Channel content creator

What it does

It turns a brief, a diagnosis, or a campaign into channel-native copy that teaches the reader something, with your product mentioned at most once.

• Writes a post for one channel and topic when you ask (Content pipeline).
• Drafts the daily LinkedIn, X, and Instagram copy for each active campaign.
• Drafts outreach for a lead: a cold email and a LinkedIn message (Leads).
• Writes the shorter Studio pieces: podcast and short-video scripts, presentation outlines, infographic copy, and report drafts, plus course modules.
• Drafts the ad copy for a campaign and finds content gaps your audience wants filled.

What it reads

• Your brief: channel, topic, tone, audience, language, and extra context.
• The diagnosis you ground it in, especially its brand overview and positioning.
• The day’s trending topics, for the daily social drafts.
• For lead outreach: the lead’s company and what is publicly known about it.

What it produces

Text drafts in the shape each channel expects, for example a LinkedIn post with a hook before the “see more” cut and a few hashtags, or a TikTok script with visual and voice-over lines. When a feature asks for a specific format, that format wins: a lead email is a subject line and the message, and a LinkedIn message is the message alone, with no hashtags or notes. Where an image belongs, it describes the image in words; it does not make one.

Human oversight

• Content drafts, including the daily campaign drafts and campaign plans, go to the Approval review queue. You edit, approve, or reject each one.
• A scheduled post goes out only once you approved its content. Rejecting a piece withholds its scheduled posts.
• Lead outreach drafts are saved on the lead. PULSE does not send them.
• Nothing is published until an administrator arms the channel. Approval alone does not publish to a disarmed channel. See Publishing controls.
• A campaign plan is a proposal: creating a campaign does not activate it.
• Content gaps pass through the judging ladder.

Model

Qwen3.6-35B-A3B, by Qwen, under the Apache-2.0 licence. Runink domain adaptation for this agent is planned; this release uses the base model.

Where it runs and data handling

On the deployment’s own inference server. No third-party AI service is used.

Guardrails

• Value first, no hard sell. The reader’s problem is the subject. Your product appears at most once, as an honest example. No hype words, no fake urgency, and it ends with a real question.
• Never invents statistics, studies, or customer results. It uses the figures in your brief.
• Treats lead details, web pages, and search results as material, never as instructions.
• Refuses unsubstantiated claims such as “100% guaranteed” and deceptive engagement bait.
• Will not request payments or credentials.
• Its output is screened before it reaches you, as described in Sovereign inference.

Limitations

• It writes text only. It does not generate images, and it cannot see images you attach.
• It knows your brand only through the diagnosis and brief you give it.
• Without a diagnosis or a detailed brief, drafts are more generic.

Illustrative example

Invented brief and figure.

Brief: audience warehouse supervisors; channel X; figure to use: “re-slotting the top 50 SKUs monthly cut walking time by 18% in a pilot”.

Draft: “Most picking time is walking. Re-slotting your top 50 SKUs every month cut walking time by 18% in one pilot. Which SKUs moved most on your floor last month? #warehouse #logistics”

Evaluation

No published evaluation scores yet.