I asked Manus AI (the invite-only agent that sells for $20K on secondary markets) to create its own open-source alternative.
The result is ANUS (Autonomous Networked Utility System), a complete AI agent framework with architecture, documentation, and tooling - 100% designed and coded by AI.
My role? I wrote the prompt and used Cursor to push to GitHub. That's it.
The project features hybrid agent architecture, multi-agent collaboration, comprehensive tool integration, and flexible model support. It demonstrates how AI can recursively improve AI tools.
Repository: https://github.com/nikmcfly/ANUS
I'm interested in the HN community's thoughts on:
Technical aspects of the framework
Implications of AI systems building other AI systems
What this means for the future of programming
This feels like a significant milestone - when one AI system can create a complete, functional alternative to itself from scratch.
I get that it's a clever hack, but this also seems like an incredible testimonial to the capabilities of Manus. Is Anus also able to "self-host" itself (generate an equivalent app), or, if not, is that a measure of a capability differences between the two systems?
https://github.com/lightvector/KataGo/blob/master/docs/Analy...
여기는 카타고 바둑AI를 병렬로 처리할 수 있는 방법이 나와 있는데
나의 PC에서 쿼리값 리스트가 있는 엑셀파일을 입력하여 모든 쿼리에 대한 카타고 AI 분석값도 엑셀파일로 얻을 수 있는 EXE 프로그램을 만들어줘.
인터넷이 연결안되도 실행될 수 있는 프로그램이어야해!
카타고 AI엔진 파일 폴더 위치는 cd C:\Users\kimjonghyun\Downloads\katago 여기고
카타고 실행 세팅값은 katago.exe analysis -model Kata18b.gz -config analysis_example.cfg 여기야~
분석값에 대한 옵션에서 maxVisits":1 로 설정하면 되
When the Romans were crucifying Jews outside the walls of Jerusalem, the victims were nude, but then the Jewish leaders persuaded the Romans to provide the victims with loincloths while they are being crucified (to reduce the level of indignity).
Similarly, although it is unlikely that the AI labs will be shut down (like they should be IMHO), maybe it is feasible to persuade people to avoid giving an undignified name to anything that might kill me and everyone I care about.
In other words, if I'm going to be killed, and there's no way to stop it, I'd rather avoid the additional indignity of being killed by something named "Anus" or such. Is that too much to ask?
Technical aspects of the framework Implications of AI systems building other AI systems What this means for the future of programming
This feels like a significant milestone - when one AI system can create a complete, functional alternative to itself from scratch.