kirby88

4 ms·

1. 1.▲
   Show HN: Conversations as a first class citizen in AI coding agent(github.com)

   2 points by kirby88 4mo ago|0 comments
2. 2.▲
   Show HN: A whiteboard for your AI coding agent [video](youtube.com)

   1 points by kirby88 4mo ago|0 comments
3. 3.▲
   by kirby88 5mo ago

   Claude code is closed source as well (at least before the leak). But point taken thank you.
4. 4.▲
   by kirby88 5mo ago

   Using a virtual filesystem to minify source code and let the LLM evolve in the minified space gives surprisingly good results and reduce tokens by 20 to 40% (depends on the language, python for instance)
5. 5.▲
   Token efficient source code representation using virtual filesystem(github.com)

   2 points by kirby88 5mo ago|4 comments
6. 6.▲
   by kirby88 5mo ago

   I've been building an AI coding agent that using the exact same prompt than claude code, but uses a virtual filesystem to minify source code + the concept of stem agents (general agents that specializes during the conversation for maxi
7. 7.▲
   Show HN: I built an AI coding agent 50% cheaper than Claude Code (same prompts)(github.com)

   4 points by kirby88 5mo ago|2 comments
8. 8.▲
   by kirby88 5mo ago

   I wonder how that compare to harness methods like MAKER https://www.cognizant.com/us/en/ai-lab/blog/maker
9. 9.▲
   by kirby88 5y ago

   I don't think so, you just forgot to add the remote server address. If you run `system "git", "ls-remote", " --upload-pack=$(touch hello) git@github.com:torvalds/linux.git", "HEAD"` it will
10. 10.▲
    by kirby88 5y ago

    You explained what this option is, I was asking why.
11. 11.▲
    by kirby88 5y ago

    I had the same reaction as you... it looks like the vulnerability is coming from git, that's a very dangerous behavior.
12. 12.▲
    by kirby88 5y ago

    I still don't get why the --upload-pack option execute its content... Seems pretty dangerous to me. How did you find that out?