Threads have a reputation problem, and it is deserved. Most of them are one idea stretched across eight posts because threads get more engagement than single posts.
Readers know. "A thread 🧵" as an opening line is now a reason to scroll past.
But the format is genuinely right sometimes, and it is worth knowing when.
When a thread works
The story has a sequence. Something happened, then something else, and the order matters. A debugging story, a launch, an experiment that changed direction.
The parts build. Each post depends on the one before. If they could be reordered without loss, it is a list, not a thread.
One post genuinely cannot hold it. Not "would be tight" but cannot. If it fits in 280 characters, it should be 280 characters.
When it does not
You have one idea and want more surface area. This is the common case and readers detect it immediately.
It is a list. Five tips are five posts or one post. Numbering them and calling it a thread adds nothing.
The first post is a promise rather than content. "Here's how I built X. A thread." You have spent your best real estate announcing.
The first post has to stand alone
The most reliable test: if someone reads only the first post and never continues, do they get something?
If yes, the thread is built correctly and the rest is bonus. If no, you have written a headline and are asking people to click.
This also improves the thread, because a first post that has to work alone is a better first post.
Length
Most stories that warrant a thread need three to five posts. Not seven, not the ten that the format seems to invite.
A rough shape when there is one:
- Hook and the implicit promise
- The problem or context
- What was tried
- What failed or changed
- The insight, and what it means
Fewer is usually better. If you are padding to reach a number, the thread ended two posts ago.
Things to drop
Numbering. The platform threads them. "1/" and "2/7" is noise, and committing to a count in advance means padding to hit it.
The thread emoji. It signals a format rather than a subject.
"Follow for more." The content earns the follow or it does not.
A summary post at the end. If the thread needs summarising, it was too long.
How we decide
Noomachy DevTracker generates threads only when the story warrants one, and the generator can return nothing.
The schema has a justified boolean. The model is explicitly told that forcing a thread onto a small story is worse than not threading it, and that a story fitting in one post should return one part with justified false.
When it comes back false, no thread is produced. The user gets a LinkedIn post and an X post and nothing else.
That was a deliberate product decision. It would be easy to always produce a thread, and every day would have more output. But a tool that reliably produces mediocre threads teaches you to ignore its threads, which costs more than the missing content.