AI Insights & Research

852 reproducible tasks drawn from workplace sessions bring artifact quality, cost and runtime into enterprise agent evaluation.

When agents retrieve internal material, how can they share what is necessary while withholding sensitive context? CI-Work turns that tension into an evaluation.

A look at designing the harness around an agent to support sustained progress on long-running application development.

Supplementary study · 13

Developers, infrastructure funds, private credit and long-term contracts underpin data centers. Compute users, asset owners and risk bearers need not be the same parties.

Core reading · 10

Data dominance, decision delegation and standard-setting reveal how algorithmic systems reshape relations among states, technology firms and other organizations.

Core reading · 05

Delegation transfers more than tasks: authority, accountability, boundaries and verification criteria. Define how outcomes will be checked before decomposing the work.

Core reading · 08

If AI erodes labor-income and consumption tax bases, how should public finance respond? Compare taxation with public participation in asset returns.

A practical guide to tasks, traces and graders—and why a single evaluation method cannot tell the whole story.

Core reading · 04

Why do hierarchies exist? Knowledge, help-seeking costs and expert time explain how autonomous and assistive AI affect organizations and distribution.

Core reading · 03

As agents lower search, communication and contracting costs, where do the boundaries between firms, markets and platforms move?