The economic index found AI inside real work

By UAF Editorial10 February 20251 min read

Anthropic’s Economic Index was an important change in how AI’s labor effects could be discussed. Instead of beginning with a forecast about jobs, it examined anonymized usage patterns and asked what people were actually doing with a model.

Usage data is not the labor market. It reflects one product, one population and one set of interactions. But it can reveal a useful middle layer between grand claims and anecdotes: the tasks where assistance is already being attempted, the occupations that appear in the data and the distinction between augmentation and automation.

The economic record should grow more empirical as systems grow more capable. We need to know not only what an agent can do, but which tasks it changes, who reviews the result and where the new coordination cost lands.

More from this desk

Society

a16z wants Silicon Valley to cosign its own reputation graph

Cosign turns endorsements and work relationships into a searchable network for people, companies and jobs. Its launch raises a larger question: who gets to curate the professional reputation graph that AI systems will use?

26 Sep 20263 min read

Autonomy

When the demo stops being the moat

Dan Greenheck’s reflection on Astra and Three.js work is a small record of a larger shift: visible demos are becoming cheaper, while judgment, maintainability and trust remain harder to automate.

22 Sep 20262 min read

Economy

The agent economy begins with task crossover

The early economic signal from AI agents is not that jobs have vanished. It is that tasks are crossing occupational boundaries. Workers are using AI to do pieces of work that were previously outside their formal role, then some of those pieces become recurring responsibilities.

22 Sep 20261 min read