Show HN: Smile – an open source language for structuring prompts (github.com)

The open-source project (: Smile introduces a novel prompt language designed to bring clear, structured syntax to prompt engineering for large language models (LLMs). By leveraging emoticon-like brackets instead of traditional code symbols, (: Smile creates an intuitive, consistent format that both humans and machines can easily interpret. This structured approach aims to reduce ambiguity and improve instruction following, addressing a critical challenge in prompt design where the slightest syntactic changes can greatly impact model performance. What sets (: Smile apart is its focus on organizing complex multi-turn conversations, intricate instructions, and large prompt contexts in a maintainable and portable way. It separates prompt language from response formatting, allowing developers to define roles, tasks, data, tone, and response styles as discrete sections marked by unique emoticon delimiters. This methodology enhances the model’s meta-awareness—prompting it to prepare, analyze, and respond thoughtfully—akin to the popular “chain-of-thought” reasoning pattern but encoded with structured markers. Compatibility across popular LLM platforms such as OpenAI’s ChatGPT, Anthropic’s Claude, and Moonshot AI’s Kimi Chat makes it a versatile tool for enterprises needing reliable, repeatable AI behavior in complex workflows. For the AI/ML community, (: Smile offers a promising pathway to more robust prompt engineering by formalizing prompt structure into a lightweight, human-readable “language.” This can improve consistency, reduce instruction drift, and enable scalable, collaborative prompt development. Its open-source nature invites experimentation and adaptation, potentially setting a new standard for how instructions are coded, shared, and understood by both AI systems and their human creators.