HN Debrief

Simplifying and Refactoring Introductory Calculus (2018)

  • Education
  • Mathematics
  • Programming
  • AI

The paper proposes a rewrite of first-year calculus. It treats differentials like algebraic objects students can manipulate directly, pushes formal limits to the end of the course, and argues that many standard rules for derivatives are really the same operation in disguise. The author’s concrete complaint is that introductory courses force students to memorize several separate procedures, hide what symbols like dy and dx are supposed to mean, and introduce second-derivative notation that breaks the algebraic story the rest of the course is trying to tell. In the comments, the author said this came out of teaching small homeschool calculus classes and later writing a textbook, not from a large institutional rollout. He also said he avoids forcing his alternative higher-derivative notation in the main text because students still have to survive conventional later courses.

If you care about technical education, the useful takeaway is not the paper’s specific notation change but its sequencing claim: lead with operational intuition, then add rigor once students have something to anchor it to. For teams hiring or training quantitative talent, poor calculus teaching still shows up later as weak comfort with modeling, rates of change, and approximation.

Discussion mood

Mostly positive about the paper’s critique of standard calculus teaching, but skeptical that a notation overhaul alone is the answer. The mood was constructive and opinionated, with recurring frustration at epsilon-delta-first pedagogy, overloaded curricula, and the gap between symbolic competence and actual understanding.

Key insights

  1. 01

    Trim the course before changing foundations

    A more useful fix is to narrow what counts as introductory calculus. The point is that students do not need every theorem, sequence argument, multivariable extension, and ugly symbolic technique in the same gateway course. Treat continuity and the common limit rules as working assumptions, get to derivatives sooner, and move laborious methods into later specialist classes. That framing shifts the problem from notation design to curriculum bloat.

    If you influence curriculum, cut scope before adding new formalism. A leaner first course will likely improve retention more than a clever redefinition of dy and dx.

      Attribution:
    • scythe #1
  2. 02

    Infinitesimals now have rigorous options

    The old objection that dx is mystical is historically dated. Several commenters noted that Leibniz-style reasoning can be formalized through nonstandard analysis and hyperreals, which let dy and dx behave more like actual quantities instead of limit shorthand. That does not settle the teaching question, because hyperreals give up the Archimedean property and introduce their own machinery, but it undercuts the claim that infinitesimal approaches are inherently sloppy.

    Do not dismiss infinitesimal-first teaching as mathematically unserious. If you are evaluating educational material, separate the rigor question from the pedagogy question.

      Attribution:
    • simonreiff #1 #2
    • u1hcw9nx #1
    • gloveone #1
  3. 03

    Concept without drill does not stick

    Visual or intuitive presentations help students get started, but they do not replace execution practice. The sharper claim here is that real understanding only appears once a student can reliably work problems, spot patterns, and recover methods from memory. Without exercises and feedback, even elegant explanations stay at the level of recognition rather than fluency.

    When choosing learning resources, prefer ones with dense problem sets and worked solutions, not just polished exposition. For internal training, pair conceptual talks with repeated hands-on exercises.

      Attribution:
    • BobaFloutist #1
    • tgv #1
  4. 04

    Delay rigor because students forget it anyway

    The author argued from teaching experience and cited papers claiming that students retain little of epsilon-delta reasoning when it is front-loaded. The practical case for postponing formal limits is that rigor lands better once students already know what derivatives and integrals are for. In that view, early handwaving is not a flaw but a scaffold, as long as it is clearly labeled as temporary.

    For onboarding difficult technical material, put formal proof after users have built operational intuition. People remember abstractions better when they can attach them to a tool they already use.

      Attribution:
    • johnnyb_61820 #1 #2
  5. 05

    Discrete sequences and zoomable graphs compete as intuition

    Two concrete teaching metaphors emerged for limits. One builds from sequences and convergence, using shrinking finite differences to reach derivatives. The other starts from continuity and the modern instinct that you can keep zooming into a smooth graph and see the same local behavior. The interesting point is that both sides are trying to replace formal epsilon-delta syntax with a more native mental model, not to defend rigor for its own sake.

    If you teach technical ideas, choose the metaphor that matches what your audience already manipulates comfortably. For today’s students, interactive visual intuition may sometimes beat theorem-first abstraction.

      Attribution:
    • rramadass #1
    • cyberax #1
  6. 06

    Integration remains the unsolved pedagogy mess

    Even commenters sympathetic to reform said the paper attacks the easier half of calculus. Differentiation can be unified cleanly through algebraic rules, but symbolic integration still depends on a grab bag of substitutions, identities, and pattern recognition. Rule-based integration systems were offered as a way to expose that structure, but nobody had a convincing simplification on the level the paper claims for derivatives.

    If you want to improve calculus education or tooling, integration is the higher-leverage target. There is still room for better systems that explain why an integration trick applies, not just which answer to write.

      Attribution:
    • ogogmad #1 #2
    • drunkboxer #1
    • srean #1

Against the grain

  1. 01

    Stewart already works for many students

    The push to refactor calculus can be overstated. One blunt response was that James Stewart’s standard textbook is already excellent, which reflects a broader truth that many students do learn perfectly well from conventional materials. That reminder keeps the conversation honest. A teaching method can be widespread because it is imperfect and still good enough for a large population.

    Do not assume your own bad calculus experience proves the entire mainstream approach is broken. When changing teaching materials, validate against actual outcomes rather than novelty appeal.

      Attribution:
    • richard_chase #1
  2. 02

    Deferring theorems can hollow out math

    Waiting a long time to connect intuition to proof risks teaching a bag of tricks instead of mathematics. The point here is not nostalgia for rigor. It is that proof is part of how students learn to reason about systems, and if formal justification arrives too late, many never integrate it with the computational side at all.

    If you sequence intuition before rigor, put a clear deadline on when the formal story arrives. Otherwise students may leave with procedures they can execute but cannot justify or extend.

      Attribution:
    • xiphias2 #1
  3. 03

    Smooth infinitesimals hide edge cases

    Smooth infinitesimal analysis was praised as elegant, but the strongest pushback was that it simplifies calculus partly by ruling away rough functions and adopting nonclassical logic. That can make the subject feel cleaner while making it harder to tell when an assumption is illegitimate. Another commenter argued that a mix of nonstandard analysis with big-O and little-o notation captures the same benefits more transparently.

    Be wary of pedagogies that become simple by baking in stronger assumptions than the later subject allows. If you use them, mark clearly where the simplification stops being safe.

      Attribution:
    • yorwba #1
    • ogogmad #1

In plain english

Archimedean property
A property of the real numbers stating that no number is larger than every whole number, and no positive number is smaller than every reciprocal of a whole number.
dual numbers
A number system like complex numbers but with a nilpotent element whose square is zero, useful for derivatives and automatic differentiation.
DX
Developer experience, the overall ease and quality of using tools, codebases, and workflows as a developer.
dy
A differential symbol traditionally used to represent an infinitesimally small change in the variable y.
epsilon-delta
The standard formal definition of limits in calculus, using two small positive numbers to state precisely how outputs must behave when inputs get close to a point.
hyperreals
An enlarged number system used in nonstandard analysis that includes infinitesimally small and infinitely large numbers.
intuitionistic logic
A system of logic that does not accept some standard classical proof principles, especially unrestricted use of proof by contradiction.
little-o notation
A notation for describing an error term that becomes negligible compared with another quantity near a limit.
nonstandard analysis
A rigorous branch of mathematics that extends the real numbers with infinitesimal and infinitely large numbers.
Smooth infinitesimal analysis
An approach to calculus that uses infinitesimals inside a different logical framework where many functions are treated as smoothly differentiable.

Reference links

Author follow-up and related papers

Alternative calculus texts and foundations

Teaching resources and visual explainers

Research and technical references