Open with the claim, not the background. Readers who land here from a link have no context and little patience, so the first paragraph should state what you are arguing and why it would matter if true. Save the history of the problem for section two — the people who need it will still be reading by then.
Keep paragraphs to a single idea each. This page is set to roughly sixty-five characters per line, which is the width at which sustained reading is easiest; if you widen it, the eye starts losing its place on the return sweep.
§ 1The problem
What is currently unexplained, and what the existing accounts get wrong. Be specific about what you are disagreeing with. A theory that argues against a vague consensus is hard to evaluate and easy to dismiss.
§ 2The proposal
The core of it. If your argument has a formal component, inline math looks like \(f(x) = \alpha x + \beta\) and sits inside a sentence without disturbing the line spacing. Displayed equations get their own line:
$$ \Phi(t) = \int_{0}^{t} \kappa(\tau)\, e^{-\lambda(t - \tau)} \, d\tau $$
Both are real text, so they are searchable, selectable, and readable by screen readers. Never paste equations as screenshots.
A sub-point
Use a third-level heading when a section has parts worth navigating to separately. If you never need one, delete the rule for it.
§ 3What would falsify it
This section does more for your credibility than any other. State what observation or argument would count against the theory. If nothing could, say so plainly and explain why — that is a substantive position, not an evasion.
Quotations from other work go here, attributed. Keep them short; paraphrase where you can.
§ 4Open questions
What you have not worked out. Listing these invites the right kind of correspondence and makes the draft status honest rather than apologetic.