AI hit the plant from both sides. The MES is still the gate.
Deep dive · Rocky Wang · October 28, 2026 · Combines the Oct 20–27 MES-inside wave (Parsec, Critical Manufacturing, L2L,…
Read More
Who runs the plant’s agents? Rockwell, Dassault and Oracle each want the orchestration layer
Deep dive · Rocky Wang · October 28, 2026 · Covers Rockwell FactoryTalk Conductor (Oct 21), Dassault Systèmes’ long-running Virtual…
Read More
别再堆子智能体了:Anthropic 建议收成一条主循环
Anthropic 近日建议:企业别急着给每个团队、每类问题各做一套子智能体,再挂编排器。多个客户部署对比里,带技能(skills)的单智能体主循环,往往质量更好,成本与延迟也更低。有客户对比后,单循环大约快一倍。 常见翻车是「康威定律」:代理架构长成组织图——意图分类、固定流水线、分客群分支,上面再套编排器。结果是交接丢上下文、子代理蔓延、系统提示互相打架、评测各评各的、多一层调用就多一层等待。一家客户面向代理塞了十几个专家子代理,品牌语气难统一,延迟也伤体验。 图:从多子智能体乱线到单主循环。 更好的目标形态是:一个模型、一个上下文窗口、开放式工具循环。硬约束(校验、审批)放进工具,不靠提示词口头规矩;领域步骤收成 skills 按需加载;工具延迟加载。工具定义超过约 30–50 个时,选型往往变差。Claude Code 对更新模型的系统提示曾缩八成以上,部分靠 skills/tools 渐进披露。 子代理仍有用:单循环上下文已可测得伤表现、需要并行且用完即弃的窗口、或选工具/指令冲突明显时。其余多数可先改成技能 + 工具。Anthropic 已放 迁移工具包,建议先做端到端评测,一次只折一个——有支持场景折进去后延迟降了,却让主代理跳过了该问专家的步骤。 我的看法: 在…
Read More
Humanoid Robots Are on the Line. Now the Plant Has to Run Them.
Key takeaways Tesla started Optimus production at Fremont — with no unit counts. The Q3 update (Oct 21, 2026) lists…
Read More
Humanoid Robots Are on the Line. Now the Plant Has to Run Them. (Deep Dive)
Deep dive · Rocky Wang · October 25, 2026 (PT) · Tesla Optimus at Fremont and humanoids in real factory…
Read More
The Critical 5%: A&D Supplier Capacity Analysis and Tracking
Deep dive · Rocky Wang · October 23, 2026 (America/New_York) · Aerospace & Defense supplier capacity, material demand, KPIs, and…
Read More
AI Agents Have Reached the Planning Desk. The Plant Still Decides.
Key takeaways Agents now sit upstream of MES. Siemens put one on the production plan in Erlangen; Arda put one…
Read More
Grok Bot as Your Apriso Engineering Agent: From Requirement to Importable GPM in a Day
How an autonomous AI agent, connected to the IACF Apriso MCP server, carried a real shop-floor use case through design,…
Read More
AI agents move into the MES, and onto the plant’s own servers
Deep dive · Rocky Wang · October 23, 2026 · Covers Parsec TrakSYS 14.1, Critical Manufacturing v12, L2L Shop Floor…
Read More
Agents reach the planning desk: what Siemens Erlangen and Arda show about AI in production planning and manufacturing engineering
Techbot deep dive for Rocky Wang · October 22, 2026 · Sources are numbered in brackets and listed with dates…
Read More
Your AI Agent Is an Insider Now. Put Two Gates in Front of It.
If 30 years in manufacturing IT have taught me one thing about access, it’s this: whatever you grant will get…
Read More
Bring Your Own LLM, or None at All: Choosing Where MoM-MCP’s AI Runs
MoM-MCP is IACF’s bridge between AI and large language models (LLMs) and the manufacturing applications a plant already runs: MES/MoM,…
Read More
Siemens and NVIDIA move PCB inspection setup into the digital twin
Deep-research brief for Rocky Wang · October 21, 2026 · Sources checked Oct 21, 2026 (America/Chicago) 1. Summary Siemens and…
Read More
AI Agents on the Shop Floor: Where They Help Today, and Where They Shouldn’t (Yet)
After 30 years in manufacturing IT, I ask one question about any new technology that wants to get near production:…
Read More
Sign Up, Hand It Off: Your AI Agent Is Ready for Work
At the start of 2026, a personal AI agent meant self-hosting one. Now you download an app, sign in and…
Read More
Siemens Intelligence Center X: what it is, how it’s built, and whether a plant should care
Research brief for Rocky Wang. Status as of October 17, 2026. Every factual claim links to its source. Numbers that…
Read More
Bigger Agents, Tighter Fences: The AI Week of October 10–17, 2026
Most weeks, AI news is a pile of benchmark charts. This one had a shape. Two frontier labs and a…
Read More
Inside the MoM-MCP Flow Engine: From Any Trigger to a Governed, Approved Action
MoM-MCP is IACF’s bridge between AI and large language models (LLMs) and the manufacturing applications a plant already runs: MES/MoM,…
Read More
From Claws to Grok Bot: A Practical Guide to the 2026 Agent Wave
Most AI hype cycles follow the same arc: a demo, a flood of posts, then the quiet realization that the…
Read More
Introducing MoM-MCP: The Governed Bridge Between AI and Every Manufacturing Application
Manufacturers want AI that does more than chat. They want it to read live production data, explain what is happening,…
Read More
