Cookie Cutter Physical AI

This week: predictive maintenance ROI (paid), train a whole-body control foundation model, cardboard boxpocalypse, dry electrode printing, generative model for structured CAD, and open WMA models.

Sep 21, 2025

∙ Paid

Shop Talk

Capturing this week's zeitgeist

Ann Clark's success demonstrates how foundational manufacturing technologies and systematic process improvement can create competitive advantages without requiring AI. The company leveraged lean manufacturing principles to reduce inventory by 55% and optimized their production equipment to match the speed and cost-effectiveness of overseas alternatives. Their flexible manufacturing setup allows rapid changeovers between 700 different cookie cutter shapes, enabling same-day fulfillment that imported competitors cannot match. This operational agility, combined with quality control from in-house production, has allowed a small Vermont manufacturer to capture 65-70% of the U.S. market while maintaining premium pricing - proving that strategic process optimization and domestic supply chain control can outweigh pure cost advantages.

Ann Clark's story raises a few questions when evaluating technology:

Can flexibility beat AI forecasting?
Is supply chain visibility overrated for vertically integrated operations? Even Musk is making electrical transformers now.
How do you balance automation with workforce flexibility?

Next up: Pack Expo, September 29-30, 2025 in Las Vegas, NV, USA

Assembly Line

This week's Industry 5.0 breakthroughs and frontier technologies of the built world.

🧠🤖 Agility Robotics explains how to train a whole-body control foundation model

Agility Robotics has developed a whole-body control foundation model for its Digit humanoid robot to ensure it remains safe and stable while performing a wide variety of tasks. It functions much like the motor cortex in the human brain, taking signals from different parts of the brain — different levels of Agility’s control hierarchy in Digit’s case — to control voluntary movements and fine motor skills.

This model is a relatively small LSTM (long short-term memory) neural network with fewer than 1 million parameters, which is trained in NVIDIA’s Isaac Sim physics simulator for decades of simulated time over three or four days.

Digit’s motor cortex is learned purely in simulation and transfers zero-shot to the real world. It can be prompted with dense free-space position and orientation objectives for the arms and torso to accomplish different goals, including walking around and picking and placing heavy objects. It can learn downstream dexterous manipulation skills on top of the motor cortex, and similarly coordinate complex behaviors using large language models (LLMs).

📦♻️ The Cardboard Boxpocalypse

Oregon's Plastic Pollution and Recycling Modernization Act /Oregon.gov/ and the National Association of Wholesaler-Distributors filed a lawsuit on July 30 challenging the constitutionality of 2021’s Plastic Pollution and Recycling Modernization Act. /Packaging Dive/

🖨️🔋 Is 3D Printing Battery Components a Reality? Inside Sakuu’s Dry Electrode Printing Technology

✍️ Author: Paloma Duran

Sakuu specializes in printing dry materials, and its Kavian platform has readapted 3D printing techniques for high-volume battery production. The company developed a process in which the battery’s current collector foil moves beneath the deposition head, rather than moving the head itself.“This allows us to increase the speed of material deposition,” Niestroj explained. “More importantly, it’s about applying the right process to each layer, controlling thickness, speed, and quality. Aligning the process, the material, and the resulting layer is the core capability that enables scaling.”

While electrode processing remains highly demanding, scalable, real-world solutions are rare. Sakuu is currently collaborating with SK On in pre-commercial testing to assess requirements for producing final electrodes and mass-production equipment. These pilot studies are intended to validate the technology, identify limitations, and inform potential adoption.

🏭📊 The Industrial Data Refinery. 101 Tech Talk.

SparkPipe: A Streamlined, Cloud-Ready Pipeline for OT Data Built on MQTT Sparkplug /N3uron/
From blue ocean to red lake: What happened to the 620+ IoT platforms, and what is ahead /IoT Analytics/

The IT/OT Insider

Industrial AI Unpacked: Introducing a new series of Articles and Podcasts on AI

“We don’t want to make the same bad decision from today faster — we want better decisions…

5 days ago · 4 likes · David Ariens and Willem van Lammeren

New Product Introduction

Highlighting new and innovative facilities, processes, products, and services

🧠📐 Introducing SGS-1: the first generative model for structured CAD

Announcing SGS-1, a foundation model that can generate fully manufacturable and parametric 3D geometry. You can try a research preview of SGS-1 here. Given an image or a 3D mesh, SGS-1 can generate CAD B-Rep parts in STEP format. Unlike all other existing generative models, SGS-1 outputs are accurate and can be edited easily in traditional CAD software.

🧠🤖 First Open-Source Operating System for Intelligent Robots

At OpenMind, we believe that needs to change. We’re excited to announce the beta release of the world’s first open-source operating system for intelligent robots - a universal platform that allows any robot to perceive, reason, and act in the real world.

OM1 is designed to be hardware-agnostic and open from the ground up. It lets developers install fully autonomous software on robots without being locked into a single ecosystem or forced to reinvent the wheel.

🧠🤖 UnifoLM-WMA-0: A World-Model-Action (WMA) Framework under UnifoLM Family

UnifoLM-WMA-0 is Unitree‘s first open-source world-model–action architecture spanning multiple types of robotic embodiments, designed specifically for general-purpose robot learning. Its core component is a world-model capable of understanding the physical interactions between robots and the environments. This world-model provides two key functions: (a) Simulation Engine – operates as an interactive simulator to generate synthetic data for robot learning; (b) Policy Enhancement – connects with an action head and, by predicting future interaction processes with the world-model, further optimizes decision-making performance.

Business Transactions

This week's top funding events, acquisitions, and partnerships across industrial value chains.

🇺🇸🦾🤖🦿 Figure Exceeds $1B in Series C Funding at $39B Post-Money Valuation

Figure, the AI robotics company developing autonomous general-purpose humanoid robots, announced it has exceeded more than $1 billion in committed capital through its Series C financing round, at a post-money valuation of $39 billion. The funding will accelerate the company's efforts to bring general-purpose humanoid robots into real-world environments at scale. The round was led by Parkway Venture Capital with significant investment from Brookfield Asset Management, NVIDIA, Macquarie Capital, Intel Capital, Align Ventures, Tamarack Global, LG Technology Ventures, Salesforce, T-Mobile Ventures, and Qualcomm Ventures.

Additionally, Figure announced two pivotal advances for Helix, Figure’s Vision-Language-Action (VLA) model for generalist humanoid control:

Project Go-Big: Internet-Scale Humanoid Pretraining. Figure is building the world’s largest and most diverse humanoid pretraining dataset, accelerated by an unprecedented partnership with Brookfield, which owns over 100,000 residential units worldwide.
Zero-shot human video-to-robot transfer. Helix has achieved a new learning milestone: after training exclusively on egocentric human video, Figure robots can now navigate cluttered real-world spaces from natural language commands like "go to the fridge"—a first in humanoid robotics.

🇺🇸 Divergent Announces $290 Million Series E to Scale Digital Manufacturing Platform and Meet Growing US Defense Production Demand

Divergent Technologies, Inc. announced the closing of its Series E financing raising a total of $290 million at a $2.3 billion valuation. The round was led by Rochefort Asset Management and consists of $250 million in equity capital and $40 million in debt capital.

Founded in 2014, Divergent is the creator of the Divergent Adaptive Production System (DAPS™), the world's first end-to-end digital manufacturing platform enabling rapid design, additive manufacturing, and automated assembly. The new capital will scale manufacturing capacity for the business and fund the development of new capabilities for upcoming product families.

🇺🇸 Lila Sciences Series A: Welcoming New Partners in Our Mission to Build Scientific Superintelligence

Lila Sciences has announced the completion of a $235 million Series A funding round, co-led by Braidwell and Collective Global. The round also included participation from Altitude Life Science Ventures, Alumni Ventures, ARK Venture Fund, Common Metal, Flagship Pioneering, General Catalyst, March Capital, the Mathers Foundation, Modi Ventures, NGS Super, the State of Michigan Retirement System, and a wholly owned subsidiary of the Abu Dhabi Investment Authority (ADIA).

Since emerging from stealth in March, they’ve accelerated progress toward an operating system for autonomous science across materials, chemistry, and life science. With this growth capital, Lila will take its platform to the next level of intelligence, scale, and speed—and move closer to enabling breakthroughs that might otherwise have taken decades.

🇩🇪 encentive taps €6.3M, led by General Catalyst, to slash industrial energy costs with AI

As energy prices soar and climate demands intensify, industrial companies face mounting pressure to cut electricity expenses and reduce emissions, or risk losing their competitive edge. encentive, a German AI software startup, is filling this gap with its platform flexOn, which directly controls industrial assets to reduce energy costs by up to 20%. The company announced it raised €6.3 million in a seed funding round led by US venture firm General Catalyst, known for its forward bets on startups like Parloa and Mistral. The new capital will enable encentive to accelerate international expansion and deepen its AI-powered energy management platform, targeting what experts estimate is a billion-euro market opportunity.