New research from Federal Reserve economists Leland Crane and Paul Soto confirms AI is already reshaping employment, particularly in coding sectors. Their analysis reveals around half a million fewer coders are working today than would have been expected if pre-LLM-era trends had continued, matching earlier findings from private payroll data.

The impact varies dramatically based on how tasks are bundled within jobs. Junior developers and contractors with "weak bundles" of discrete coding tasks face greater displacement, while senior developers whose roles combine programming with domain expertise see AI as an assistant rather than replacement. This explains why hiring for senior software roles continues to outperform junior positions.

AI agents are rapidly expanding the complexity of tasks they can handle independently, potentially affecting even experienced workers sooner than expected. The technology currently impacts early-career jobs most severely, but as AI capabilities advance, experienced professionals may find themselves in AI project management roles rather than creative positions, fundamentally altering career progression paths in tech.