The PromptifAI Project: Architecture of the Rotary Language Machines

From its trenches came to light the Promptifai project, a combination of 'prompt'—formal instruction—and 'ifai'—a contingency in artificial intelligence, a theoretical and practical investigation in computational literature. 

The prompt is not merely an accessory tool or a technical engineering sequence; it is an autonomous textual genre capable of undergoing a dense, recursive, self-aware, and metaprogrammatic writing process. 

This 'Living Archive,' structured on a Drupal 11-based platform internally named drupal-rlm, or Rotary Language Machines, operates like a light novel. It functions as a dynamic environment for compiling and executing instructions, designed so that human and non-human agents can act as translators, copyists, scrambles, shifters or modifiers, and text displayers, as well as generators of multimedia, text, images, video, and sound. 

The transformation system is called the Rotary Language Machine (RLM). It operates in three steps: a source text or multimedia file and a prompt retrieved from the site are processed by the user model, and the result is added in the final step, in the output, as a text and/or multimedia object. 

Prompts are code snippets, compact blocks of instructions or descriptions, that the user copies and pastes into any artificial intelligence application to obtain an output: transformed text, image, or video. 

Prompts are a narrative genre, and each narrative fragment behaves simultaneously as a parameterized semantic data structure. There is no division between narrative and technical instruction. The testbed consists of a matrix of 'bit!' type nodes, with revisions and "comments," which in turn can become new bits! Each node can operate as the snippet that serves as an instruction to co-produce a new bit! Well, not really; for now, there are bits! and rotators. 

With them, users can obtain textual, narrative, and metatheoretical outputs, with linguistic models that induce LLMs such as Gemini, DeepSeek, Vertex API, or Claude to continue the diegesis or to perform sociological, aesthetic, and philosophical reflections on the structural conditions of their own production, or even multimedia outputs with dense sensory descriptions, incorporating syntactic variables of cinematography, framing, lighting, optical textures, camera movements, and rhythmic transitions suitable for processing in engines such as Runway Gen-2, Pika Labs, or Stable Video Diffusion. 

* * * 

As demonstrated in the primordial JOURNEY 0 framework.