← Home From Native to Derived Content types Completeness
The whole picture New

One keyword, every form it takes.

Everything the app makes for one keyword. Read it left to right.

Derivation map

  • Input list
  • Native
  • Derived
  • Derived · 2nd level
  • Derived data
Derivation map The input is the Topic's keyword list, made in one of three ways: typed (A) or uploaded (B), which are native-style, or generated from the Topic by an AI model (C), which is derived-style. Each keyword is passed in one at a time. A second input, the Filter list, is extracted from the words in the PDF text, transcripts and EXIF data and supplies the checkbox options in Filter Words Search. Tasks fetch native content from the Internet: images, video and PDF articles (URLs from serper.dev for the Topic and keyword), news, shopping, Google Trends and Power Search. From the keyword, an AI model derives the definition, and Stable Diffusion derives pop art. From native files come JPG/EXIF readouts, filter words (each group's vocabulary, with parts of speech), PDF glossaries and summaries, AI long-tail keywords, and popularity stars and dynamic links for each news and shopping site; the long-tail keywords seed Power Search, a native grid of searches on seven engines. At the second level, the definition becomes narration MP3 and a wordcloud; narration plus images, pop art and video frames becomes the slideshow, navigated keyword by keyword; EXIF, PDFs and transcripts become fun facts. Filter words and long-tail keywords feed Filter Words Search, which searches the full corpus of PDF text, EXIF data or transcripts by word or part of speech. Repeated for every keyword in the Topic, the whole Topic group is published in one click as a website, with a Topic-wide full-text search over every PDF, transcript and EXIF record. Everything is produced by dynamically generated tasks, one for each user, topic, keyword and content type, run by background workers. INPUTKEYWORD LISTATypedBUploadedCAI modelA · B: native-styleC: derived-styleNATIVE · INTERNET, UNALTEREDDERIVEDDERIVED · 2ND LEVEL tasks fetch from the Internetserper.dev · Topic + keywordURLs for images · video · PDFsAI modelStable DiffusionextractorAI modelextractorAI modelAI long-tail variations seedPower Search: 7 search enginesAI-found sitesDynamic linksKokoro TTSAI modelimages + video framescheckbox optionsto search the corpusPDF text · EXIF · transcriptsby word or part of speech · AND/ORnavigate keyword by keyword+ PDFs · YouTube transcriptsDefinitionPop artImagesVideoArticles (PDF)NewsShoppingGoogle TrendsPower SearchJPG / EXIFFilter listPDF Glossaries& Summaries“…”Long-tail keywords★Popularity starsNarration MP3WordcloudSlideshowFun FactsFilter Words Searchrepeated for every keyword in the Topic, then:Topic groupevery keyword, native + derivedPublishone click · % completePublished websitetopic cards · browse by keywordTopic-wide searchfull text of every PDF, transcript, EXIFTHE ENGINEdynamically generated tasksuser›topic›keyword›content type= one task, run in the background · tracked in Completeness TasksONE KEYWORDborder collie+ Topic: Dog breeds

Scroll sideways to see the whole map.

  • Input. The Topic's keyword list, one keyword at a time. Typed (A) or uploaded (B) lists come from you, like native content. An AI-generated list (C) is derived.
  • Native. Tasks fetch it from the Internet and show it unaltered. serper.dev supplies the image, video and article URLs for each Topic + keyword.
  • Derived. AI models, text-to-speech and extractors generate it from the keyword and its native files.
  • Derived, 2nd level. Made from derived content, e.g. Definition → Narration MP3 → Slideshow.
  • Filter list. Extracted from the words in the PDFs, transcripts and EXIF data. It becomes the checkbox options in Filter Words Search.
  • Bottom row. Repeated for every keyword, then the whole Topic is published as a website with one search box.
  • The engine. Every box is built by its own background task: user › topic › keyword › content type.
Native & Derived Content New

Two kinds of content. One interface.

Every piece of content is one of two kinds. Both open the same way.

Native content

From the Internet, unaltered

An endpoint on the Internet, such as a photo, a YouTube video, a PDF article, a news story, a product page or a Google Trends chart, brought in for the keyword and displayed as is.

  • Origin: a real address on the Internet, and every item traces back to it.
  • Altered? Never. What you see is what its publisher put online.
  • Role: the source files that derived content is generated from.
  • Examples: 40 photos of a border collie, a dozen PDF articles, today's headlines.
Derived content

Generated from native files

New content we create by running the keyword and its native files through AI models, text-to-speech, image generation and extractors. It didn't exist on the Internet until we made it.

  • Origin: generated here, by an AI model, Kokoro text-to-speech, Stable Diffusion or an extractor.
  • Altered? It is the alteration: a summary, a reading, a readout or an image that's new.
  • Stacks: derived content can be derived again. The definition becomes narration, and the narration voices a slideshow of images, AI pop art and video frames.
  • Examples: an AI definition and its MP3 reading, fun facts written from PDFs, EXIF data and transcripts, long-tail keywords, and the popularity stars and dynamic links for each news and shopping site.

One universal content UI

Every type, native or derived, is a button in the same Content Type grid, with the same controls. New types join the grid the same way. A button lights up when its content is ready and greys out when it isn't.

Native contentDerived content
Where it originatesAn endpoint on the Internet: a web page, YouTube, a PDF, an outlet, a store, Google TrendsHere, generated from the keyword and its native files
How it's shownUnaltered, exactly as publishedAs new content: a definition, a reading, a summary, a readout, an image
What makes itIts original publisherAI models, Kokoro text-to-speech, Stable Diffusion, extractors
Also coversThe pages and files themselves, plus live searches through Power SearchThe data that organizes them: filter lists, long-tail keywords, popularity stars and dynamic links. An AI-generated keyword list (Method C) is derived too.
Can it be derived again?It's the starting pointYes. Narration comes from the definition, and the slideshow combines the narration with images, pop art and video frames
How you open itThe same way: a button in the Content Type grid, shown in the middle section
← HomeContent types →