Technology Radar Vol 35

Thoughtworks Technology Radar Vol. 35, published in November 2026, is a twice-yearly guide to tools, techniques, platforms, languages and frameworks, based on the experience of its global teams. It classifies each item as Adopt, Trial, Assess or Caution, indicating how strongly teams should consider using it. This volume identifies four broad shifts in technology practice, especially as AI changes software development and enterprise infrastructure.

Harness engineering is maturing into a discipline: teams must build and maintain the context, memory, feedback loops and safeguards that make AI agents effective, potentially through enterprise-grade harness platforms. Model choice is also widening beyond proprietary frontier systems. Open-weight and local models, specialized models and model routing offer alternatives that can reduce costs, protect data and limit vendor dependence. Developers are increasingly running multiple coding-agent sessions or delegating longer tasks, but review capacity and delivery pipelines can become bottlenecks; the guidance is to expand agent use in step with oversight and production capacity. Finally, AI economics depends on infrastructure as well as model prices. Kubernetes-based “AI factories” coordinate compute, model serving, inference, storage and networking as programmable resources, making efficient platform design central to scaling AI sustainably.