Indie hacker Irtiza Hammad built Promptle around a single question: can you look at an AI-generated image and reconstruct the prompt that made it? Players submit their best guess and get scored on how close they are to the original text. The solo daily puzzle is fine. The PvP mode is what's drawing the crowd.
In head-to-head play, two players race to guess the same prompt within two minutes. That constraint changes everything — what would otherwise be a contemplative exercise becomes a pressure test, and the Indie Hackers comment thread suggests players are finding it addictive. An Elo ranking system gives the competition teeth; there's a rating to defend and a leaderboard to climb. A new image drops every 24 hours, borrowing the daily-ritual structure that made Wordle a phenomenon.
For the agent-ecosystem audience, the interesting angle isn't the game itself — it's the skill it trains. Working backwards from an image to the language that produced it forces players to think about how generative models interpret phrasing, style references, and compositional specificity. Several commenters noted that a handful of rounds measurably sharpened their prompting intuition. Hammad built the entire project with AI assistance, which is at least on-theme.
Some rough edges remain — a focus bug in the PvP text input was flagged by early users and acknowledged by the creator. But the core loop holds up. Promptle isn't marketed as a tutorial, and it doesn't need to be.