As you can see, there is no real consensus on the “best solution” to the simple “how do you change things” problem. My 5c: none of these tools give the model a stable, verifiable identifier for the lines it wants to change without wasting tremendous amounts of context and depending on perfect recall. They all rely on the model reproducing content it already saw. When it can’t — and it often can’t — the user blames the model.

[...]

Now bear with me here. What if, when the model reads a file, or greps for something, every line comes back tagged with a 2-3 character content hash:

`11:a3|function hello() { 22:f1| return "world"; 33:0e|}`

When the model edits, it references those tags — “replace line 2:f1, replace range 1:a3 through 3:0e, insert after 3:0e.” If the file changed since the last read, the hashes (optimistically) won’t match and the edit is rejected before anything gets corrupted.