Three channels ask the reader to do different things
A newsletter introduction invites further reading. A carousel needs steps that stand on their own. A video needs language that works aloud and a visual plan someone can produce. We used our published research-assistant article to make all three.
This task produced three Chinese drafts, three English adaptations and a review record in Codex. The card preview typesets the actual copy. The video remains a script: no email, social post or finished video was published by this exercise.
Choose one source of facts
We retrieved the published Chinese research-assistant article and saved it in the pack. It records six sources, a 62-article inventory, two selected directions and two deferred ones. Those are the facts used by the adaptations. The English article offers the same context.
The source distinguishes a fully read automatic transcript from other video references, and identifies Codex as the environment where the work completed. Shortening the story must not turn that into all videos watched or a successful Claude Code test.
Download the working pack. It includes the source snapshot, a brief, six actual drafts and review.md. No dependency installation is required to read it.
The newsletter introduction points to a useful next read
Here is the body from the actual English adaptation:
A folder full of tutorials can leave you unsure where to start. We took one editorial assignment and gave an assistant source notes, a reader brief and an existing article inventory. The output was a set of topics, reasons to defer other ideas, and concrete next actions.
The task left two selected directions and two deferred ones. The article includes the actual materials and a working pack you can try with your own tutorials.
The review question is concrete: can someone find the promised material after following the link? The cards and working pack exist. No measured efficiency improvement or reader feedback was supplied, so none appears in the introduction.
Give each card one action
The three cards ask the reader to name a deliverable, record how deeply each source was read, and request an action card. That sequence helps someone prepare without reading the entire article first.
Input counts belong in supporting context, not on every card. The figure typesets the copy created in this task; it is neither a platform screenshot nor evidence of social reach.
A voice script needs a visual plan
The script has four shots: the saved-tutorial problem, three inputs, four editorial decisions, and a next exercise. For the third shot, the voice explains that our task selected two directions and deferred two. Its visual plan shows those actual source decisions and labels them as this task’s record.
The opening bookmark list is a proposed illustration. It must not be presented as a real person’s browsing history. The script is also not labeled as a sixty-second video: timing requires reading it aloud and accounting for pauses, transitions and visual explanations. The video files in the pack are inputs for that later production work.
A reusable request
Read the confirmed article and use only traceable facts.
Create three drafts:
1. Newsletter intro: reader problem and material in the full article.
2. Three cards: one actionable step per card.
3. Video storyboard: separate voice copy from planned visuals.
Keep the source link. Add no results, prices, features or testimonials.
Separate actual artifacts, proposed visuals and actions not performed.
Save drafts only; do not send or publish.
Check every promise against the source before handing them over.
This request packages the editorial requirements used here. Specify audience and channel purpose first. If the output merely cuts the article into three lengths, identify the missing reading task and revise its structure.
Hand over the copy with its status
Our review preserves several distinctions: two selected topics are not a general success rate; selection is not publication; a script is not a finished video; a typeset preview is not a platform screenshot. Every draft retains a source link and adds no tool capability or performance claim.
These checks establish correspondence with this source, not future clicks or conversions. Actual publication should have its own record of channel, final version, date and feedback.
The notes-to-article workflow and this adaptation task now provide a connected set of artifacts. Try another article with the same method before deciding which stable steps deserve a reusable Skill.