Fundamentally, I agree with him. If you want to make a game, use technology that is already out there. Use the stuff that is best suited to your specifications. Using a 'game-maker program' isn't uncool, it's efficient, and anyone who wants to continue asserting that the only way to do anything is to write it all yourself (preferably in assembly) can GJOAC.
However, I disagree with a few points - mostly the one about our not needing to make tools because our existing ones are 'mature.' That's right, they are; they've been in development for years now, refining the arts of things like per-polygon character modelling.
Guess what? Per-polygon character modelling is useless in the long run.
As the standard polygon budget for a highest-detail character averages out at six figures, we can't continue using tools that were created to work using five-year-old paradigms. The same goes for engines that were designed without any concurrency - rumour has it that all the next gen consoles are going to be multiprocessing, and we've already got tech like hyperthreading / dual-core appearing on the PC.
Writing my own engine is fun (hell, it's a lot more fun than doing the content, that part sucks ass).