Before you begin
- Math ladder: R17, R18, R21

Lesson 01 · Section 1 of 7 · Reading
The original feedback had four pieces. "I don't know what it is" — Chapter 0. "I don't understand the words" — Chapter 2. "I don't know how to calculate" — Chapters 3 and 4.
The original feedback had four pieces. "I don't know what it is" — Chapter 0. "I don't understand the words" — Chapter 2. "I don't know how to calculate" — Chapters 3 and 4.
This is the fourth: "I don't know how to decide the right things."
And it's the one the arithmetic doesn't answer. Chapter 4 told you what the engine computes and why. This chapter is about the moment after the engine has finished — when the numbers are on screen, correct, and a choice is still yours.
Every operation in this chapter was taught in Chapter 3 or 4: expected value, marginal thinking, asymmetry, decomposition, base rates. Nothing new to learn mechanically.
What's new is the question you ask of the numbers, and it turns out most bad decisions in this domain are not arithmetic failures. They're one of about a dozen recognisable moves made in the wrong place.
"Ignore sunk costs" is a slogan. "Continuing the deal you've spent 60 hours on destroys $73k against redeploying" is an argument.
So every idea in this chapter is anchored to a figure a real module produces, and the anchors are unusually large. Three examples of what a soft criterion costs:
| The soft version | What it costs |
|---|---|
| "Work the biggest logo" | 55.8% of achievable value (pql-01) |
| "We're so close, let's finish it" | $76k against redeploying (pql-03) |
| "Same quota for everyone, that's fair" | half the team's output — $1,200k of $2,400k (scd-04) |
Those aren't rhetorical. They're what the engines compute.
| # | Piece | The move it teaches |
|---|---|---|
| s01 | What a decision actually is | Three parts — and if one is missing, it isn't a decision |
| s02 | Evidence versus instinct | Instinct is fast, and 56% optimistic here |
| s03 | Framing | "What caused it?" and "what should we do?" have different answers |
| s04 | Find the lever | Three of four plausible actions are worth exactly $0 |
| s05 | Base rates | Start from the population; the modal outcome is nothing |
| s06 | Probability applied to a deal | Calibration: saying 70% and meaning it |
| s07 | Sunk cost | The 40 hours appear in neither number, and that's the point |
| s08 | Opportunity cost | The whale gets cut — highest value, still cut |
| s09 | Expected value as a rule | Then check the downside |
| s10 | Asymmetric errors | Which way is cheaper to be wrong — computed, never assumed |
| s11 | Deciding at the margin | The next unit, not the total |
| s12 | Reversible versus irreversible | One discount is $30k; a precedent is $1,500k |
| s13 | Bias versus noise | Two problems, opposite fixes, and sign matters |
| s14 | Decision versus outcome | The better bet loses 3 times in 10 |
| s15 | Power and alternatives | The +$20k came from away from the table |
| s16 | One yardstick, and the journal | The habit that makes the rest stick |
Then D06 — Decision Drills (three stations, about two minutes), and a recall set.
Same taxonomy as Chapter 4: twelve outcome traps where the naive choice loses money you can point at, and three reading traps where the naive number is real and simply isn't the quantity the decision turns on.
But this chapter has one thing Chapter 4 didn't, and it's worth watching for. In sls-rnx-02, two completely different wrong instincts produce the identical wrong answer: "save the biggest logos" and "save the most at-risk accounts" both return $152k against an achievable $322k.
Two independent reflexes agreeing feels like corroboration. It isn't — they're wrong for unrelated reasons and happen to land in the same place. That's the most dangerous shape a mistake can take, because the usual check (does anyone else read it differently?) returns a false all-clear.
None of that is about numbers. All of it is about which number you look at.