MonkeysPaw – a prompt-driven web framework in Ruby (worksonmymachine.ai)

MonkeysPaw is a new Ruby web framework that turns every page into a prompt: you write “wishes” in markdown (title, hero, features, etc.), and an LLM generates the full HTML/CSS/JS page, routing follows file structure, and layouts/styling are driven by natural-language descriptions. Released as a Ruby gem (sublayerapp/monkeyspaw on GitHub), it caches generated output to avoid unnecessary re-renders and exposes typical web behavior (buttons, interactions) by translating vague gestures into working JavaScript. The project grew from a RubyConf talk about applying Postel’s law to LLMs—accepting whatever the model produces rather than forcing exact outputs—and treats hallucinations as a feature rather than a bug. For the AI/ML community this is a concrete example of “natural language as source code” and content-first development: it lowers the barrier from idea to working UI and reframes developer intention as the primary artifact. That opens creative workflows but brings predictable trade-offs—performance costs, unpredictable or slightly-off outputs, and brittle behavior for complex interactions that may require precise wording or retries. The author plans improvements such as component-level caching, image generation, preloading, and dynamically generating ERB partials by rescuing NameError. MonkeysPaw is primarily an experiment in expression-first frameworks; it’s useful for rapid prototyping and exploring how LLMs reshape frontend development, but it requires careful prompt design and user testing.