wtf is meat?

meat is a command line tool that takes a git commit or diff and uses an AI model to shrink it down into what the project calls a reading diff, a version of the change that keeps only the parts worth a human's attention. The idea behind it is that AI coding models have gotten reliable enough that a person reviewing their output no longer needs to check every line for style issues, missing null checks, or import statements. What still needs a careful human look is the conceptual stuff: which algorithm was chosen, how the architecture changed, and why. meat tries to strip a diff down to just that. You install it with a single Go install command, then run the meat command by itself to review the most recent commit in a repository, or pass it commit references similar to how you would with git to review something else. Processing a commit takes some time, since the model has to read through the change and decide what actually matters, so the project suggests wiring meat into your development tools so a commit gets pre-processed automatically before you actually want to read it, rather than waiting on demand. For very large changes, meat splits the diff at file and code hunk boundaries and processes it in chunks of up to a few megabytes at a time. Even a very large commit still ends up as a single combined reading diff at the end, it just takes correspondingly longer to produce. The project is written in Go and does not state a license in the excerpt provided.