LinkedIn allows 3,000 characters. Almost nothing should use them.
Working ranges
60 to 120 words. A single observation, a small shipping update, one sharp point. Underused and very effective.
120 to 220 words. The standard range. Room for a hook, the thing that happened, and why it mattered. Most posts live here.
220 to 350 words. A genuine technical breakdown where the detail is the value. Justified when someone is learning something they could not get in a paragraph.
Beyond 350. Rarely justified. Usually one idea with excessive setup, or three ideas that should be three posts.
The real question
Length is a symptom. The question is whether any sentence is doing no work.
Common passengers:
Preamble. "I wanted to share something I have been thinking about." Delete. Start at the thought.
Restating the hook. Saying the same thing twice in different words, once at the top and once as you get into it.
Justifying the post. "This might be obvious but..." If it is obvious, cut the post. If it is not, cut the hedge.
Manufactured reflection. A closing paragraph about what this taught you about perseverance, added because posts seem like they should end that way.
Cut those four and most posts lose a third of their length and improve.
The truncation point
LinkedIn cuts posts off after a few lines with a "see more" link. The exact point varies by device and changes, so building around a precise character count is fragile.
What holds is that the opening has to earn the expansion. If the hook and the first line do not create a reason to continue, the rest is unread regardless of quality.
This is an argument for front loading the interesting part, not for a specific length.
Short posts feel risky
Most people find a 70 word post uncomfortable to publish. It feels like it is not enough, as though effort should be visible.
It is not. A short post that says one thing clearly is read completely, which is more than most long posts manage. Effort is visible in the specificity, not the word count.
If you are worried a short post is too thin, the check is whether you cut anything of value. If you cut only padding, it is not thin. It is edited.
The mechanical version
Noomachy DevTracker targets 100 to 220 words for LinkedIn by default, with an explicit instruction not to fill toward the limit, and the character count is enforced after generation rather than being left to the model's judgment.
That instruction exists because the default behaviour of a language model given a 3,000 character budget is to use a meaningful fraction of it. Length has to be a constraint, not a suggestion, or you get 800 words of well-formed padding every time.
The same is true when you write by hand. The limit is not the target.