Show HN: I built "Docker for code", isolate AI logic into semantic containers

**Show HN: I built "Docker for code", isolate AI logic into semantic containers**

Hello HN, i want to share Glupe, a hobby project I've been working on for the last 30 days. Nothing fancy.I've been pretty annoyed by the "all or nothing" approach of most AI coding tools, which are useful, yes, but also risky. If i tell an AI to "fix this bug" it may probably do it, but it can (or will) hallucinate a fake library, mess up my manually optimized code or subtly delete important code without explaination or notice.Glupe summarized is this: "Stop giving AI root access to your code, Glupe isolates AI logic into semantic containers, so your code stays safe".Instead of sending your whole file to the LLM and pray it does not break, with Glupe you can surgically tell the AI were to write code and what to do using the $${ logic }$$ syntax. Code outside the container is preserved as is. Instructions inside the container are turned into source code and embedded into the output file.Containers can be named and are cached and hashed, Glupe detects which containers changed in a source file; when you run `glupe project.txt -make` it only calls the AI to update the "dirty" logic containers and uses cached code for the clean ones, saving time and API calls and allowing for incremental builds using AI.What if there is a syntax error on the output code...

---

**[devsupporter 해설]**

이 기사는 Show HN에서 제공하는 최신 개발 동향입니다. 관련 도구나 기술에 대해 더 알아보시려면 원본 링크를 참고하세요.