Skip to content

PreviewAll content, scores and forecasts here are illustrative sample data — not reporting, and not measurements.What this means

Agitology

AGI Index dimension · weight 9%

Planning

Goal decomposition and recovery when a plan fails.

Constructing a sequence of actions toward a goal, monitoring progress, and revising the plan when the world does not cooperate. Distinguished from reasoning by the presence of an environment that pushes back.

The argument

Evidence and counter-evidence

Both sides are published at equal weight. A framework that only records what raises a score is not measuring anything.

Raises the score

  • Multi-hour software and research tasks are completed end to end at rates that were near zero two years ago.
  • Replanning after a failed step is now common rather than exceptional, and systems increasingly detect failure without being told.

Holds it down

  • Success rate falls off steeply with task horizon; the curve is improving but its shape has not changed.
  • Systems rarely abandon a doomed strategy early — they persist, which is expensive and occasionally destructive.
  • Plans degrade badly under partial observability, where the cost of an unrecoverable action is highest.

What the score reads from

  • Long-horizon task completion

    Roughly half of multi-hour tasks; sharply lower past a day.

  • Recovery from injected failure

    Detects the failure more often than it finds a route around it.

Related coverage

Everything bearing on planning

News

Research

Argument

Context

Where this sits against the rest

Full Index
How these scores are producedMethodology v1.2