Where AI Actually Is

The AI conversation sits in two camps. One says we're plateauing. The models look the same, the advancements are incremental, and when companies start to pay for actual compute cost, the ROI of many AI initiatives is going to be in the red. The other says we're sprinting toward the "singularity": agents will replace white-collar workers in eighteen months, and civilization is at stake.

In this issue, we look at reports from major institutions about the state of AI today. Capability is accelerating. Adoption is moving faster than the internet did. The massive workforce disruption hasn't happened yet, but employment for junior software developers fell nearly 20% in the last year, and a third of organizations expect AI to reduce their workforce in the coming year.

The map of what AI can and can't do well, what is often called the Jagged Frontier, has a higher ceiling than it did a year ago. Yet it's still jagged enough that "AI can do everything" and "AI is overhyped" are both still wrong by the same amount.

This issue is a reality check: where is AI right now, on April 30, 2026?