Format and publish a finished article from a transcript and screenshots

Andrew Best, who builds the screen recorder Podsplice with his brother, recorded an unscripted tutorial video and took five screenshots of the key steps. He pasted the transcript into ChatGPT to shape a draft article, edited it so it sounded like him, and then had ChatGPT write instructions for its desktop agent. In a single prompt the agent was asked to put the screenshots in order, draw red circles and arrows on the buttons readers should click, save them with descriptive file names and alt text, upload them into the right sections, format headings and links, embed the YouTube video, pick tags, fill in the excerpt and SEO fields, and publish. He did this for a how-to article on the Podsplice website and again for the Medium post describing it.

Best says the agent worked out the screenshot order on its own, made the markings, uploaded and placed the images, embedded the video and published both articles. He checked the result afterwards and made a couple of small changes. He gives no timings beyond saying the manual version usually takes him about half an hour per article.

The writing stayed human: the agent only did the layout and publishing chores, which is the point of the post. Best is promoting his own product, and he does not say which editor his website runs on or whether the agent needed him to log in for it. Give it the order of steps and check the live page before sharing it.