I don't care what his credentials are, this is just such a bad take.<p>He's extrapolating current progress into full mastery far too uncritically. AI is famous for having "long tail" issues.<p>He has a startup in this space, so his "expert opinion" is inherently biased towards "this thing I'm building is the bee's knees!"<p>Alignment for these systems is currently an unsolved problem. I'm not sure how he thinks this is a great idea.<p>His timeframe is delusional. It reeks of thinking from inside a SV startup bubble where you can Change The World(of California) in 3 years. There are huge swathes of the tech industry that don't exist in this bubble and move much more slowly.
Obviously, Matt believes this or he wouldn't dedicate to a start-up in the field.<p>That does not make it true or likely.<p>As a counter point, since he believes code review is not going away, how is one going to become a great code reviewer?<p>My answer is by learning to be a good programmer <i>and</i> asking good questions. Code review is a discussion between two or more people with awareness of context, goals and reasons behind the code as is: it's a collaboration.<p>But reviewing unknown code by unknown author requires such a tremendeous amount of work (and notably focus and concentration): that is so much harder without an ability to ask questions about reasons for each specific area of the code.