Training
Learn to steer by evidence.
AI makes explanations cheap. Judging whether they're true is still your job. Each lesson starts with a real problem, asks what you think before teaching anything, then tests whether the idea carries over to a situation you haven't seen. Competitive games are where it starts: a small world with a full record of what happened.
Problem first
You reason before you're told
Every lesson opens with an ambiguous situation. You write your reading of it, then see evidence that may change it.
Notes that come back
Mark what you don't understand
Highlight any passage or add a note. Confusing passages and questions come back for a revisit, where you record whether your understanding moved.
Transfer, not recall
Measured on new situations
Mastery counts only questions set in a different domain, answered right first time. A before-and-after check measures whether your reasoning changed.
Read
the material behind the lessonsField manual · about 3 h
Steering by Feedback
The field manual on cybernetics: where the idea came from, the anatomy of a control loop, balancing and reinforcing loops, gain and delay, variety, learning loops, the traps, and where to intervene. Nine modules with diagrams and models.
Reading · about 20 min
What to learn for the AI era
Fifteen subjects that gain value as producing code, text and analysis gets cheap: why each matters more with AI, where to start, and one resource. Mark the ones you want to learn.
Subjects
10 subjects · 19 lessons01 · 6 lessons
Cybernetics and feedback loops
How systems use information about their own state to steer: feedback, delay, stability, reinforcing and balancing loops, and learning that questions its own goals.
Why does a solution sometimes make the original problem worse?
02 · 2 lessons
Causal reasoning
Telling what happened from why it happened: confounding, reverse causality, counterfactuals, and why a good predictor is not necessarily something to change.
Did our intervention actually cause the improvement?
03 · 3 lessons
Statistics and probability
Reading numbers without being misled by them: base rates, changes in mix, and results that drift back toward average on their own.
Does a statistic tell us what we think it tells us?
04 · 1 lesson
Systems thinking and system dynamics
How parts interact to produce outcomes none of them would on their own: bottlenecks, recurring traps, and where intervention actually works.
Are we solving the underlying problem, or moving it somewhere else?
05 · 2 lessons
Decision theory and second-order thinking
Choosing under uncertainty and following consequences past the first step: expected value, reversibility, and effects that build up slowly.
What happens next, and what happens after that?
06 · 1 lesson
Cognitive psychology and epistemology
How minds build explanations and decide what counts as knowing: confirmation bias, separating observation from story, and naming what would change your mind.
What evidence would convince me that I'm wrong?
07 · 1 lesson
Scientific method and experimentation
Testing an explanation instead of defending it: controls, random assignment, measures chosen in advance.
How would we test whether this actually works?
08 · 1 lesson
Game theory and incentives
Outcomes that emerge from people with different goals: principal and agent, gaming a measure, and Goodhart's law.
Why do people act against the goals they're measured on?
09 · 1 lesson
AI literacy and model evaluation
What models do, how they fail, and how to judge their output: leakage, base rates, drift, and why fluent is not the same as right.
How do we know whether an AI recommendation is trustworthy?
10 · 1 lesson
Human factors and behaviour change
Why correct information doesn't guarantee correct action: attention, load, feedback timing and habit.
Why might someone understand advice and still not act on it?
Four questions for any claim
- What do we actually know? Separate the observation from the story told about it.
- What else could explain it? Other causes, missing information, how the data was selected.
- What would change our confidence? The evidence that would support or weaken each explanation, named before you look.
- What happens if we act on it? Immediate effects, delayed effects, and how we'd know whether it worked.