Show HN: Umme.ai understands your codebase without LLM https://bit.ly/3zat8yP

Show HN: Umme.ai understands your codebase without LLM Hello HN! We are excited to introduce Umme.ai. Umme.ai is an innovative AI model designed to help developers understand, optimize, and transpile their codebases across various programming languages with ease. Umme represents a paradigm shift in AI understanding by redefining the AI experience. As innovation consultants, we faced the challenge of automatically rewriting complex legacy applications written in multiple programming languages. Current AI assistants like GPT or Claude aren't well-suited for most logical or automated tasks, and we couldn't rely on time-consuming prompt designs. Therefore, we built Umme for everyone. From a technological perspective: - Umme has an internal world model - Umme directly interprets multiple languages without relying on programming language syntax. Many products use custom build systems and macros, making it challenging to understand a codebase exclusively based on official programming language grammar. - No LLM, no neural net: While the industry is saturated with neural-first AI models, our vision sets us apart by adopting a logic-first approach ideal for automated tasks and workflows. - Focus on reasoning, not creativity: Our model prioritizes logical reasoning over creativity, which is why it's not based on LLMs or neural nets for understanding your codebase. - U programming language[2]: At its core, Umme leverages our proprietary programming language, U. More details on U will be shared soon. - Extensive benchmarks: We use extensive codebases such as Node.js, Chrome, LLVM, Java, and C# as benchmarks. Our goal is for Umme to be able to accurately rewrite these codebases. Umme is designed to understand our world, starting with your codebase. We’re still in the early stages and would love to hear your feedback and understand your use cases. Don't miss out, get your early access now! Or contact us: contact at umme.ai [1]: https://bit.ly/4cb3etA [2]: https://bit.ly/4caPVcs https://bit.ly/4cb3etA June 20, 2024 at 08:00AM