Where it breaks, and why
A thread reads badly when the cuts land mid-thought, so the splitter works down a ladder: it keeps whole paragraphs together when they fit, then packs whole sentences, and descends to word boundaries only when a single sentence outruns the limit. Words are never cut in half (the lone exception: a single token longer than an entire post — usually a monster URL — which gets flagged).
Two controls put the author above the algorithm. A line containing only --- forces a break exactly there — the move for punchlines that must end a post and section turns the math can’t see. And the output isn’t frozen: each post is an editable box with a live count, so the last 10% of craft happens in place instead of in a second round-trip through your notes app. Numbering, when you use it, reserves its space before packing — a post can never go over the limit because “7/9” got stapled on.
The character math each platform actually uses
Most “over the limit” surprises aren’t about length — they’re about counting rules. X weighs characters (most Latin script = 1, emoji and CJK = 2) and flattens every URL to 23 characters via its link wrapper, rules documented in X’s developer documentation. Mastodon borrows the 23-character URL rule; Bluesky counts graphemes, so a four-code-point family emoji is one character, not four. The splitter measures with each platform’s own rules:
| Platform | Limit per post | How it counts |
|---|---|---|
| X (Twitter) | 280 | Weighted: most Latin characters count 1, emoji and CJK count 2, every URL counts 23. |
| Threads | 500 | Plain count, 500 per post. URLs count at their full length. |
| Bluesky | 300 | 300 graphemes — a flag or family emoji counts once, not per code point. |
| Mastodon | 500 | 500 on the default instance; every URL counts as 23 regardless of length. |
A thread is an edit, not an export
Splitting solves the arithmetic; it doesn’t make the thread good. The difference between a thread that travels and one that dies at post two is almost always editorial:
- Rewrite the hook by hand.Post one’s only job is to earn the next tap. The first sentence of a long draft is rarely the strongest opener — promote a number, a tension, or the payoff.
- One idea per post.After splitting, read each box and ask what it’s about. If the answer is “two things,” move a sentence — that’s exactly what the editable boxes are for.
- End posts on tension, not closure. A post that resolves invites leaving; a post that opens a loop invites the next one. The
---marker exists so you can place those cliffhangers deliberately. - Spend the last post. Recap in one line, then one ask — follow, reply, or the link. One ask, not three.
Frequently asked questions
Why does one emoji eat two characters on X?
X doesn't count characters the way a text editor does — it counts weight. Most Latin-script characters, digits, and punctuation weigh 1, but emoji, CJK characters, and most other symbols weigh 2, and every URL costs a flat 23 no matter how long it really is. That's why a 270-character draft in your notes app can refuse to post. This splitter applies the same weighting X's own API enforces, so the count you see here is the count that matters.
Can I control where the thread breaks?
Two ways. Before splitting: put --- on its own line anywhere in the draft and the splitter treats it as a mandatory break — useful when a punchline must end a post. After splitting: every post is an editable text box with a live count, so you can move a sentence between posts, tighten a line that's two characters over, or rewrite a weak transition. Changing the draft or settings re-splits from scratch and replaces those edits.
Should the numbering go at the start or the end of each post?
Start-numbering (1/7 …) tells readers immediately that they're inside a thread and how much is coming — best for longer, structured threads. End-numbering keeps the first words of every post clean for the feed preview, which matters most on the hook post. There's no engagement penalty either way; pick one and keep it consistent. Skip numbering entirely for two-post threads — the reply chain is obvious on its own.
How long should a thread be?
As long as the idea needs and no longer — in practice most ideas exhaust themselves between four and ten posts. Attention decays monotonically through a thread, so put the strongest material in the first three posts rather than building to a finale most readers won't reach. If the draft splits into twenty posts, that's usually two threads or an article wearing the wrong format.
What makes a strong first post?
The first post does one job: earn the tap on "show more." State the payoff or the tension in concrete terms — a number, a contrarian claim, a before/after — and cut every word of preamble. The splitter labels post one "the hook" because it deserves different treatment: after splitting, rewrite it by hand rather than accepting whatever sentence happened to come first in the draft.
Do hashtags and mentions belong in every post of a thread?
No. If you use hashtags at all, one or two on the first or last post is plenty — repeating them in every post reads as spam to both readers and ranking systems. Mentions belong only where the person is genuinely relevant, and remember that on X, starting a post with a mention turns it into a reply with limited reach; lead with a word, not an @.
Is a thread better than one long post?
They do different jobs. A thread creates a re-entry point at every post — each one can be quoted, bookmarked, and surfaced on its own, and the format forces the skimmable rhythm feeds reward. A single long post keeps an argument intact and suits readers who hate fragmentation. If the idea has natural beats, thread it; if it's one continuous argument, post it long where the platform allows and link it elsewhere.