“Zach Lloyd assesses the current capability of AI coding agents as a "six out of ten," able to handle medium-complexity tasks but not large projects or fundamental architecture decisions.”