The three letters the industry measures its dreams with - and why nobody can say when it's “officially yes”.
The horizon has one property: however long you walk towards it, it stays ahead of you. AGI behaves the same way. Ten years ago, “a machine that talks like a human” sounded like the finish line. Today we carry one in our pockets and nobody calls it general intelligence - we moved the line. A computer beat the best chess player: the line moved. It writes code, passes bar exams: the line moves again.
That is why the AGI debate is rarely a debate about technology. It is a debate about definitions. Some measure by economics: a machine that can do most paid mental work. Others by science: a system that discovers new knowledge on its own. Others still by everyday life: give it a task like you would a colleague, and never check behind it. By the first measure we are close. By the third we are not.
Why the word carries weight when nobody has seen the thing
Because money and contracts move around it. The promise “we are building AGI” lifts company valuations and justifies costs that otherwise don't add up. And there are contracts where the word itself carries legal weight: declaring AGI changes who owns what. When a definition is worth billions, don't wait for someone to pin it down voluntarily.
Our read: watch what the systems do, not what they are called. Whether a model is “AGI” is a label. Whether it can do the work a person used to do - that is the question worth asking. And it has a measurable answer every single week.