You've got a stack of blank notebooks. Which means a favorite pen. Maybe even a fresh LaTeX template open in a browser tab. And you're staring at the first page thinking — okay, now what?
Starting a math journal sounds simple. But the difference between a notebook that gathers dust and one that actually changes how you think? It is simple. That comes down to a few habits most people never talk about.
What Is a Math Journal, Really
It's not a diary. It's not a scrapbook. And it's definitely not just "showing your work" for a grade.
A math journal is a thinking tool. A place where you externalize the messy, non-linear process of understanding something mathematical. Proofs that don't quite work. Definitions you're trying to internalize. So questions that wake you up at 2 a. Practically speaking, m. Connections between topics that only make sense after you've written them down three different ways.
Some people keep one journal for everything. Others separate by course, by research project, by phase of life. There's no correct structure — only the structure you'll actually use.
The Two Main Flavors
Learning journals live alongside a course or textbook. You're working through Rudin, or taking a topology class, or self-studying category theory. The journal captures your struggle with the material — not just the clean final answers.
Research journals are different. They're where ideas go to grow. Conjectures. Failed approaches. Literature notes. Conversations with advisors. The half-formed thought that might become a paper six months from now.
Both serve the same purpose: they make your thinking visible to your future self.
Why Bother When You Have Notes Already
Here's what most students miss: class notes record what someone else said. A math journal records what you thought Most people skip this — try not to..
That distinction matters more than it sounds.
When you copy a proof from the board, you're transcribing. When you rewrite that proof in your own notation, questioning each step, asking "why this hypothesis?" — you're learning. Practically speaking, " or "what breaks if I drop this condition? The journal is where that second process lives Took long enough..
It sounds simple, but the gap is usually here.
Research backs this up. Day to day, writing-to-learn studies in STEM consistently show that students who regularly explain concepts in their own words retain more and transfer better. But you don't need a study to know this. You've felt the difference between reading a solution and reconstructing it Worth keeping that in mind..
No fluff here — just what actually works.
What Goes Wrong Without One
You solve a problem. Which means two weeks later, the same technique appears in a different context — and you don't recognize it. Feels good. Because you never articulated why it worked the first time Still holds up..
Or you read a paper. Nod along. Plus, close the PDF. Three days later, the key lemma is gone. You didn't write down the one sentence that would have anchored it.
A journal prevents that. Not perfectly. But enough to matter.
How to Start One That Doesn't Die in Week Two
Pick a Medium You'll Actually Use
Paper has advantages. No notifications. Muscle memory for symbols. The physical act of writing slows you down just enough to think. Many mathematicians swear by it — Terence Tao, for one, has spoken about his paper notebooks.
But digital works too. Obsidian, Notion, a simple folder of .LaTeX if you want beautiful output. txt files. The best medium is the one that adds the least friction between "I have a thought" and "it's recorded But it adds up..
I use both. Paper for deep work sessions. Digital for searchable reference later. Do what fits your actual workflow, not your aspirational one.
Create a Simple Entry Template
Blank page paralysis is real. A lightweight structure kills it.
My entries usually have:
- Date and context — what I'm working on, where I'm sitting, how long I have
- Goal — one sentence: "Understand the proof of Urysohn's lemma" or "Debug why my induction fails"
- Working — the messy middle. Day to day, calculations. And diagrams. Half-sentences. Now, crossed-out lines. - Insight or question — the takeaway. That's why what clicked? What's still fuzzy? Which means what do I need to ask someone? - Next step — concrete. "Re-read Munkres 33.2" or "Email Prof.
Some disagree here. Fair enough That's the whole idea..
That's it. Five minutes to set up. Then you just fill it in.
Write Like You're Explaining to a Slightly Less Prepared Version of Yourself
Not your professor. Not a journal reviewer. *You, six months from now, after you've forgotten the details Worth knowing..
That means:
- Define your notation every time
- State the theorem you're using, don't just cite "Theorem 3.2"
- Explain why you chose this approach
- Flag where you're uncertain — "I think this works because..." is honest and useful
Future you will thank present you. I promise Nothing fancy..
What to Actually Put In It
Definitions in Your Own Words
Textbook definitions are optimized for precision. Your definitions should be optimized for you.
Take the epsilon-delta definition of a limit. Think about it: the book gives you the formal statement. Your journal gets the picture: "No matter how tight a horizontal band I draw around L, I can find a vertical band around a that keeps the graph inside the horizontal one — except possibly at a itself.
Draw the picture. Use colors. Practically speaking, write the informal version and the formal one. Connect them Easy to understand, harder to ignore..
Failed Attempts — Especially the Failed Ones
This is the part everyone skips. Don't.
When a proof falls apart, write down where and why. "Tried to use uniform continuity here, but the domain isn't compact — that's the gap.Also, " That failed attempt is now a signpost. Next time you see a similar structure, you'll recognize the trap.
Failed attempts are also where new questions hide. Think about it: "What would make this work? " That question might be more valuable than the original problem Worth keeping that in mind. Surprisingly effective..
Connections You Notice
"Wait — this is exactly the same trick as in the proof of the Hahn-Banach theorem, just in a metric space setting."
Write that down. Also, explicitly. "Same trick as Hahn-Banach: extend linearly on a subspace, use Zorn's lemma to get a maximal extension.
These cross-connections are how expertise builds. On top of that, they don't happen automatically. They happen when you force yourself to articulate them.
Questions You Can't Answer Yet
"Why does this theorem need the Hausdorff condition?" "What's a concrete example where this fails without separability?" "Is the converse true?
Keep a running list. Some become research problems. Some you'll answer next week. Some you'll ask a colleague. The list itself is valuable — it maps the boundary of your current understanding.
Common Mistakes That Kill Journals
Treating It Like a Performance
If you're writing for
Common Mistakes That Kill Journals (Continued)
Treating It Like a Performance
(Continued from where it cut off)
If you're writing for an imaginary audience—trying to sound clever, hiding confusion, or polishing every sentence until it gleams like a published proof—you've missed the point. This journal isn't a performance; it's a workshop. Future you doesn't need flawless prose; they need to see where the thinking got stuck. Did you assume a space was compact without verifying it? Write: "Assumed X is compact here because it’s closed and bounded—but wait, that only works in R^n! In infinite dimensions, closed+bounded ≠ compact (e.g., unit ball in l^2). Need to check if we’re in finite dimensions or use another argument." That raw, uncertain note is gold. It’s the exact thing future you will curse past you for omitting when they stare at the same gap six months later. Perfectionism here is procrastination in disguise. Scribble messy thoughts. Circle doubts. Let it look like thinking, not a lecture.
Skipping the "Why This Step?"
It’s tempting to just write the next line of a proof: "So, by Urysohn’s lemma, there exists a continuous function f..." But why Urysohn? Why not Tietze? What property of the space are we actually using? If you don’t pause to explain why you chose that tool, future you will have to reverse-engineer your reasoning from scratch. Instead, try: "We need a function separating point p from closed set C. Urysohn’s lemma applies here because X is normal (which it is, as a metric space), and it gives us f(p)=0, f(C)=1 directly—Tietze would require extending a function defined on C∪{p}, which is more work. Normality is the key; we used metric → normal earlier (Theorem 2.1)." This isn’t redundancy—it’s building a trail of breadcrumbs for your future self through the forest of alternatives.
Ignoring the Boundary Cases
Compactness arguments often hinge on edge cases: what happens if the cover has no finite subcover? What if the sequence has no convergent subsequence? Journals die when you only record the "success" path. Suppose you’re proving a continuous function on a compact set attains its maximum. Don’t just write: "By extreme value theorem, max/min exist." Instead, note the near-misses: "Tried to use sequential compactness: take sequence x_n where f(x_n) → sup f. Need convergent subsequence—got it by compactness. But what if X wasn’t compact? Example: f(x)=1/x on (0,1] is continuous, bounded below but no minimum (inf=0 not attained). The sequence x_n=1/n has no limit in (0,1], so sup f isn’t attained. Compactness prevents this 'escape to the boundary'—here, the boundary {0} is missing." Recording these near-failures trains you to spot where compactness is actually doing work, not just where it’s cited.
Forgetting to Reset Notation Mid-Thread
You start a page discussing compactness in metric spaces, using (X,d) for the space. Halfway down, you switch to talking about function spaces and reuse (X,d
Forgetting to Reset Notation Mid-Thread
You start a page discussing compactness in metric spaces, using $(X,d)$ for the space. Halfway down, you switch to talking about function spaces and reuse $(X,d)$ for the domain of functions, while introducing $Y$ for the codomain—but now $d$ refers to a different metric, or worse, the same symbol means two different things. Future you will waste precious time untangling whether that $d(x,y)$ refers to distance in the original space or in the function space. Instead, build notation resets into your workflow: "Switching context: now working in $C(X,Y)$ where $X=(X,d_X)$ is our original metric space and $Y=(Y,d_Y)$ is the target space. The metric on $C(X,Y)$ will be $d_\infty(f,g) = \sup_{x\in X} d_Y(f(x),g(x))$—note this requires $Y$ to be complete and $f,g$ to be bounded." Treat notation like variable scope in programming—be explicit about where each symbol lives and what it means in that context.
Hand-Waving the "Obvious" Isomorphisms
"Clearly, $L^2([0,1])$ is isomorphic to $\ell^2$." No—it's not obvious, and the isomorphism depends on choosing a basis, which requires the axiom of choice for infinite dimensions. When you skip the construction, you lose the computational tools that come with explicit representations. Instead: "Choosing the orthonormal basis ${e_n(x) = e^{2\pi i n x}}$ for $L^2([0,1])$, we get the isomorphism $f \mapsto {\hat{f}(n)}_{n\in\mathbb{Z}}$ where $\hat{f}(n) = \int_0^1 f(x)e^{-2\pi i n x}dx$. This maps $L^2([0,1]) \to \ell^2(\mathbb{Z})$. Note: this is unitary, so it preserves inner products—crucial for our later Hilbert space arguments." The "obvious" steps are often where the real mathematical content hides Not complicated — just consistent. Nothing fancy..
Not Recording Failed Intuitions
Your first instinct might be wrong, and that's valuable data. "Intuition: if $f_n \to f$ pointwise and each $f_n$ is continuous, then $f$ should be continuous. Counterexample: $f_n(x) = x^n$ on $[0,1]$ converges pointwise to $f(x) = 0$ for $x<1$, $f(1)=1$, which is discontinuous. So pointwise convergence doesn't preserve continuity—need uniform convergence instead. This explains why we require uniform convergence in the Arzelà-Ascoli theorem." Failed intuitions are signposts pointing toward deeper understanding.
The Meta-Habit: Write to Your Future Self
The common thread through all these gaps isn't sloppiness—it's forgetting that mathematical writing is fundamentally about communication across time. Still, every proof you write is a message in a bottle, tossed into the future ocean of your own understanding. The question isn't whether you'll need to revisit this material, but when, and how much harder you'll make it on yourself if you don't leave a clear path back.
Start treating your notes like archaeological sites. Was that compactness argument necessary, or could it be weakened? Even so, why did you choose that particular norm? Even so, future you isn't just reading your conclusions—they're excavating your thought process, looking for the layers of reasoning that led to each insight. What happens if you remove assumption X?
The next time you're tempted to clean up those messy margin notes or delete the "redundant" explanation of why you chose one theorem over another, remember: the goal isn't to produce publishable prose. It's to create a trail of intellectual breadcrumbs so complete that six months from now, you can pick up exactly where you left off, doubts and all. Your future self will thank you for the mess.
