Deric Cheng and I just published Windfall Trust's first AI Economics Brief - a bi-weekly newsletter tracking developments in AI economics research and policy. I'll summarize the first one below:
Conflicting automation benchmarks: Remote labor benchmarks from OpenAI and the Center for AI Safety differ dramatically depending on task content and context. The latter's Remote Labor Index found AI agents could complete less than 2.5% of real freelance tasks successfully, while OpenAI's GDPval found Claude 4.1 Opus reaching near-parity (47% preference rate) on well-specified professional tasks.
New automation theory: Philip Trammell's Stanford paper suggests automation may arrive later but faster than expected. His workstream perspective reveals that automation has significant effects only once most tasks in a sequence are automated, reconciling the limited current impacts with credible effects for full automation. This matters for policy timing: if automation doesn't arrive gradually, policymakers must prepare for its arrival.
Anthropic's collection of policy proposals: Anthropic convened economists and policymakers in DC to discuss the economic impacts of AI across various scenarios. They presented various policy ideas including a sovereign wealth fund, tax code changes, and adjustment assistance for displaced workers.
Each issue will feature key papers with analysis, plus opportunities and recent news at the intersection of AI and economics.
Read and subscribe here!
