Metacognition
Cognitive Tools: 7 Ways to Think Differently
Цели урока
- Master inversion thinking and apply it to system design
- Understand Predictive Processing as an accelerated learning tool
- Learn to switch between focused and defocused modes
- Master Dimensional Hopping for escaping dead ends
Four thinking tools used by Nobel laureates, billionaires, and 10x engineers. None of them require high IQ.
- Munger uses inversion for investment decisions
- Feynman won the Nobel Prize thanks to a defocused-mode insight
- Testing effect confirmed in 200+ experiments
- Google X uses dimensional hopping for moonshot projects
Предварительные знания
Inversion Thinking: Start From the End
**Charlie Munger**, Warren Buffett's partner, said: *"Tell me where I'm going to die, so I'll never go there."* This isn't a joke - it's one of the most powerful cognitive tools.
**Inversion** - instead of 'How do I succeed?' ask 'How do I **guarantee failure**?' Then invert every point.
Why inversion works: the brain is much better at generating **concrete failures** than abstract successes. 'What can go wrong' produces stories. 'What should go right' produces wishes.
**Practice:** Before your next code review, try inversion. Instead of 'Is this good code?' ask 'How could this code break in production?'
You're designing a notification system. Applying inversion, what question should you ask FIRST?
Predictive Processing: Predict to Learn
Neuroscientist Karl Friston proposed a radical idea: **the brain is a prediction machine**. It constantly generates expectations and updates them when reality doesn't match.
As a cognitive tool, Predictive Processing is simple: **make a prediction BEFORE you learn the answer**. Prediction error = learning signal.
This is experimentally confirmed. **Testing effect** (Roediger & Karpicke, 2006): attempting to recall before seeing the answer improves retention by 50-100%. **Generation effect**: a self-generated answer (even if wrong) is remembered better than one that was read.
**Important:** The prediction must be SPECIFIC. 'Probably something about math' is useless. 'I think this algorithm is O(n log n) because it uses divide-and-conquer' is useful. Specificity = ability to be wrong = learning.
Prediction errors mean you're thinking poorly
Prediction errors are GOOD. They're the most powerful learning signal
The brain updates its world model precisely at the moment of surprise (prediction error). No surprise = no learning. If you always guess right, you're not learning - just confirming what you already knew.
You've opened docs for a new API. How do you apply Predictive Processing?
Defocused Attention: Think Out of Focus
In 1966, physicist Richard Feynman told how the idea for quantum electrodynamics (which earned him a Nobel Prize) came to him: he was watching a student toss a plate in the cafeteria. Not thinking about physics. Just watching.
The brain operates in two modes - and both are necessary:
Research by Martindale (1995) showed: creative people differ not in the strength of focus or defocus, but in their ability to **switch** between them. Ordinary people get stuck in one mode.
**Practical protocol:** Focused → Defocused → Focused. Work 25-45 minutes in focus → take a break WITHOUT screens (walk, shower) → return. The first 5 minutes after returning are the most productive. Your brain processed the problem in the background.
This isn't mystical. The Default Mode Network (DMN) - a brain network active during the defocused state - literally searches for connections between disparate elements. It doesn't turn off when you 'rest' - it **works on the problem differently**.
You've spent 2 hours unable to find an architectural solution. All options seem bad. What do you do?
Dimensional Hopping: Add a Dimension
Any problem can be described in N dimensions. Most people get stuck in the 2-3 obvious ones. Adding a new dimension opens **an entire class of solutions invisible from the familiar axes**.
Technique: when stuck, ask 'What DIMENSION haven't I considered yet?' Typical hidden dimensions: time (when?), cost (how much?), social (who else?), emotional (how does it feel?), energy (how much effort?).
**Dimensional Hopping + Inversion = combo.** Inversion in a new dimension yields even more unexpected insights. Example: 'How to INCREASE cache latency?' (inversion) in the predictability dimension → 'Oh, we could deliberately slow down cold starts to motivate pre-warming.'
You need to optimize a task across all dimensions simultaneously
Dimensional hopping is a SEARCH tool, not an optimization tool. Find the right dimensions, then optimize 2-3 key ones
Optimizing across 10 dimensions simultaneously is NP-hard. The value is in FINDING the non-obvious dimension that changes everything, not trying to optimize everything.
A team is discussing CI/CD pipeline improvements. Everyone is focused on build speed. What NEW dimension to add?
Summary
- Inversion: think from failure to success. Failures are more concrete than dreams
- Predictive Processing: predict before you learn. Error = learning
- Defocused Attention: switch focused → defocused → focused. DMN searches for connections in background
- Dimensional Hopping: when stuck - add a dimension you haven't considered yet
Next
In the next lesson - Recursive Self-Modeling: how to turn knowledge about your thinking into a superpower
- Recursive Self-Modeling — Applying cognitive tools to yourself
Вопросы для размышления
- Which of the four tools feels most natural to you? Which do you ALREADY use unconsciously? And which feels most alien - that's the one to start practicing.