5,127 Failed Prototypes and a Cartridge in a Gear Tree

This week: Dyson's failed prototypes, 2D Images into G-code for Additive, street sweeper manufacturing dark factories, Waymo foundation model, grocery automation, natgas turbines, agentic BOM engine

Dec 14, 2025

∙ Paid

Shop Talk

Capturing this week’s zeitgeist

There is no better way to spend the holiday season than upskilling from the comfort of your own home. Whether you are physically building a DIY Arctos arm or looking to master the theory behind industrial automation, the tools are ready for you. The Apple Manufacturing Academy is now available virtually with admission through an online application.

Get 50% off for 1 year

Assembly Line

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

🗣️ Conversation With James Dyson, Founder of Dyson

After developing 5,127 failed prototypes and being rejected by every major manufacturer, Dyson launched his own company and reshaped the vacuum industry by the 1990s. He became known for his iterative engineering approach, his cyclonic separation technology that eliminated bags, and his ability to bring products to market against fierce opposition. Around half of Dyson’s global team are engineers and scientists, with research interests spanning robotics, AI, machine learning, solid-state battery development, material science, and high-speed electric motors.

⚙️🧠 AI Framework Translates 2D Images into G-code for AM

Researchers from Carnegie Mellon University have developed Image2Gcode, a novel deep-learning framework that generates printer-ready G-code directly from 2D images, effectively bypassing the traditional, multi-step workflow of CAD modeling, meshing, and slicing.

Using a diffusion-transformer model trained on over 100,000 STL-G-code pairs, the system interprets visual inputs—ranging from digital slices to hand-drawn sketches and photographs—to synthesize continuous extrusion paths and feasible infill structures without rule-based programming. This end-to-end approach not only lowers the barrier to entry for non-experts by creating a direct link between visual concepts and physical fabrication but also demonstrates a 2.4% improvement in path efficiency compared to standard slicers, paving the way for autonomous, intent-aware manufacturing systems.

🏭 Inside Schwarze Industries: How Street Sweepers Are Manufactured

Throughout the manufacturing process, Schwarze has shown a strong commitment to cost optimization and operational efficiency. A major source of savings has come from re-evaluating their make-versus-buy decisions across critical components.

🏭 Inside China’s ‘dark factories’ where robots run the production lines

China is expanding highly automated factories where robots outnumber workers and operate with minimal lighting. One car plant can build up to 1,200 electric vehicles a day, cutting labour costs and reducing errors. Humans are still needed for maintenance and final quality checks, but staffing is seven times lower than in traditional plants.

🏭 Ford’s Car of the Future, Hatched in a Skunk Works Near Los Angeles. /NYT/ Desperate to catch up with Chinese automakers, Ford is redesigning its fleet with a Silicon Valley-style team. Is it too late?
🏭 Hadrian to Build AI-Powered Production Cell at Lockheed Missile Factory /Design and Development Today/

🚘🧠 Demonstrably Safe AI For Autonomous Driving

The Waymo Foundation Model is a versatile, state-of-the-art world model powering our AI ecosystem. Its innovative architecture provides significant benefits over the pure end-to-end or modular approaches.

In particular, the model leverages the full expressibility of learned embeddings as a rich interface between model components and supports full end-to-end signal backpropagation during training. At the same time, its additional compact, materialized structured representations like objects, semantic attributes, and roadgraph elements allow for:

Powerful correctness and safety validation at inference time in the Driver
Highly efficient, physically-correct and realistic closed-loop Simulation at extremely large scale
Strong verifiable feedback signals for evaluation by the Critic and reinforcement learning during training

New Product Introduction

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

⚡ AI Needs More Power Than the Grid Can Deliver. Supersonic Tech Can Fix That

The future of flight just unlocked the future of power. Superpower is Boom’s 42-megawatt natural gas turbine built to power AI datacenters. The same supersonic technology drives both the Superpower turbine and Boom’s Symphony jet engine: an all-new engine core designed for sustained and efficient high power output, even in challenging thermal conditions.

Read more on Boom and CEO Blake Scholl on X

💰 Boom Supersonic to Power AI Data Centers With Superpower Natural Gas Turbines; Adds $300 Million in New Funding /Business Wire/
McMaster vision of turbine part numbers? /X/

💿⛓️🧠 Wizerr AI Launches the Agentic BOM Engine for Electronics Manufacturing, Built with NVIDIA

Wizerr AI introduced the Agentic BOM Engine, powered by Wizerr’s patent-pending component intelligence engine (ELX). ELX reads and interprets complex component datasheets with engineer-level precision and scales that capability to millions of components, turning unstructured documents into AI-transformed, reasoning-ready data for multi-agent workflows across the electronics manufacturing value chain.

The Agentic BOM Engine delivers engineer-grade analysis and procurement-grade intelligence to every BOM decision. Teams get validated alternates, supplier-merge opportunities, and clear visibility into cost and sourcing risks in minutes, helping engineering, sourcing, and reliability teams make faster, better-aligned decisions from design through production.

Unitree Debuts the World’s First Humanoid Robot “App Store”

Chinese home appliance giant Midea unveils 6-arm robot for factory work /SCMP/
Mercado Libre and Agility Robotics Announce Commercial Agreement to Deploy Humanoid Robots /Business Wire/

Oracle Runs Into More Trouble as Bonds Looks Like Junk /Bloomberg/ while Marvell Eyeing Connectivity as the Next Big Thing in AI /EETimes/ positioned to deliver next-gen optical interconnects /Tom’s Hardware/

Business Transactions

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

📑 Future-Ready Manufacturing Study 2025 /TCS/ with 88% of leaders anticipating AI will capture at least 5% of operating margin; 1 in 4 expect returns above 10%
🇸🇪 Industrifonden Exits Crosser Following Acquisition by Aveva / Schneider Electric /Industrifonden/
🇦🇪 Strategic Development Fund announces investment and initial agreement with Machina Labs /Zawya/

🇺🇸💿 Unconventional AI Startup Hits $4.5 Billion Valuation in Seed Round

Unconventional AI raised $475 million in a deal co-led by Andreessen Horowitz and Lightspeed Venture Partners, said Naveen Rao, who left Databricks in September and is chief executive officer of the startup. Other investors include Lux Capital and DCVC. Databricks and Amazon founder Jeff Bezos also participated in the round, and Rao said he invested $10 million of his own funds at the same terms as other investors.

Unconventional AI is looking to biology for inspiration, as living organisms are always constrained by energy and can be quite efficient about using it. The human brain, for instance, can do tremendous work and runs on just about 20 watts of energy, Rao pointed out. (That’s enough energy to power, say, a small light bulb.) Rao said Unconventional AI is also focusing on the analog origins of computing, rather than the digital chips of today, and how the company may be able to use the intrinsic physics of semiconductors to run AI models.

Read more at Bloomberg and a16z/Naveen Rao on X

🇺🇸♨️ Fervo Energy Raises $462 Million Series E to Accelerate Geothermal Development and Meet Surging Energy Demand with Clean, Firm Power

Fervo Energy, the leader in geothermal energy development, announced the closing of its oversubscribed $462 million Series E funding round, led by new investor B Capital, a global multi-stage investment firm. This significant financing reflects strong market confidence in Fervo’s opportunity to make geothermal energy a cornerstone of the 24/7 carbon-free power future.

The financing will accelerate Fervo’s growth, supporting the continued buildout of Cape Station and early development of several other projects across its expanding pipeline. Located in Beaver County, Utah, Cape Station will begin delivering 100 megawatts (MW) of firm clean power to the grid beginning in 2026, with an additional 400 MW online by 2028 for a total of 500 MW. Cape Station is expected to be the largest next-generation geothermal development in the world. Construction is progressing rapidly, demonstrating best-in-class operational performance with drilling times decreasing and efficiencies improving with each new well.

Read more at Business Wire and CEO Tim Latimer on X

♨️ Geothermal energy’s massive leap forward /The Economist on YouTube/
🔋 Amazon Leads $81 Million Financing for Battery Startup Blue Current /WSJ/
🔋 South 8 Technologies Secures $11M in Additional Funding to Accelerate LiGas Battery Production for Defense and Aerospace /Business Wire/

🇺🇸 Steelhead Technologies Announces $84M Growth Capital Investment from Mainsail Partners

Steelhead Technologies, an all-in-one ERP software purpose-built for metal finishers and fabricators, announced an $84 million growth investment from Mainsail Partners, a growth equity firm that specializes in partnering with vertical SaaS businesses. The funding will be used to expand the platform’s capabilities and Steelhead’s suite of AI-driven tools that are designed to help shops maximize their efficiency and profitability.

🇺🇸👕 Rotostitch Raises $1M in Pre-Seed Funding to Revolutionize Apparel Manufacturing

Rotostitch, an innovative manufacturing startup in automated textile manufacturing technologies, announced that it has raised $1M in an oversubscribed pre-seed funding round, led by Boost VC and Nova Threshold. The new capital will accelerate product development, scale manufacturing capabilities, and expand go-to-market efforts.

Rotostitch Co-Founder and CTO Anson Tsang, a former Princeton materials science PhD candidate and ex-Tesla/Apple engineer will help co-lead the company’s hardware platform. “The company’s proprietary manufacturing process will deliver dramatically greater speed and versatility compared to today’s technologies. This will drive the adoption of sustainable fashion technology and automation across the industry.”