The “Changes” tab shows a task’s changes grouped by repository and its pull request: the repository and the PR number, the number of added and deleted lines, the number of files.

Under the repository heading there is a list of files: the folder in grey, the file name more prominent, a mark of what happened to the file (A added, M modified, D deleted, R renamed, “binary”), and +added −deleted. All files are collapsed: the screen opens equally fast for three files and for three hundred.
Clicking a row opens the file; the “Expand all” button expands every file in the repository. The content is loaded on expanding — first a skeleton stands in for the code, then the lines appear. A long file arrives in parts as you scroll: at the bottom you see “Loaded N of M lines”; nothing is cut off, the whole file is readable.
While you scroll an expanded file, its header stays pinned at the top — with the repository name, the path and the counters, so you always know whose code you are looking at. The next file’s header replaces the previous one.
The “Patch” button downloads the module’s changes as a single .patch file — you can apply it on your side (git am).
If the task changed the development tools (the environment for building and running tests), a short summary of them sits at the top, with how many files and lines changed there. The tool files themselves are not shown.

Next to it is the “Discuss” tab — a discussion thread about the code from which you can propose a finding. The “Ask the assistant” button in a finding opens the same thread with a question about that finding already asked.