How knowledge is called and tasks get done知识调用与任务执行
How knowledge is called on, how tasks are carried out, and how an individual's capability is extended.
知识怎样被调用，任务怎样被执行，个人能力怎样扩展。
The levers worth studying: software, compute, automation and collaboration.
可研究的杠杆：软件、算力、自动化与协作。
Questions this concept asks这个概念追问的问题
- Which step of the task does this capability actually change, and which step still needs a person?这项能力实际改变了任务中的哪一步？哪一步仍要人来做？
- Does it do things for the individual, or let the individual do what they could not do before?能力扩展是替个人做事，还是让个人能做以前做不到的事？
- How do cost, reliability and the review burden change?成本、可靠性与审核负担怎样变化？
Common readings we test against evidence我们会用证据检验的常见看法
- A higher benchmark score means faster work.基准分更高 = 工作更快。
- Faster output means the task gets done faster.更快的输出 = 更快完成任务。
- An agent can replace the whole workflow.agent 能替代整个流程。
Evidence that settles it需要的证据
- Trigger:触发事件： First-hand release notes for the model, tool or feature (version, conditions, price).模型、工具或功能发布的一手说明（版本、条件、价格）。
- Mechanism:机制： A breakdown of the task flow, stated limits, price and latency data.任务流程拆解、限制说明、价格与延迟数据。
- Counter-evidence:反方证据： Independent evaluations or failure cases.独立评测或失败案例。
- History:历史对照： The real adoption curve of the previous generation of tools.上一代工具的真实采用曲线。
WAKENEX pieces built on this concept运用这个概念的 WAKENEX 内容
- More code, not more software: review is the binding step.更多代码，不等于更多软件：瓶颈在人工评审这一步。 ·
- Agents now compete on cheap human steering, not on generation.编码智能体如今比拼的是“让人低成本地掌舵”，因为限制产出的是人的监督注意力，而不是生成能力。 ·
Source: WAKENEX core thinking V1.0 · II. Three windows · AI来源：WAKENEX 核心思维 V1.0 · 二、三个观察入口 · AI技术 · All concepts全部概念 · Markdown