Models

The most useful mental models I've come across, and how I use them.

• Model 1

  First principles thinking

  Break a problem down to the basic facts you know are true, then reason up from there.

  Instead of reasoning by analogy (“this is how it’s always been done”), ask what is physically or logically required. It is how SpaceX challenged the cost of rockets by pricing the raw materials. Useful whenever an industry’s accepted costs or limits look like habit rather than physics.

  Origin: Aristotle; popularised by Elon Musk

• Model 2

  Inversion

  Solve a problem backwards by asking what would guarantee failure, then avoid it.

  “Invert, always invert.” Rather than asking how to make a system secure, ask how you would break it. Listing the ways something fails is often easier and more concrete than describing success, and it surfaces risks optimism hides.

  Origin: Carl Jacobi; Charlie Munger

• Model 3

  Second-order thinking

  Look past the immediate result of a decision to the consequences of those consequences.

  First-order thinking asks “what happens next?”; second-order asks “and then what?”. A cheaper energy source, for example, changes demand, grids and geopolitics, not just bills. Useful for judging new technologies whose biggest effects are indirect.

  Origin: Howard Marks

• Model 4

  Vertical integration

  Own more of the chain that makes your product, from raw inputs to the customer, instead of buying each step from others.

  Carnegie owned the mines, railways and mills behind his steel. SpaceX builds most of its rockets in-house, and Apple designs its own chips. I use it to ask where a company’s real advantage comes from: controlling the key steps can cut cost, speed up iteration and protect quality, but it also ties up capital and focus, so it only pays where the step really matters.

  Origin: Business strategy; Andrew Carnegie, and today SpaceX, Tesla and Apple

• Model 5

  Systems thinking

  See a problem as a set of connected parts, feedback loops and delays, rather than a chain of isolated causes.

  Meadows showed that a system’s behaviour comes from its structure: stocks, flows and the loops between them. I use it when a fix seems obvious, to ask what pushes back. Power grids, AI adoption and security incidents all produce surprises once delays and feedback kick in, and the most effective levers are often goals and incentives, not individual parts.

  Origin: Jay Forrester; Donella Meadows

• Model 6

  The 5-step algorithm

  Question every requirement, delete parts or steps, simplify, accelerate, then automate, in that order.

  1. Make every requirement less dumb, and give each one a named owner. 2. Delete any part or process you can; if you don’t end up adding some back, you didn’t delete enough. 3. Simplify or optimise only what’s left. 4. Speed up the cycle time. 5. Automate last. I find the order is the point: the most common mistake is optimising or automating something that shouldn’t exist at all.

  Origin: Elon Musk

• Model 7

  How to think and write critically

  Three levels of analysis that move from facts, to meaning, to implications.

  1. Descriptive analysis: who, what, where, when. 2. Insightful analysis: how and why. 3. Foresight analysis: so what, what if, and what’s next. I use it to structure my take on each piece in the weekly roundup: get the facts straight, explain what’s driving them, then say why it matters and where it could lead.