The Enginuity Digest
After being nominated as word of the year, this week’s newsletter aims to uncover the concept of “Vibe-coding” and the democratisation in potential this provides to engineers. Gone are the days of engaging in 6-month projects to build simulations or programs, prototypes can now be developed in days.
Here’s what will be covered in today’s newsletter:
News Update:
Meta’s Leading AI Scientist Pivots to World Models for Spatial Intelligence
Razor Labs Releases Mobile Fleet Predictive Maintenance AI
Datamine–Aereo Partnership Aims to Modernise Mine Surveying & Planning with AI and Drones
Vibe Coding: What it is, what it means and some key considerations
Tool of the Day: Cursor/Codex/Claude Code
AI in Application: Predictive Modelling - Water Systems
What’s been happening with AI?
Leading AI Scientists Pivot to “World Models” for Spatial Intelligence (Link) |
Two of the most influential figures in AI research, Yann LeCun (Meta) and Fei-Fei Li (The AI “Godmother”), have independently shifted their primary research efforts toward world models, a new paradigm where AI develops an internal representation of physical reality.
Traditional LLMs are incredibly powerful when it comes to learning patterns in text to predict the next word in a sequence, but have very limited understanding of objects, space, and cause-and-effect.
These world models aim to give AI the ability to simulate its environment, understand how objects move and interact, and make decisions grounded in physics. This is basically the difference between understanding how different ingredients interact versus memorising recipes.
The shift is driven by a growing consensus that current AI systems are hitting a ceiling on applicability towards robotics and complex environmental interaction, areas that human cognition and vision excel at.
Why This Matters: This is a major architectural pivot that could redefine autonomy. World models are foundational for advanced robotics and hence for Australia’s mining and heavy-industry sectors, this signals the next leap where machines can proficiently interact with their environment rather than simply observing it.
Razor Labs Releases Mobile Fleet Predictive Maintenance AI (Link) |
Razor Labs has launched a new AI platform specifically built for predictive maintenance of mobile mining fleets, covering the full spectrum of heavy vehicles including haul trucks, wheel loaders, graders, and excavators.
The platform uses sensor data, vibration patterns, temperature signals, and historical failure datasets to identify emerging faults weeks before they become operational issues.
The solution integrates into existing fleet telematics and maintenance systems, enabling automated diagnostics, early failure alerts, and detailed root-cause analysis.
Razor Labs positions this as the final piece of a complete predictive ecosystem, complementing its fixed-plant AI and visual anomaly detection systems to create a unified view of both stationary and mobile assets.
Why This Matters: Mining companies lose millions annually to unplanned shutdowns and catastrophic component failures. This shift to AI-first maintenance has the potential to extend equipment lifespan, improve safety, and materially increase throughput.
Datamine–Aereo Partnership Aims to Modernise Mine Surveying & Planning with AI (Link) |
Datamine has partnered with Aereo to combine drone-based imaging, aerial monitoring, and AI analytics into an integrated mine planning and operational intelligence platform.
The system can automatically convert drone captures into 3D terrain models, volumetrics, compliance reports, and plan-vs-actual updates.
By merging Aereo’s high-frequency aerial data with Datamine's mine planning suite, operators can now access near real-time insights on material movement, bench geometry, stockpile changes, and road condition analysis.
Why This Matters: This partnership moves the mining sector closer to continuous digital oversight, where operational decisions are made using current, high-resolution data rather than periodic survey snapshots. This kind of integrated AI imaging platform will drive safer haul roads, tighter planning accuracy, and more efficient pit development.
Vibe Coding: A Breakdown
Vibe coding sounds a whole lot better than my usual frustration coding
So, what is it?
2025 has brought with it the exponential rise in vibe coding, so much so that it was even chosen as the Collins Dictionary Word of the Year. At its core, vibe coding means using natural language (i.e. plain English), not Python or C++, to tell an AI what kind of system you want to build. Instead of writing syntax, you describe the intent or “vibe” of the code, and the AI takes care of the details.
You tell an AI system something like:
“I want a small tool where I can adjust turbidity and see how it affects a filter’s loading rate.”
“I want a dashboard that displays pump energy use and lets me compare peak vs off-peak periods.”
“I want a calculator that shows how different materials affect thermal expansion in a pipeline.”
The AI generates the working prototype based on your description, including buttons, graphs, logic, and all and then you refine it through conversation. It’s basically engineering problem-solving expressed in plain language, with the AI doing the technical assembly behind the scenes.
Vibe Coding as a Democratising Force
Notably, this approach is now democratizing early-stage innovation. Engineers who understand processes, constraints, safety, and system behaviour (essentially people who may be domain experts but not software experts), can now turn ideas into testable tools without engaging a full software team across weeks. Ultimately this leads to:
Faster prototyping
More experimentation
Clearer communication with clients
Better alignment with software teams later in the process
The Reality of Vibe Coding
Whilst an insanely powerful tool, vibe coding still isn’t magic. It relies on real code underneath, and that still means someone has to understand how to integrate, debug, and secure what the AI produces. LLMs can misinterpret intent or generate brittle logic if left unchecked. So for organisations, this isn’t a replacement for software teams, it simple allows for better early-stage prototyping and ideation. The best results come when experienced engineers use it to get ideas moving quickly, then collaborate with software experts to create real production ready software.
In short: vibe coding lets engineers build more, test more, and innovate more, all without needing to become programmers. It’s a shift worth watching.
Tools of the Week: Cursor/Codex/Claude Code
Since we’re talking about vibe coding and intent-driven development, it’s worth highlighting the tools that make this possible. While there are dozens of emerging platforms, three stand out as the most capable, most widely adopted, and most relevant for engineers
Codex: Created by OpenAI, Codex is usable by anyone with a ChatGPT subscription, and integrates quite nicely with workflows on Github. For engineers who already dabble in Python, MATLAB-style computations, or data analysis, Codex-powered tools are a strong starting point.
Claude Code: Claude Code is rapidly becoming the favourite among professional software developers. It’s particularly strong in long-context reasoning—meaning it can interpret and work across large, complex codebases or multi-file projects.While it may be more powerful than what most engineers need for early-stage prototyping, its stability and accuracy make it a standout for those collaborating with software teams or managing more involved digital projects.
Cursor: Cursor is the closest thing to a “pure vibe coding experience.” It’s fast, lightweight, and extremely user-friendly. After downloading it, users can start describing ideas in plain language and watch the system scaffold applications, create files, and update logic as they refine their intent. The free plan is generous enough for experimentation, making Cursor an excellent entry point for engineers who want to explore concepts.
For simplicity and approachability, Cursor is the best place to start, especially if you want to experiment with vibe coding and see how quickly ideas can turn into functioning digital tools, however the best choice ultimately comes down to your preferences and what you have access to.
AI in Application: Predictive Modelling in Water
Now all this vibe coding talk is interesting, but what does it actually look like in application.
Imagine a regional water utility who is responsible for managing a distribution network with multiple reservoirs, pressure zones, and booster pumps. These systems are complex: water age fluctuates, chlorine residual decays over distance, seasonal demand spikes stress certain mains, and operational decisions (like adjusting pump speeds or opening a bypass valve) have cascading flow and water-quality impacts.
Traditionally, exploring these “what-if” scenarios would require planning and engaging in a comprehensive, multi-month development project to make a full hydraulic modelling software (or large engagement of a consultancy). This, however, is where vibe coding offers a powerful new workflow.
Using a tool like Cursor or Codex, an engineer could describe the desired functionality in simple terms: “Build a dashboard where I can adjust pump speeds, tank levels, and valve positions, and see how pressures, flows, and chlorine residual change across the network.” This engineer could provide specifications, context, and a breakdown of the key equations governing the hydraulics (or not!). The AI can then scaffold a working prototype: sliders for operational inputs, a basic hydraulic approximation, simple chlorine-decay calculations, and graphs showing system response.
This prototype is likely to be far from right, however through iteration and testing, a water specialist engineer could explain pain points and ideate solutions with the AI vibe coding assistant to rapidly prototype and test ideas.
Ultimately this system would enable for the answering of key operational questions such as:
What happens if a reservoir is taken offline?
How does reducing pump speed affect downstream pressure?
Does water age exceed limits in low-demand periods?
The real value is speed. Instead of weeks of development, a prototype can emerge in hours. Iteration becomes conversational, adjusting equations, adding limits, and refining assumptions directly with the AI. Then, once the logic is validated, the prototype can be handed to a software team for integration with real data, SCADA systems, or digital-twin platforms.
As such vibe coding can, and already does, empower engineers to create interactive tools that capture their domain knowledge, help stakeholders understand system behaviour, and accelerate the journey from concept to usable digital product.
In Other News
OpenAI releases GPT-5.1, with improvements to efficiency, reasoning, and personalities. This release was unexpected but widely praised as exceeding expectations - Link
Waymo, Google’s autonomous vehicle startup, becomes the first robotaxi to offer driverless rides on freeways after over 100 million miles of successful testing - Link
If you haven’t seen it, Russia’s first humanoid robot was presented to the world to the backdrop of victorious music, before falling flat on its face 30 seconds in. Video - Link
Jeff Bezos’ Blue Origin company succesfully landed its rocket’s booster for the first time, something only previously done by SpaceX - Link
Berkshire Hathaway, Warren Buffet’s company that is typically averse to investments in tech, reveals a $4.3 Billion investment in Alphabet (Google) - Link
Scientists built a Digital Twin of Earth (yep, you read that right!) at 1km resolution, potentially unlocking the most powerful weather forecasting system in existence - Link
The biggest solar flare in nearly two decades just hit, reigniting fears about how a major flare could disrupt or permanently damage satellite electronics and the infrastructure we rely on every day - 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.
