Is AI About to “Eat Everything”? (It’s Not.)

The gist

1. The Meter chart tracks only how well AI handles coding tasks at a 50% success rate, not general intelligence or any 12-hour human task.
2. Newport says coding gains came largely from handcoded expert-system logic built into coding tools, not smarter underlying models.
3. Progress looks exponential on coding benchmarks but only linear on broader capability indexes, showing improvement is application-specific.
4. Newport argues transhumanist communities have pushed apocalyptic framings that distort how AI companies, investors, and policymakers talk about the technology.