Content

Turning what the listeners heard into something worth publishing. The unit is evidence, not a topic: anyone can ask a model for blog ideas, and what this has that a generic writer does not is dozens of real people asking the same question in public, with quotes and dates.

​

Ideas

Every listener mode produces themes — clusters of signals saying the same thing. A run can raise a content suggestion on its own on eligible plans, or you can ask for one, or name a topic yourself. Most runs suggest nothing, and that is the feature. The bar for volunteering an idea is a setting, so you can make it harder or easier.

​

Targets

A channel is who ranks the post; a target is one publishable artefact on it. One idea produces one piece per switched-on target, each written from the shared brief and never from another target’s output. An unconfigured workspace gets blog and LinkedIn. Each target costs one piece of the monthly allowance, so an Instagram post and an Instagram story are two — see openpulse.cloud/pricing for what your plan includes.

​

Six channels, six generators

Not one prompt with a length parameter — each channel is governed by a different ranking system and different community rules:

• Blog opens with the evidence, because that specificity is what search and answer engines cite. Internal links come only from your own sitemap; anything else is blocked, because an invented internal link is a 404 published under your name.
• LinkedIn writes something worth keeping rather than applauding, because saves and dwell outrank likes. A URL in the body is flagged as blocking — it suppresses reach, and the link belongs in the first comment.
• X is checked against 280 characters in code rather than asked for in a prompt, and a link counts as 23 characters however long it is.
• Reddit replies to a real thread, never posts. An idea with no Reddit signal behind it is declined rather than turned into the promotional post that gets accounts banned.
• Instagram produces a script — frames with copy, visual direction and mandatory alt text — written from the evidence rather than from another channel’s draft.
• Facebook keeps the page post and the group version as separate fields, and a group version that is the page post reworded is flagged.

No channel ever sees another channel’s output.

​

Reviewed before you see it

Every draft goes through a deterministic linter — character counts, link provenance, promotional phrasing — and then an editorial review on a different model family from the writer, because a model reviewing its own prose agrees with the choices it just made. At most one fix round.

​

Editing, versions and export

Prose is edited in place. Structured parts (blog sections, X posts, Instagram frames) are changed by asking, so the shape stays valid. Every prior state is kept and restoring one is itself an edit, so history is append-only. If a colleague saved while you were editing, you are offered “take theirs” or “keep mine” rather than having the two merged unasked. Export is Markdown for the blog and plain text for everything else. The AI-disclosure note is kept separate from the copy rather than appended to it: several jurisdictions now require significantly AI-generated content to be labelled, and LinkedIn, Meta and TikTok run their own composer-level markers. The export tells you where to tick those, without adding a sentence you might publish by accident.

​

Images

Generated alongside the copy, at the exact size each platform needs. Redraws are budgeted per piece rather than per month, so the question “can this piece afford another go” has an answer before you press the button. A carousel is one draw and six pictures. See openpulse.cloud/pricing for the redraw budget on your plan.