The Enginuity Digest
Welcome back. While the US ramps up its "AI Arms Race" rhetoric, a quiet revolution is happening in the industry control rooms and workshops, moving from systems that fix things when they break to systems that refuse to let them break in the first place.
Here’s what will be covered in today’s newsletter:
News Update:
The New Manhattan Project: The US Commission calls for a state-led race to AGI.
Aristotle Solves the Unsolvable: Harmonic’s AI cracks a 30-year-old math problem, ushering in the era of "Vibe Proving."
Deep Dive: AI for Asset Management - From Reactive to Autonomous.
In Other News: Updates in engineering, AI and technology.
What’s been happening with AI?
US Launches "Genesis Mission" for AI in Science (Link) |
President Trump has signed an Executive Order launching the "Genesis Mission," a massive initiative led by the Department of Energy to turn the US government’s network of databases and supercomputers into a unified AI development platform.
The mission aims to double the productivity of American science and engineering within a decade by creating "scientific foundation models" capable of automating research and testing hypotheses.
The White House has explicitly likened the scale of this effort to the wartime Manhattan Project. It involves creating a "closed-loop AI experimentation platform" that links 17 national laboratories, aiming to accelerate discoveries in fusion energy, new materials, and biotechnology.
Why this matters: This marks a pivotal shift where AI moves from a tool for scientists to becoming the scientist itself. However, experts warn this could change research "for better or worse," raising concerns about AI hallucinating clinical trial data or opening the door to new biosecurity risks if not globally coordinated.
Aristotle Solves 30-Year-Old Mathematics Problem (Link) |
Aristotle, an AI system built by Harmonic, has independently solved a version of Erdős Problem #124, a combinatorics problem that has remained open since the 1990s.
The system didn't just guess; it solved the problem in six hours and then formally verified the proof in the Lean programming language in just one minute.
Founder Vlad Tenev calls this the arrival of "vibe proving"—where AI explores the "vibe" or intuition of a proof, then locks it down with machine-verifiable rigor.
Why this matters: We are moving toward mathematical superintelligence. If AI can solve open problems in pure math, it can theoretically verify complex engineering control logic that is currently too complex for humans to prove "bug-free." It turns advanced mathematics from a field for the elite few into a scalable, collaborative tool.
Autonomous Asset Management
If you walk into a traditional processing plant, you’ll likely see a maintenance planner staring at a spreadsheet, trying to figure out if Pump A needs a service now or if it can wait until next month’s shutdown. It’s a high-stakes guessing game. Service it too early? You waste money on parts and labor. Service it too late? You face a catastrophic failure, downtime, and safety risks.
Recently however, a shift has begun for asset management, moving it from what was once a strategy of "Run to Failure" to now seeking Autonomous Reliability.
The Evolution of Maintenance
To understand where we are going, we have to look at the hierarchy of maintenance maturity:
Reactive (Run-to-Failure): The "don't touch it until it smokes" method. It’s cheap upfront but expensive when a $50 bearing failure destroys a $500,000 gearbox and halts production for three days.
Preventive (Calendar-based): : A shift to replacing parts based on Mean Time Between Failures (MTBF). If a bearing was rated for 10,000 hours, we replaced it at 9,500, regardless of its condition. This traded the risk of catastrophe for the certainty of waste.
Predictive (Condition-based): Using sensors (vibration, temperature, oil analysis) to monitor health and therefore allowing us to understand the machine, and not just read the calendar (I.e.The vibration is high, so let's schedule a fix) This is where many top-tier industry players have been for the last decade.
Prescriptive & Autonomous (AI-driven): The new frontier. The system doesn't just tell you something is wrong; it tells you what is wrong, when it will fail, and how to fix it. In fully autonomous systems, it might even order the spare part and schedule the work order itself.
The Potential For AI
AI, specifically Machine Learning (ML), changes the game because it excels at pattern recognition across multivariate data sets.
A human engineer can look at a vibration trend and say, "That bearing is loose." But an AI can look at the vibration, the motor current, the ambient temperature, the discharge pressure, and the fluid viscosity simultaneously. It can learn that when the ambient temp drops and pressure spikes, a specific vibration signature is actually normal, but if the pressure is stable and vibration rises, it indicates mechanical fatigue developing on a main internal bearing surface.
Furthermore, one of the most powerful applications is creating "virtual sensors." Imagine a critical pipe where you can't physically install a flow meter due to slurry abrasion. By training an AI on the pump speed, power draw, and valve position, the model can infer the flow rate with 99% accuracy without a physical sensor ever touching the fluid. This is risky, but becoming more and more viable as the technologies develop.
Real World Examples:
Here are two case studies demonstrating the scale and ROI of AI in asset management.
1. Shell: Scaling to 10,000 Assets Shell faced a massive "scaling" problem: building a bespoke model for one control valve is easy; maintaining bespoke models for 10,000 valves across 20 countries is a logistical nightmare.
The Strategy: They partnered with C3 AI to build "Type Models"—one robust AI model for a specific class of valve. They then deployed this single model to thousands of instances, using "Transfer Learning" to tweak it for local operating conditions (e.g., adjusting for the ambient temperature difference between Alaska and Nigeria).
The Outcome: The system now monitors over 10,000 pieces of equipment, identifying degradation weeks in advance. It proves that standardization is the secret sauce to scaling AI.
2. Rio Tinto: The Autonomous Mine In the Pilbara, Rio Tinto’s "Mine of the Future" program has moved beyond just driverless trucks—it has moved to system-level preservation.
The Strategy: Their autonomous haulage system doesn't just predict engine failure; it predicts road failure. By analyzing suspension data from the trucks, the AI identifies potholes and corrugations in the haul road that are damaging the tires. They fix the road to save the truck.
The Outcome: Autonomous trucks now operate 700 hours more per year than human-driven ones (no shift changes, no fatigue). It is a prime example of AI creating consistency that physically extends the life of the asset.
Integrating AI (It’s Not Just Magic)
Unfortunately, You can't just sprinkle "AI dust" on a 1980s coal plant and expect miracles. There is a hierarchy of needs for AI integration:
Sensor Density: You need eyes and ears. This means retrofitting assets with cheap IoT vibration sensors, acoustic monitors, and thermal cameras.
Data Historian: You need clean, time-series data. If your data is locked in a silo or overwritten every week, the AI can't learn the history of failure.
Connectivity: High-bandwidth, low-latency connection (Private 5G or Starlink) is crucial for remote sites. The AI needs to send the data to the cloud (or an Edge compute device) to process it.
The "Human in the Loop": This is critical. AI makes recommendations; it doesn't sign off on safety critical isolation permits. The reliability engineer’s role shifts from data gathering to data interpretation and strategy.
The Bottom Line AI in asset management isn't about replacing the maintenance crew. It's about giving them a superpower. It’s about ending the 3 AM call-outs because a pump failed unexpectedly. It’s about squeezing that extra 2% of efficiency out of a billion-dollar asset. In an economy defined by efficiency and volatility, that 2% can be the difference between profit and loss.
In Other News
The Australian resources sector is seeing a record surge in mergers in Gold and Copper. With inflation sticking and geopolitical tension rising, the big players (like Northern Star) are choosing to buy existing mines rather than risk the decade-long timeline to build new ones. (Link)
Toyota claims to have solved the durability issue with solid-state batteries, promising a 1200km range EV by 2028. (Link)
MIT Study Finds AI can replace 11.7% of US jobs at current capabilities, particularly in the administrative and financial services sectors (Link)
DeepSeek, China’s leading AI startup has released two new AI models to compete with ChatGPT and Gemini, excelling in particular in mathematical benchmarks (Link)
An AI-generated song just hit #1 on Billboard as many large music labels partner with music generating AI companies. For a deep dive into this check out this article: (Link)
This newsletter seeks to engage and challenge the way engineers see AI and its potential for application in industry. Any thoughts, questions or arguments are welcome! Finally, if you enjoy the content, please refer it on to your friends and colleagues, I would love the support.
