Results Writeup

Write the strongest honest version of the result. Calibrated writing wins more arguments with reviewers than confident writing.

The core problem this skill solves

Two failure modes dominate scientific writeups, and good prose has to avoid both:

1. Overclaiming. Treating a correlation as a mechanism, a single dataset as a general law, p < 0.05 as truth, or an absence of effect as proof of no effect. Reviewers catch this and it sinks papers.
2. Underclaiming / mushiness. Hedging everything into uselessness ("results may suggest a potential trend"), burying the actual finding under qualifications, or refusing to commit to any interpretation. This wastes the reader's time and signals the authors don't believe their own work.

The target is in between: state precisely what was found, precisely what it means, and precisely where it stops — in that order. If you find yourself reaching for "may" or "potentially" twice in one sentence, the claim is either wrong or you haven't figured out what it actually is.

Workflow

1. Gather findings with provenance

Before writing anything, list every quantitative claim the writeup will make and note where each comes from — a specific table, a logged experiment, a notebook cell, a script output. If a number doesn't have a traceable source, it doesn't go in the writeup. This is the single most important step; everything else is downstream of it.