RadarRADAR

Agents are starting to do a lot more than write code. They can delegate work to other agents, coordinate tasks and increasingly interact with software through the same interfaces we use. This week, What Is Agent Orchestration? looks at how agents can work together, while Build Your First Multi-Agent Workflow puts it into practice. We’re also looking at Agents That Use Software Like You Do (agents that can navigate, click, type and work directly with applications).

There’s plenty happening outside multi-agent systems too. Running capable models on your own hardware is becoming increasingly practical, so Run AI Locally explores what you can actually do locally, where it makes sense and what you give up compared with the cloud.

And as agents connect to more tools, there’s an interesting question around how they should access them. In MCP vs CLI, we’re testing whether agents really need purpose-built MCP integrations when they can already use the command line. Finally, if you want something you can put to work straight away, the Product Engineer role includes Getting Up To Speed on Any Codebase, a practical workflow for using AI to understand an unfamiliar project faster.