The Parent's Toolkit

Four tools for the actual moments: when she's stuck, when you've forgotten something, when you want to check whether she really has it, and when you're deciding whether to say anything at all.

Who this is for

This document is addressed to a parent, and it talks about a student in the third person. She is welcome to read every word of it — nothing here is meant to be kept from her, and it is better that she knows what it says than that she finds out.

It is not an authority. It is a second voice on the same material, written so that there is someone else in the house who can follow what she is doing. Where it disagrees with her teacher, her teacher is the one grading.

Tool 1 — The Rederivation Map

The point of the whole guide is that almost nothing in BC needs to be memorized in isolation. This is the dependency map: if she blanks on the thing in the left column, rebuild it from the thing in the right.

15 of the 20 rows open into the steps. The other 5 are one-liners already — there is genuinely nothing more to say about them, and padding them out would make the arrows meaningless.

Forgotten	Rebuild from	Comes up in
	Product rule. Q = f/g ⇒ f = Qg ⇒ f′ = Q′g + Qg′ ⇒ solve for Q′.	Unit 2 onward, constantly
Rebuilding Quotient rule Let Q = f/g, so f = Qg. Differentiate with the product rule: f′ = Q′g + Qg′. Solve for Q′: Q′ = (f′ − Qg′)/g. Put Q = f/g back in and clear the fraction: Q′ = (f′g − fg′)/g². Worked at length in Rebuild 1 below.
	Divide sin²+cos²=1 by cos². (By sin² gives cot/csc.)	Units 3, 6, 9
Rebuilding tan²+1 = sec² Start from the identity nobody forgets: sin²θ + cos²θ = 1. Divide every term by cos²θ: tan²θ + 1 = sec²θ. Divide the same identity by sin²θ instead: 1 + cot²θ = csc²θ. Worked at length in Rebuild 2 below.
	Quotient rule on sin/cos, then sin²+cos²=1 on top.	Unit 3
Rebuilding d/dx[tan x] Write tan x = sin x / cos x and apply the quotient rule. Numerator: cos x·cos x − sin x·(−sin x) = cos²x + sin²x. That is 1, so the derivative is 1/cos²x = sec²x.
d/dx[sec x]	Write as (cos x)⁻¹, chain rule.	Unit 3
	Velocity on the unit circle is position rotated 90°: (cos t, sin t) → (−sin t, cos t).	Unit 3; reappears in Unit 9
Rebuilding Both sine and cosine derivatives at once A point on the unit circle is (cos t, sin t). Its velocity is tangent to the circle, the same length, turned a quarter turn forward. Rotating (x, y) by 90° gives (−y, x), so the velocity is (−sin t, cos t). Read off componentwise: (cos t)′ = −sin t and (sin t)′ = cos t. Both derivatives, from one picture.
	sin y = x or tan y = x, differentiate implicitly, solve, convert with the Pythagorean identity.	Unit 3; reversed in Unit 6
Rebuilding d/dx[arcsin], d/dx[arctan] Write y = arcsin x as sin y = x. Differentiate implicitly: cos y · y′ = 1, so y′ = 1/cos y. Convert back: cos y = √(1 − sin²y) = √(1 − x²), giving y′ = 1/√(1 − x²). Same route for arctan: tan y = x ⇒ sec²y·y′ = 1 ⇒ y′ = 1/(1 + x²), using sec² = 1 + tan².
	e^y = x ⇒ e^yy′ = 1 ⇒ y′ = 1/x.	Unit 3
Rebuilding d/dx[ln x] Write y = ln x as e^y = x. Differentiate both sides: e^y·y′ = 1. e^y is just x, so y′ = 1/x.
d/dx[a^x]	a^x = e^{x ln a}, chain rule.	Unit 3
	Sector squeeze: sin x ≤ x ≤ tan x, from three nested areas.	Unit 1; powers all of Unit 3’s trig
Rebuilding sin x / x → 1 Draw a unit-circle sector of angle x, with the inscribed triangle inside it and the tangent triangle outside. Compare the three areas: ½sin x ≤ ½x ≤ ½tan x. Divide through by ½sin x: 1 ≤ x/sin x ≤ 1/cos x. cos x → 1, so the outer bounds close and the squeeze forces sin x / x → 1.
	The same inequality chain. It generates all of them.	Unit 1
Rebuilding Any other trig limit at 0 Everything reduces to sin u / u → 1, so make the argument match the denominator. (1 − cos x)/x: multiply above and below by (1 + cos x) to get sin²x / (x(1 + cos x)) → 0. The same manoeuvre over x² leaves ½ — which is why those two limits differ. tan x / x = (sin x / x)(1/cos x) → 1.
Why +C exists	MVT: same derivative ⇒ differ by a constant.	Units 5 and 6
Sector area ½r²θ	Fraction θ/2π of the full circle πr².	Units 1 and 9
Polar area ½∫r²dθ	The sector formula, used as a slice.	Unit 9
	ds² = dx² + dy². Factor out dx (or dt).	Units 8 and 9
Rebuilding Arc length A short piece of curve is the hypotenuse of a tiny right triangle: ds² = dx² + dy². Factor out dx²: ds = √(1 + (dy/dx)²) dx. Factor out dt² instead, for a parametric curve: ds = √((dx/dt)² + (dy/dt)²) dt. Integrate ds across the interval. Every arc-length formula in the course is this one line, factored differently.
	S − rS telescopes to a − arⁿ. Divide.	Unit 10
Rebuilding Geometric series sum Write S = a + ar + ar² + … + arⁿ⁻¹. Multiply by r: rS = ar + ar² + … + arⁿ. Subtract. Everything in the middle cancels: S − rS = a − arⁿ. Divide: S = a(1 − rⁿ)/(1 − r). If \|r\| < 1 then rⁿ → 0 and the sum is a/(1 − r).
	Differentiate the generic power series n times, set x = a. The n! is what n-fold differentiation makes.	Unit 10
Rebuilding Taylor coefficients Suppose f(x) = Σ c_n(x − a)ⁿ and go looking for the coefficients. Set x = a: every term but the first dies, so c₀ = f(a). Differentiate once and set x = a again: c₁ = f′(a). Differentiate n times and the nth term becomes n!·c_n. Hence c_n = f⁽ⁿ⁾(a)/n! — the factorial is what repeated differentiation leaves behind.
	Substitute / differentiate / integrate / multiply one of the four memorized ones.	Unit 10
Rebuilding Any Taylor series Memorise only four: e^x, sin x, cos x, and 1/(1 − x). Substitute into one: cos 2x from cos u with u = 2x. Differentiate or integrate one term by term: ln(1 + x) from 1/(1 + x). Multiply by a power of x: x·cos 2x from the series for cos 2x. Worked at length in Rebuild 3 below.
	It’s the next term you dropped, with the derivative at its worst case.	Unit 10
Rebuilding Lagrange error bound The error is simply what you threw away — everything past the last term you kept. It is dominated by the first omitted term, the one of degree n+1. You do not know the derivative there, so use its worst value on the interval: \|R_n\| ≤ M\|x − a\|ⁿ⁺¹/(n+1)!. For an alternating series you can do better: the error is smaller than the first omitted term outright, no M needed.
	Both are “step along the tangent line.” Set the tangent to zero, or follow it for h.	Units 4 and 7
Rebuilding Newton’s and Euler’s formulas Both say the same thing: stand at a point you know and step along the tangent line. Newton wants a root, so he steps to where the tangent crosses zero: x_n+1 = x_n − f(x_n)/f′(x_n). Euler wants the next value, so he follows the tangent forward a distance h: y_n+1 = y_n + h·f′(x_n, y_n). One picture, two questions. If you can draw the tangent you can rebuild either formula.
	It cancels a common factor using tangent lines, so both must pass through zero. Hence 0/0.	Unit 4
Rebuilding Whether L’Hôpital applies It applies only to 0/0 or ∞/∞. Check the form first, every single time. The reason: near a shared root, f and g both look like their tangent lines. Those are f′(a)(x − a) and g′(a)(x − a). The (x − a) cancels, leaving f′/g′. If it is not an indeterminate form there is no common factor to cancel, and the rule is not merely unhelpful — it is false.

Three of these worked out, because “rebuild it” is easier said than done

Rebuild 1 — the quotient rule, in four lines

When this happens: mid-test, she needs the derivative of (x²+1)/(x−3), and she can’t remember whether it’s f′g − fg′ or fg′ − f′g on top. Guessing is a coin flip and a wrong sign kills the whole problem.

Let Q = f/g. Then f = Q·g.
Product rule on that: f′ = Q′g + Qg′.
Solve: Q′ = (f′ − Qg′)/g.
Substitute Q = f/g and multiply top and bottom by g: Q′ = (f′g − fg′)/g².

Faster sanity check if she just needs the sign: try f = x, g = 1. The answer must be 1. Plugging in, (1·1 − x·0)/1 = 1. ✓ The reversed version gives −1. ✗ Ten seconds, no derivation.

Rebuild 2 — a trig identity she half-remembers

When this happens: a Unit 6 integral needs ∫sec²x dx or ∫tan²x dx, and she recalls there’s a relationship between tan and sec but not which way it runs.

Start from the one identity nobody forgets: sin²θ + cos²θ = 1.
Divide every term by cos²θ: (sin²/cos²) + 1 = 1/cos².
Translate: tan²θ + 1 = sec²θ.

So ∫tan²x dx = ∫(sec²x − 1)dx = tan x − x + C. The identity she couldn’t recall was five seconds of division away, and dividing by sin² instead gives the cotangent version if that’s what she needs.

Rebuild 3 — a Taylor series she never memorized

When this happens: Unit 10 asks for the Maclaurin series of x·cos(2x) through the x⁵ term. Computing five derivatives of that product by hand is fifteen minutes and three chances to slip.

Start from the memorized cosine series: cos u = 1 − u²/2! + u⁴/4! − ⋯
Substitute u = 2x: cos 2x = 1 − 4x²/2 + 16x⁴/24 − ⋯ = 1 − 2x² + (2/3)x⁴ − ⋯
Multiply by x: x − 2x³ + (2/3)x⁵ − ⋯

Thirty seconds, no derivatives taken. The general rule: never compute Taylor coefficients from the definition if the function is built out of one of the four standard series. Substitution, multiplication, differentiation, and integration all pass straight through.

Say this to her once

There is no formula sheet on the AP exam. Every formula must be recalled or reconstructed. A student who can rebuild the quotient rule in four lines when it deserts her under time pressure has a real, concrete advantage over one who only memorized it — this is not an aesthetic argument, it's a points argument.

Tool 2 — Questions Worth Asking

Questions that separate real understanding from successful pattern-matching. Each is answerable in one sentence by someone who has it, and produces hand-waving from someone who doesn't.

How to use these without it becoming a quiz

One question, in passing, when she's already comfortable. Never during homework, never after a bad test, never two in a row.

The good version is asking because you actually want to know — which, for most of these, you will. "I got stuck on why radians matter and I can't work it out" is a real sentence. "Let me check whether you understand radians" is not, and she'll hear the difference instantly.

If the answer is shaky, don't correct it on the spot. Note which section of the guide covers it and find a reason to mention that thing a few days later.

Unit 1 — Limits

Give me a function where the limit at 3 is 7 but f(3) = 100.

Good: any function with a hole, explicitly redefined at that point — she should be able to invent one in ten seconds, e.g. "f(x) = x + 4 everywhere except f(3) = 100."
Shaky: "That can't happen." That's the misconception the whole unit exists to break, and it will cost her on every 0/0 problem.
Ask it when: she's just finished the limits unit and thinks it was easy.

Why does sin x / x → 1 only in radians?

Good: because the sector-area formula ½r²θ only holds in radians — in degrees you'd carry π/180 through every trig derivative forever.
Shaky: "Because that's the rule." Fine at this stage, but it means the sector-squeeze picture didn't land, and that picture is the source of every trig limit she'll need.
Ask it when: she complains that her calculator was in the wrong mode. Which she will.

Units 2–3 — Derivatives

Why isn't the product rule just f′g′?

Good: the rectangle picture — growing both sides adds two strips and a corner; the corner vanishes in the limit but the strips don't. Or just: "try f = g = x; you'd get 1 instead of 2x."
Shaky: "Because it's a different rule." She'll still get the problems right, but she has no error-check when she misremembers it under pressure.
Ask it when: she's fluent with the rules and bored.

Nothing in the power rule differentiates to 1/x. Why not, and what fills the gap?

Good: the power rule always lowers the exponent by one, so landing on −1 means starting at x⁰, which is constant with derivative 0. ln x fills it.
Ask it when: Unit 3 introduces d/dx[ln x] = 1/x. It reframes a memorized table entry as the answer to a question, and it sets up the Unit 6 payoff where ln x gets defined as an area.

Why are you allowed to treat a circle as a function when you differentiate implicitly?

Good: locally it is one — and dy/dx = −x/y blows up exactly at (±5, 0), which is where the circle is vertical. The formula announces where the trick stops working.
Ask it when: implicit differentiation is going well. This is the best "the textbook didn't tell you" moment in the fall.

Units 4–5 — Applications

Can a particle be speeding up while accelerating in the negative direction?

Good: yes, if it's also moving in the negative direction. Same sign means speeding up.
Shaky: "No, negative acceleration means slowing down." This is the most-missed motion question on the exam, every year.
Ask it when: during the motion unit. This one is worth actually resolving rather than leaving.

If f′(c) = 0, is there a maximum at c?

Good: no — x³ at 0. Critical points are candidates, not conclusions.
Ask it when: she starts curve sketching. Also a fast check on whether she's justifying or just computing.

Why is a tangent-line estimate sometimes too big and sometimes too small?

Good: concavity. Concave up puts the tangent below the curve, so it underestimates.
Ask it when: linear approximation appears. The same fact reappears for Euler's method in Unit 7, so it pays twice.

Unit 6 — Integration

What is an integral, if not the area under a curve?

Good: accumulating a product where one factor varies. Area is one picture of it.
Shaky — and this is the important one: "the antiderivative." That's a computational method, not a definition, and a student who believes it cannot make sense of work, mass, expected value, or anything in Units 7 and 8.
Ask it when: right after the Fundamental Theorem, which is exactly when the misconception forms.

Why is the Fundamental Theorem surprising?

Good: slope and area are unrelated-looking questions, and it took two thousand years to notice they're inverse.
Shaky: "Is it?" Which is a fair response to how it's usually taught — forty minutes, presented as a computational rule.
Ask it when: she's just learned it and is unimpressed.

Where does the +C come from?

Good: the Mean Value Theorem — two functions with the same derivative differ by a constant.
Shaky: "It's a rule." Extremely common. Worth fixing because it also explains why FTC Part 1 gives the same answer no matter which antiderivative you pick.

Unit 10 — Series

The terms of 1 + ½ + ⅓ + ¼ + … go to zero. Does it converge?

Good: no. And ideally: it diverges like ln n, agonizingly slowly.
Shaky: "Yes." This single question catches the most consequential misconception in the largest unit on the exam. If the answer is yes, she will misapply the nth-term test all spring.
Ask it when: series begins. This is the one question on this page I'd insist on asking.

What's the connection between the integral test and the p-series rule?

Good: they're the same threshold. Σ1/n^p and ∫dx/x^p converge together, both exactly when p > 1.
Ask it when: the convergence tests feel like an arbitrary zoo — which they will, for about three weeks. Seeing two of them as one fact makes the rest feel learnable.

Why is there an n! in the Taylor coefficients?

Good: differentiating (x−a)ⁿ exactly n times produces n!, so the coefficient has to cancel it.
Ask it when: Taylor series is introduced. It converts the formula from arbitrary to inevitable.

The one to ask all year

Highest-yield sentence in this document

"Why do you think they teach it in that order?"

It works because: it's a real question you don't know the answer to, she can actually answer it, and answering it requires understanding both topics and the relationship between them.

Examples of where it bites:

Why limits before derivatives? — Because the derivative is defined by one. But historically it went the other way: calculus worked for 190 years before limits were rigorous.
Why Riemann sums before the Fundamental Theorem? — So you know what an integral is before you learn the shortcut. Otherwise "integral" just means "antiderivative," and Units 7 and 8 stop making sense.
Why partial fractions before logistic growth? — Because solving the logistic equation requires them.
Why series last? — It needs limits, integrals, improper integrals, and derivatives of every order.

Each of those is a two-minute conversation, and each one makes the curriculum feel designed rather than arbitrary.

Tool 3 — The Pacing Calendar

Approximate, but the shape holds for most BC classes. Official College Board unit numbers in brackets, since that's what her teacher will use.

When	Topic	Watch for
Late Aug – Sep	Limits, continuity [1]	Abstract and unmotivated. She's not confused — it genuinely hasn't paid off yet.
Sep – mid Oct	Derivatives, all rules [2,3]	Taught at double speed in BC. Chain-rule fluency is the thing that matters.
Mid Oct – Nov	Applications, MVT, optimization [4,5]	First real modeling. Related rates is where word problems start biting.
Dec – Jan	Integrals, FTC, techniques [6]	Heaviest unit on the exam, 17–20%. If Riemann sums get rushed, fill that gap.
Feb	Differential equations [7]	Usually goes well. Slope fields are easier than they look.
Feb – Mar	Applications of integration [8]	Volumes with a shifted axis of rotation is the error zone.
Mar	Parametric, polar, vector [9]	Polar area limits. Sketch first, always.
Mar – Apr	Sequences and series [10]	17–18%, hardest unit, worst timing. This is the one.
Apr – early May	Review	Fatigue is the real enemy. The exam is in the first half of May.

The single most useful thing on this page

Exam weighting is inverted from how the year feels. Verified against College Board's published multiple-choice weightings:

Units 1–4 (all of limits, all of basic differentiation): 5–10% each
Units 5 and 9: 10–15% each
Units 6 and 10: 15–20% each — tied heaviest, up to two-fifths of the section together

So the fall races, at double speed, through the lowest-weighted material. The heavy units land February through April, when everyone is tired and the exam is close.

Be most available in the spring, not the fall. That's counterintuitive, because September is when a course feels new and a parent feels useful.

Format change for May 2027 onward

A student starting BC in autumn 2026 sits a revised paper. Multiple choice Part A: 29 questions in 62 minutes (was 30 in 60). Part B: 13 questions in 38 minutes (was 15 in 45). Forty-two questions rather than forty-five.

Course content is unchanged. The only practical effect: prep books printed before 2026 have the old counts, so timed-practice pacing will be slightly off.

Tool 4 — Triage: What to Do When She's Stuck

The most valuable judgment here isn't mathematical. It's working out which of five things is happening, because they need opposite responses and the wrong one makes it worse.

Symptom	Cause	Response
Sets the problem up wrong, then executes cleanly	Modeling gap — can't translate words into an equation	Draw it together. Ask what varies and what's fixed. Don't touch the calculus.
Sets it up right, gets the wrong number	Algebra slip, not conceptual	Have her check her own work backwards. Re-teaching insults her.
Right answer, can't say why	Pattern-matching — will fail on a variant	One diagnostic question, later, not now.
Doesn't know which tool applies	Selection problem, not a knowledge problem	The decision procedures: limits tree, series tree, LIATE. She has the tools.
Frustrated, avoidant, snappish	Not a math problem	Stop. Nothing you explain will land.

Four of these, as they actually look

Scenario 1 — the modeling gap

"A ladder is sliding down a wall…" She writes x² + y² = 169, then immediately substitutes x = 5 and y = 12, differentiates, and gets zero. She's frustrated because the calculus was easy and the answer is obviously wrong.

What's happening: she substituted before differentiating, which turned two variables into constants. This is the single most common related-rates failure and it is not a calculus error — it's a modeling error about which quantities are allowed to move.

What to do: ask "which of those numbers is true only at this instant?" Don't explain the fix. That one question locates it, and she'll see it herself. Then the rule sticks: letters for what varies, numbers for what's fixed, values in last.

What not to do: walk her through the corrected solution. She'll follow it, agree, and make the same error next week, because the error was never in the steps she watched.

Scenario 2 — the algebra slip

She sets up a washer-method volume perfectly — correct axis, correct radii, correct limits — and gets a negative number.

What's happening: almost certainly a sign or a squaring error inside the integral, or π∫(R−r)² instead of π∫(R²−r²). Nothing conceptual is broken.

What to do: "the setup looks right to me — a volume came out negative, so something downstream flipped a sign." Name the class of error and let her hunt. Finding her own arithmetic mistakes is a skill and it only develops if she does it.

Why this matters more than it sounds: a student who has been rescued from arithmetic all year has no error-detection habit on exam day, when nobody is available to check her work.

Scenario 3 — pattern-matching that will fail later

She's getting every u-substitution right. You ask, idly, why substitution works at all. She says "you just let u be the inside part."

What's happening: the procedure is solid and the concept is missing. She will be fine until a problem where the derivative is almost present — ∫x·e^(x²)dx, where du = 2x dx but only x dx is there — or until Unit 10, where she needs to run series operations in reverse.

What to do: nothing right now. She's succeeding; interrupting that is pure cost. Note it, and a few days later mention the actual answer as a thing you found interesting: substitution is the chain rule read backwards, so what you're really hunting for is a function and its own derivative sitting in the same integrand.

Timing rule: concept lands after fluency, not before, and not during.

Scenario 4 — the selection problem

Series unit. She's staring at Σ n²/(n⁴+3) and has tried three tests without getting anywhere. She knows all the tests.

What's happening: this is not a gap in knowledge, it's a gap in ordering. She's picking tests at random instead of running a procedure.

What to do: hand her the decision tree, don't narrate it. "Does aₙ → 0? Yes, so nothing's ruled out. Any factorial? No, so skip the ratio test. Rational-looking? Yes — so limit-compare, keeping only the dominant powers: n²/n⁴ is 1/n², a convergent p-series."

Why this is the most fixable category: selection problems look like knowledge problems and feel like stupidity to the student. They're neither. Ten minutes with a decision procedure resolves what an hour of re-teaching won't.

Scenario 5 — and the honest one

It's 10:40pm, there are two problems left, and she's snapping at you.

What's happening: not calculus.

What to do: stop. Say the problem set will be there tomorrow, or that a wrong answer on one homework question costs nothing. Nothing you explain at this point will be retained, and the association between you and the subject is worth more than two problems.

The failure mode is you, and it's worth naming plainly

Read this one twice

You will want to give her the beautiful structural picture. What she needs on a Tuesday night is to get the derivative of x·sin(x²) right on a timed test.

Conceptual depth doesn't substitute for procedural fluency; it follows it, usually by months. Parents who lead with "but do you see why" often read as (a) not actually helping and (b) mildly disappointed.

The extensions are the right move for a bored kid and the wrong move for a struggling one. Diagnose before you deploy.

Rules of engagement

Follow her lead; offer one thing. The "one thing to show her" items in the guide are one per unit deliberately. That's a ceiling, not a target.
Never deploy an entice item when she's behind. A struggling student hearing "isn't it beautiful that—" experiences it as a change of subject away from her actual problem.
Don't front-run her teacher. L'Hôpital in September, or the difference-quotient shortcut before she's drilled the algebra, costs her points and skips practice the unit exists to build. The guide flags these in purple.
Frame as interest, not assistance. "I got curious about why they teach limits first and the answer is kind of a mess" lands completely differently from "let me help you with limits."
She may not want to be taught by a parent at all. That's common, it isn't about you, and the offer itself can be the problem. The history-and-context layer survives that — it's a thing you can mention at dinner without it being a tutoring session.

What you uniquely have

The actual value proposition

She's a sophomore in BC who likes math and was taught well. She does not need you to teach her calculus.

What you can offer that her course structurally cannot: why the ideas are ordered this way, what the historical fights were, what's still unresolved, where it goes next, and what other countries do differently.

That's a different thing from help. It's also what makes a subject feel like a place rather than a hurdle — and it's the part a good teacher would love to give her but has 150 students and a May deadline.

Appendix — Three Conversations Worth Having

1 · The foundations were broken for 150 years

Newton and Leibniz had working calculus by 1670. Berkeley demolished its logic in 1734 — infinitesimals had to be both zero and not-zero, "the ghosts of departed quantities," and he was right. Nobody could answer him. Weierstrass's limits (1860s) dodged the problem rather than solving it. Robinson finally vindicated Leibniz in 1966 with nonstandard analysis.

Why it's useful: when she treats dy/dx as a fraction and cancels — which separation of variables basically instructs her to do — the honest response isn't "don't." It's "that's Leibniz's instinct, it isn't justified in the framework you're being taught, and it took 300 years to make rigorous." Much better conversation than a correction.

2 · Notation as competitive advantage

Newton used dots. Leibniz used dy/dx and ∫. The priority war made Britain loyal to Newton's notation for a century, and British mathematics fell behind the continent — because Leibniz's notation suggests the correct manipulations and Newton's doesn't.

Why it's useful: it's the best available argument that interface design is not cosmetic. Applicable well outside mathematics.

3 · Differentiation is an algorithm; integration isn't

Every elementary function has an elementary derivative — turn the crank, always works. Most have no elementary antiderivative, and Liouville proved it in the 1830s. Not "nobody found one." There isn't one.

And yet ∫e^−x² over the whole real line is exactly √π. The most important function in statistics can't be integrated by any technique in her course, but its total area is clean.

Why it's useful: it tells her that Unit 7 being a bag of tricks isn't a failure of her textbook or of her. It's a fact about mathematics. Students who don't know this assume they're missing an insight.