Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

github.com

Post reply on HN
Print view View original

41–49 of 49 posts

Re: Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

#41

Post by neilunlimitess »

shrishdwi wrote: post #27
xhrpost wrote: post #23

I intuitively and perhaps naively thought that Claude using the LSP server would negate a lot of grep use. Is this tool solving the same problem or something else?

LSP is for when you already know the symbol and want an exact answer, like go-to-definition or find-references. graft is for the step before that: you ask "where does auth happen" and it hands the agent the right files and lines to start from, so it reads a lot less to get oriented. so graft solves for making claude code understand your codebase every session. And I feel you can save a lot more tokens and also increa…

[flagged]

Re: Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

#42

Post by subscribed »

skerit wrote: post #7

Does Claude's grep still prepend the relative path of the file before _every_ single line? Because that is nasty, especially in java projects.

Absolute for me. In the next project I'm going to twist its sandbox further and just bind mount the actual project directory as /wtf and make it work there.

Re: Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

#43

Post by subscribed »

bdangubic wrote: post #40
jatins wrote: post #29

slop README. If you couldn't be bothered to write it, don't expect people to read it.

Why is this a problem?! I actually prefer to see README written by the author of the code

README is for humans if a human comes up and says "this project is for humans".

Slop is excessively redundant. Basically "purple prose" of the technical documentation.

But primarily: if its not worth writing, its not worth reading.

(I use Claude to prepare half of my technical documentation but then i review EVERY single file. And every single one gets extensively edited despite being coherent from the outset. I just feel like it would be disrespectful to get LLM spit something out and ask my colleagues to read it. Like.... come on.)

Re: Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

#44

Post by bdangubic »

subscribed wrote: post #43
bdangubic wrote: post #40

Earlier quoted context omitted.

Why is this a problem?! I actually prefer to see README written by the author of the code

README is for humans if a human comes up and says "this project is for humans". Slop is excessively redundant. Basically "purple prose" of the technical documentation. But primarily: if its not worth writing, its not worth reading. (I use Claude to prepare half of my technical documentation but then i review EVERY single file. And every single one gets extensively edited despite being coherent from the outset. I just…

README should be written by the author of the code, it this case it is fully LLM written code so let the brother write the README.

Also “README is for humans” might have been true awhile ago, these days README’s are being tailored to Agent as they write most of the code, not humans

Re: Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

#45

Post by jatins »

bdangubic wrote: post #40
jatins wrote: post #29

slop README. If you couldn't be bothered to write it, don't expect people to read it.

Why is this a problem?! I actually prefer to see README written by the author of the code

As I said: If you couldn't be bothered to write it, don't expect people to read it. I find it disrespectful to your audience

That said my own views on what is/isnt slop is evolving i understand your pov

Re: Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

#46

Post by subscribed »

bdangubic wrote: post #44
subscribed wrote: post #43

Earlier quoted context omitted.

README is for humans if a human comes up and says "this project is for humans". Slop is excessively redundant. Basically "purple prose" of the technical documentation. But primarily: if its not worth writing, its not worth reading. (I use Claude to prepare half of my technical documentation but then i review EVERY single file. And every single one gets extensively edited despite being coherent from the outset. I just…

README should be written by the author of the code, it this case it is fully LLM written code so let the brother write the README. Also “README is for humans” might have been true awhile ago, these days README’s are being tailored to Agent as they write most of the code, not humans

README is the first document of the repo read by the human stumbling upon the repository. Even if it's written by LLM it should be edited and managed by the human maintainer.

Sharing unedited documentation like this is sloppy and speaks for itself ("I can't be bothered to control it"). Leaving readme only to agents is no class, since an alternative is one line of reference to the complete documentation for bots - no need to hijack the repo's first document designation with _that_.

Re: Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

#48

Post by jmavid »

peter_d_sherman wrote: post #5

>"The problem: Every task, your coding agent starts blind. Before it changes anything, it re-explores the repo: grep a term, open a file, follow an import, back out, try again. It is rebuilding a picture of a codebase it mapped an hour ago and threw away. That rediscovery burns most of a run's tool calls, tokens, and latency, and it is pure overhead " The author of this article brings up a very interesting problem --…

Try AgisHub's 30 tools for AI agents at https://api.agishub.com/try. No signup, no API keys.".
Post reply on HN