And who knows what the economic incentives are in couple years. Maybe the focus shifts towards one-shotting complex million SLOC codebases which are order of magnitude more sloppy than the current outputs.