The new diff format is awful

[nonspecialist]

Is your feature request related to a problem? Please describe.
Moving the "diff" output into an S3 bucket as a set of separate gzipped tarballs is just awful. The workflow to review changes in a pipeline has gone from "try and parse the diff output in a single pane from the CodeBuild job, which is sucky enough" to "download a bunch of tarballs to a local machine, unpack them, and go through the resulting files one by one". We already have people just blindly clicking "Approve" because they can't be bothered going through this new process. Adding cognitive load to an already not-great system is not the right direction.

Describe the feature you'd like
A diff stage in the pipeline which just shows what changed, instead of all the CDK guff that you have to pick through to work out what changed. Also, not having to download a dozen tarballs.

Additional context

[richardkeit]

While I might not agree with the tone, I agree that this is a lot more involved than a simple review of an output in a codebuild job.

There is a theme of having the DIFFs being more digestible, like:

• Feature Suggestion: Summarize Changes for Approval with Amazon Q #697

[nonspecialist]

Yes, I was a little snarky, but I stand by the assertion that this most recent change was a step backwards in usability.

A better way would be to:

• identify all the diffs that contain no changes (trivial) and remove them
• emit the actual diffs in JSON or a machine-readable format
• reformat as a single table and include that in a "summary" section before the manual approval, similar to what v1.10.1 did