I can build almost any piece of software I can dream up, the data never leaves my machine, and there's literally no way for anyone to charge me any price ever.
I'm struggling to understand the point of your comment. Why would that affect whether or not I am over the pricing barrel with a frontier lab? Like is it easier to verify Opus code than Qwen code? The code is the code.
But the tools we have to do this suck. We don't have much beyond "here's the source code as text, look at it, and think really hard".
Another comment[0] suggested having AI generate "code walks" through the code. This seems better than our current approach of "my PR contains a diff, good luck".
I'm working on a tool in Emacs that allows me to mark certain lines, or entire projects, as needing review, and then helps me track which parts have been reviewed. I need something to help me find parts of the code I have never reviewed before.
When lines change, they are put up for review. When lines haven't been reviewed for awhile, they are put up for review. Lines are ranked based upon the complexity of the context they are in, and complex areas of the code are reviewed more often.
Basically, stop focusing on individual PRs and instead focus on continuous review of the entire system with tooling that helps identify those parts of the code that are most in need of review.
Reviewing PRs is wasteful because the optimal way to review a system does not match the way a system naturally grows through individual commits.
Our job, much like a locomotive operator, is to make sure the agents don't go off the rails. We call ourselves software engineers in recognition of that.