AI Capabilities Look Overblown, Gains Stay Real

The strongest read is that AI is overblown as a prophecy and underrated as a tool. The gap is reliability, not raw usefulness.

• –Current models still fail too often for unsupervised high-stakes work.
• –Capability gains have been fast enough to keep resetting expectations.
• –The biggest near-term impact is leverage: one person can now do work that used to need a team.
• –Public debate keeps collapsing separate questions into one: autonomy, intelligence, and economics are not the same problem.