I was so hoping that this would not be about AI, and actually talk about how we need to do better as an industry and start using objective measures of software quality backed by government standards.

Nope, its about AI code reviewing AI, and how that's a good thing.

Its like everyone suddenly forgot the old adage: "code is a liability".

"We write code twice as fast!" just means "we create liability twice as fast!". It's not a good thing, at all.

This is advertising for an AI product. Slightly more interesting background story than most articles doing so, but still an ad for a product that probably won't actually work.

Adversarial agents seem to be unreasonably effective, and probably we're just at the beginning of grasping how useful they can be to almost any agentic use case.

Wherever you have competing priorities and trade-offs (i.e. everywhere), there is likely value in a sub-agent that does its best to make the case for every alternative, as well as sub-agents devoted to refuting each other's arguments. Then at the top level, you have a synthesis agent that parses all the debate and determines the ideal course of action.

The human brain seems to work similarly: we have many relatively independent cognitive subsystems with their own areas of focus, and these are often at odds with each other. The "executive function" process weighs the clamoring of every subsystem and then decides what to actually do in each moment.