What makes a build-in-public post on X get 10x engagement

The takeaway is not "post about your tax problems." It is that the single biggest hits for build-in-public accounts were unpredictable, off-topic, personal moments, the kind you cannot schedule. The product-relevant formats further down the list (an AI-does-my-job post, a system map, a save-list) reached lower multiples individually, but they are formats a creator can choose to write on purpose, which makes them more useful for a content calendar than hoping for a viral rant.

This is the opposite of what our LinkedIn research finds, where images beat text-only posts by 36%. On X, format alone does not create the outlier: what the post says appears to carry more weight than whether it has a picture.

Read this as two different jobs: write short (a hook plus 2 to 4 short lines) for consistent, predictable engagement rate, and treat the occasional long-form post as your highest-variance swing, not your default. It will not always be your best-performing post by rate, but when a creator's long post connects, it tends to beat their own baseline by more than any short one does.

Note the n and creator count next to every bar in the chart above: the highest bars (system map, launch, reply gate) are proven on very few tweets from very few creators, so treat them as promising, not universal. If you can only build one repeatable habit from this data, a save-list is the best-evidenced choice: common enough that most of the 15 creators used it, and consistently above their own baseline when they did.

• This is a selected, outlier-skewed sample. We gathered each creator's own top-liked tweets, so both the outlier rate and the format and length findings describe a creator's better posts, not a random sample of everything they post.
• Causation. These are observational comparisons on a small set of accounts. A creator choosing to write a save-list after a good week is not the same as save-lists causing the good week.
• Views, saves and impressions data were incomplete for some tweets. Engagement-rate comparisons (format and length) use the 1,188 tweets with 1,000 or more recorded views; the outlier, timing, hook and pattern comparisons use likes against each creator's own median across all 1,209 core tweets.
• Small n on the sharpest findings. The posting-time and hook-signal comparisons rest on only 43 outlier tweets. Read the directional shape, not the exact percentages, as the takeaway.
• 15 creators, not a market. This describes what worked for this specific set of build-in-public accounts on X, not every niche or every account size.

For a size-for-size comparison, our State of LinkedIn 2026 study measured the same kind of within-author effects on 134,522 LinkedIn posts: there, images beat text-only posts by 36% and longer posts kept winning past 2,400 characters, the opposite of what this X data shows. The two platforms reward different things.