How To Make Math

Building Mathematical Content From Scratch

I spend most of my time converting hand-written derivations into clean, publishable documents. The workflow I use isn't fancy. It involves LaTeX, a few helper packages, and a habit of checking every fraction before I commit it to the source file. The first thing people get wrong is assuming you need special software to make math look decent. You don't. A plain text editor and a compiler are enough. Everything else is optional overhead that adds cost without proportional benefit.

How To Make Math Look Right Without Wasting Time

Start by choosing your markup system. LaTeX remains the default in technical fields because it handles spacing, alignment, and symbol rendering better than anything else built for casual use. MathML exists. Word's equation editor exists. Neither will save you when a reviewer asks you to adjust a subscript placement in twelve different equations across a forty-page document. I learned this the hard way during a project where the journal rejected a submitted PDF because their automated typesetter couldn't parse nested fractions inside a displayed equation. It took me three hours to find the offending line and rewrite it using separate \frac blocks instead of stacking them. Now I avoid nesting more than one level deep unless the journal specifically asks for compact notation. Here is the practical setup I recommend. Install TeX Live if you are on Linux or Mac. On Windows, MiKTeX works fine if you keep it updated. Download an editor like TeXstudio or VS Code with the LaTeX Workshop extension. The editor does not change your output quality; it only changes how fast you notice errors.

The core document class you should reach for is amsart or revtex depending on whether you are writing a standalone paper or targeting a specific society journal. For personal notes and internal docs, book or article classes are sufficient. Do not waste time customizing margins on your first draft. The template you pull from arXiv or your target journal already has the right dimensions. When writing the actual equations, stay in math mode. Type $E = mc^2$ for inline, and \[ ... \] or the equation environment for displayed formulas. The align environment is your main tool for multi-line derivations. Use \label and \ref for cross-referencing equation numbers rather than typing them by hand. Numbers change when you edit. Labels do not. A common mistake beginners make is treating LaTeX like a word processor and manually adjusting spacing with \, and \;. You almost never need to. The math engine has built-in spacing rules that produce correct results ninety-five percent of the time. Manual adjustments become visible only when someone compares your document side by side with typeset professional work.

For symbols that are not in the default font set, load the amssymb package. It gives you everything from \mathbb for blackboard-bold numbers to the proper calligraphic letters. If you need something rarer, like a specific operator from a particular subfield, package that supports it usually exists. Searching CTAN by keyword is faster than asking someone online. I encountered an edge case once involving a custom commutative diagram that broke under a newer version of TikZ after a routine system update. The diagram had been working for two years. The fix was to pin the TikZ package version in the document preamble and rebuild from source rather than relying on the distribution's automatic updates. I stopped trusting auto-updates for anything related to publication workflows after that. If you want to generate math content programmatically instead of typing every symbol by hand, Python with SymPy or SageMath can output LaTeX directly. The .latex() method on a SymPy expression produces clean source you can paste straight into your document. This cuts down the time on long derivations significantly. A twenty-line symbolic computation that would take an hour to transcribe manually comes out in about ten minutes including verification time.

There is a tradeoff here. Programmatic generation introduces a layer between your thinking and the final output. You can end up pasting something that looks correct but is structurally wrong because the symbolic engine simplified an assumption you did not intend. Always render the PDF and inspect the output before including it. A two-minute check saves three hours of debugging later. For collaborative work, keep everything in version control. A simple git repo with your .tex files and a Makefile that runs pdflatex or xelatex on commit prevents the kind of disaster where three people edit the same draft and nobody knows which version is current. I have seen entire projects lose a week of work because someone overwrote a shared directory without checking modifications first. The compilation chain matters more than people admit. A single run of pdflatex is rarely enough. Run it twice, check the .log file for undefined reference warnings, then run bibtex if you have citations, then run pdflatex two more times. The log file tells you exactly where things broke. Most errors are not fatal; they are just unresolved cross-references that need another pass to resolve.

When sharing your work, export to PDF. Do not share source files with people who are not familiar with the toolchain. They will complain about missing packages and ask you to convert everything to Word. Refuse politely and send the PDF instead. Word can open PDFs but it does not preserve equation formatting. You lose more than you gain. Performance-wise, a typical forty-page math document compiles in under thirty seconds on a modern machine. Documents over one hundred pages with hundreds of equations and complex bibliographies can take several minutes. If your compile time exceeds five minutes consistently, check for unused packages, large embedded graphics, or recursive cross-references that force unnecessary rebuilds. Cleaning those up usually drops compile time by half or more. The main limitation of this approach is that it has a steep entry curve. You will struggle for the first few weeks. Equations will not align. References will break. The error messages look like gibberish. This is normal. Working through the errors builds familiarity faster than any tutorial can teach you. I stopped looking up basic syntax after about two months of regular use. Before that, I spent more time reading documentation than actually writing content.

If LaTeX feels too heavy for your needs, Overleaf is a browser-based alternative that handles compilation on their servers and provides templates for most major journals. It removes the local installation headache but introduces dependency on their platform. Projects move slower when you cannot work offline or need to install custom packages that the platform does not support. The bottom line is that making math look right is mostly about discipline, not talent. Type consistently. Check your references before you submit. Keep your packages updated but pinned at known-good versions for active projects. The tools do most of the heavy lifting once you stop fighting them.