Builders over safety: agentic hiring outpaces; Microsoft opens TauGrid

Risk alarms are getting louder while hiring leans harder into building. Posts tout agentic AI and calls to pace the frontier, yet the jobs picture tells a different story: engineering demand vastly outstrips evaluation and safety, and enterprise adopters, not nonprofits, lead the hiring. In our own data, that skew is stark; our tracker shows 1,681 modeling roles vs 44 in evaluation and safety. Across the 192 employers we track, 4,294 AI roles were open as of 2026-09-18, down 497 on the week. Our model puts new AI jobs created month-to-date at roughly 95,028. Taken together, the market is still adding AI capacity while tilting dollars toward shipping systems rather than staffing evaluators. Microsoft’s open-sourcing of TauGrid points to where the skills curve is bending: simpler AI operations on Kubernetes and rising expectations for AI orchestration. If agent swarms and large-scale deployments are the direction of travel, infrastructure jobs will shift toward teams that can run and coordinate these workflows. For people whose work is affected, the signal is consistent across stories and numbers: the strongest demand is for builders who can put agentic systems into production, while safety and evaluation hiring remains thin and led by enterprises rather than nonprofits.