at the time of death as well as rigpa, Trekcho and Togal, he speculates that - such deep contemplations can potentially result in a primary sense of cellular and even subatomic processes taking place within the human body. - https://hyp.is/go?url=http%3A%2F%2Fdocdrop.org%2Fvideo%2FsDyu39FCAMk%2F&group=world - It seems that the multi-scale competency architecture would be a good scientific framework to explore these questions.
So we're good with three-dimensional space, - but imagine if we had a primary sense of our own blood chemistry. - If you could feel your blood chemistry - the way that you currently see and smell and taste things that are around you, - I think we would have absolutely no problem having an intuitive understanding - of physiological-state space - the way we do for three-dimensional space.
claim - Lifetime practitioners of Tibetan meditation claim they have a primary sensation of their own impending death suggestion - Suggest to Michael Levin to investigate such phenomena from a multi-scale competency architecture perspective - What else can the expert meditators directly experience? And how do they achieve this? How can deep contemplative practice result in such profound experiences? Would expert meditators resonate with Levin's framework?
Venkatesh Rao thinks that the Nazi bar analogy is “an example of a bad metaphor contagion effect” and points to a 2010 post of his about warren vs plaza architectures. He believes that Twitter, for example, is a plaza, whereas Substack is a warren: A warren is a social environment where no participant can see beyond their little corner of a larger maze. Warrens emerge through people personalizing and customizing their individual environments with some degree of emergent collaboration. A plaza is an environment where you can easily get to a global/big picture view of the whole thing. Plazas are created by central planners who believe they know what’s best for everyone.
楣樑(柱頂楣構) Note: 主要用於古典建築物，是古典柱頭頂部三大主要部份中位置最低的一個，擱在圓柱頂板上，包含以過樑方式橫跨圓柱或角柱的水平橫樑。古典柱頭頂部的另外兩個部份為挑檐和檐壁。環繞門或窗，類似柱頂過樑的鑄模框架，稱為「楣樑飾邊」。環繞拱門外部弧線的裝飾板條，則稱「拱門緣飾」。
example of meta/hypothesis-on-glasp annotation
entablature 楣樑；柱頂楣構
order: 古典建築柱式(style, rule) e.g. Doric order
楣樑(柱頂楣構) Note: 主要用於古典建築物，是古典柱頭頂部三大主要部份中位置最低的一個，擱在圓柱頂板上，包含以過樑方式橫跨圓柱或角柱的水平橫樑。古典柱頭頂部的另外兩個部份為挑檐和檐壁。環繞門或窗，類似柱頂過樑的鑄模框架，稱為「楣樑飾邊」。環繞拱門外部弧線的裝飾板條，則稱「拱門緣飾」。
entablature 楣樑；柱頂楣構
order: 古典建築柱式(style, rule) e.g. Doric order
for: social tipping point, multi-scale competency architecture, MET, major evolutionary transition of individuality
Title: Using emergence to take social innovation to scale
Emergence is how lifecreates radical change and takesthings to scale
Defections from large-scale anatomical goals, such as those that occur due to an inappropriate reduction of gap junctional connectivity [74], present as cancer, cause reversions of cell behavior to ancient unicellular concerns which lead to metastasis and over-proliferation as the cells treat the rest of the body as external environment.
fresh perspective
adjacency between
multiscale competency architecture of life
paraphrase
comment
adjacency statement
question
summary
biology uses a kind of multi-scale Competency architecture of nested problem solvers 00:03:24 and that navigation is a really Central concept
Structure of Automation
Tuile
┌─────────────┐ ┌──────────┐
│Lansdcape │ 1:N │Type │ 1:N
│(déploiement)├──────►│(playbook)├───┐
└─────────────┘ └──────────┘ │
│
┌──────────────────────────────────┘
│
│ ┌────────┐ ┌───────────────────┐
│ │Function│ 1:N │Component │
└►│(rôle) ├──────►│(task file du rôle)│
└────────┘ └───────────────────┘
collections contain various modules and Ansible Roles that collectively (intentional use here) serve the end user and must (emphasis on this word) collectively work together, be tested together and be released together.
...collectively work together, be tested together and be released together.
Communication is important so that the different divisions in the company know which services are available.
Lien à faire avec le modèle de capacité d'affaires.
Five core layers: 1. Presentation Layer 2. Application Layer 3. Domain Layer 4. Persistence Layer 5. Database Layer
In the layered architecture pattern, each layer fulfills a specific responsibility and role within the application. Some focus on user interface logic, while others handle the execution of business rules. These layers complement each other's unique purposes, but they aren't explicitly reliant on one another to perform their own tasks.
In software architecture, layers act as individual processes within the infrastructure of an application. These layers typically form a pattern, also called the n-tier architecture pattern.
Payment Information Centrality Revision Project "Ways to fund your education" (or similar) instead of "Minimize Financial Barriers". "Verified Sources" as a phrase is problematic for users. Desire is to provide assurance to users that there are individualized options yet still have those options located in a more streamlined, centrally located way.
These are the other pages that have similar content: https://www.waubonsee.edu/admissions/costs-and-payments/paying-tuition
https://www.waubonsee.edu/admissions/costs-and-payments/paying-tuition/payments-verified-sources
https://www.waubonsee.edu/admissions/costs-and-payments/paying-tuition/making-payments
https://www.waubonsee.edu/admissions/costs-and-payments/paying-tuition/payment-plans
https://www.waubonsee.edu/admissions/costs-and-payments/paying-tuition/tuition-refunds
https://www.waubonsee.edu/admissions/costs-and-payments/paying-tuition/tax-credits-education
The second purpose of skip connections is specific to transformers — preserving the original input sequence.
Skip connections serve two purposes. The first is that they help keep the gradient smooth, which is a big help for backpropagation. Attention is a filter, which means that when it’s working correctly it will block most of what tries to pass through it.
The central object in the transformer is the residual stream.
One of the main features of the high level architecture of a transformer is that each layer adds its results into what we call the “residual stream.”Constructing models with a residual stream traces back to early work by the Schmidhuber group, such as highway networks and LSTMs, which have found significant modern success in the more recent residual network architecture . In transformers, the residual stream vectors are often called the “embedding.” We prefer the residual stream terminology, both because it emphasizes the residual nature (which we believe to be important) and also because we believe the residual stream often dedicates subspaces to tokens other than the present token, breaking the intuitions the embedding terminology suggests. The residual stream is simply the sum of the output of all the previous layers and the original embedding. We generally think of the residual stream as a communication channel, since it doesn't do any processing itself and all layers communicate through it.
A transformer starts with a token embedding, followed by a series of “residual blocks”, and finally a token unembedding. Each residual block consists of an attention layer, followed by an MLP layer. Both the attention and MLP layers each “read” their input from the residual stream (by performing a linear projection), and then “write” their result to the residual stream by adding a linear projection back in. Each attention layer consists of multiple heads, which operate in parallel.
Defenders can design their own infrastructure to be im-mutable and ephemeral, as is becoming an emerging trend inprivate sector defense through the practice of Security ChaosEngineering
Immutable: unchangeable infrastructure components, such as ssh access disabled by default.
Ephemeral: short-lived servers for single processes, serverless infrastrucure
Deploy engines as separate app instances and have them only communicate over network boundaries. This is something we’re starting to do more.
Before moving to this microservice approach, it's important to consider whether the benefits are worth the extra overhead. Jumping to microservices prematurely is something I've seen happen more than once in my career, and it often leads to a lot of rework.
we need to consider how the architecture might make the UX suck.
Architecture can make UX suck.
my strong sense is that we’re currently papering over major UX problems that are linked to core architectural properties.
