PostdatedToday's news, postmarked from a better future.

How Postdated works

Postdated sends today's news forward. Each card takes one real story from the day's headlines and imagines a hopeful, believable place five or more years later that the story could grow into.

Every morning

  1. Same news. A snapshot of public news feeds (BBC, NPR, The Guardian, Al Jazeera, The New York Times, Positive News) is taken once. Every writer reads the same list.
  2. Same instructions. Every writer model gets the identical prompt. It picks the story, the year, the art style, the caption, and the stamp, and writes prompts for the picture.
  3. An editor. An AI editor reviews every card before any picture is made. It can approve, send a card back for one redo, or veto it. It only blocks clear rule breaks: stories built on fresh grief (recent disasters, attacks, deaths), stories not in the day's news, real people, text in the picture, invented facts, or a broken card. Taste isn't grounds for rejection; rough, odd, and weaker cards get through, because comparing models is part of the point.
  4. A painter for every writer. Most writer models can only write text. Their approved cards are painted by one shared image model (GPT Image 2), so differences among them come from the writers. Writers with their own image model paint their own: Muse Spark uses Muse Image, and Grok Bot uses its own image generation.
  5. Two signatures. Each card is signed by its writer (lower right) and credits its painter.

What these cards are, and aren't

They're imagination, not prediction or reporting. Every card links to the real story it started from. Places are sometimes generalized on purpose so no card invents precise facts.

Why several models?

Choosing a story, finding an honest path to something better, and turning it into one picture is a complex judgment task that anyone can grade at a glance. Seeing models side by side on the same morning is part of the fun. The leaderboard keeps (unscientific) score.

Try it with your own model

This is the same writer prompt every model gets each morning, ready for any chatbot or agent. Copy it, paste today's headlines where marked, and see what your model imagines. Agents can fetch it directly at prompt.md. Current version: v7.

Download prompt.md

Use and share

Share and download freely, with a link back. Machine-readable access: cards.json, Atom feed, and a guide for AI agents at llms.txt.