[Discussion] Claude cutting segments of files to move them speed up

So I've noticed that when Claude is moving large chunks from one location (or one file) to another location or file, it's very slow, and scales with the length of the chunk.

Is the slowness due to Claude having to interpret the entire code chunk to determine where the end is, to do the cut?

If so, is there not a way to speed this up dramatically using core util file tools, or custom scripts? For example, if Claude knows the start and end line, it can call to a local script to move it, i.e.

``move infile_start infile_end outfile_start outfile_end filepath``

Where move can be whatever (executable, bash script, python script, etc). I'm sure there's also a bunch of core util tools available for this that I'm unaware of.

Claude can also use local scripts or exes to parse the file if it is a known programming language as well, to parse the tree and determine function start/end locations, and use this to inject into other files.

Point being, it feels like there may be a lot of wasted execution and inference happening that can be optimized out using clever strategies, which should result in a massive speed-up, cost savings and free up a bunch of compute for Anthropic.