Unpopular opinion: fidelity is the most overrated metric in digital twins. A low-fidelity twin that changes a decision is worth more than a photorealistic model nobody opens.

Digital twins are not an IT project. They're an operations project with technology in it. Get the org chart wrong and the technology doesn't matter.

The biggest risk in applied AI isn't the model being wrong. It's the model being right and nobody trusting it enough to act.

"We need more data" is the industrial equivalent of "we need more meetings." You don't need more data. You need better questions.

AI without a digital twin is a brain without a body. It can think. It can predict. But it can't see the operational context it's predicting into.

Stop calling it a proof of concept. Start calling it a proof of value. POCs are designed to impress. POVs are designed to change something.

If your digital twin vendor can't explain the problem they're solving in one sentence, they don't understand the problem.

The industrial metaverse will fail for the same reason most digital twins fail: Technology looking for a problem, instead of problems looking for technology.

new skill unlock : AssignBlame.md

ahhhh... HR involved..... 2nd unlock: AvoidBlame.md

Hot take: the best digital twin you can build right now might be a well-structured Claude skill that uses real world data and that everyone actually uses. Start there. Then evolve.

3D visualisation is a feature, not a strategy. If your digital twin strategy starts with "let's build a 3D model" you've already lost.

Every failed twin I've seen started with the same sentence: "Let's just get all the data in first and the use cases will emerge." They don't.

Your digital twin doesn't need to be real-time. It needs to be right-time. Some decisions need data every second. Most need it every shift.

Adoption isn't a training problem. It's a design problem. If people aren't using your twin, you designed it for the technology, not the user.

Unpopular opinion: most organisations aren't ready for a digital twin. They need a better spreadsheet, a cleaner data model, and an agreed definition of the problem first.

Applied AI in mining isn't about replacing humans. It's about giving a shift supervisor the same situational awareness at 3am that they have at 10am.

The real AI adoption curve in heavy industry: 1. Excitement 2. Pilot 3. Pilot works in controlled environment 4. Pilot fails in production 5. Blame the data 6. Actually fix the data 7. Quiet, boring success

AI predictions without operational context are just numbers. Digital twins turn numbers into decisions.

The gap between an AI model and a changed outcome in the real world is usually: - Trust - Context - A frontline worker who was never consulted Technology is the easy part.