AI is changing both **how we build software and who we're building it for**. For October's **Practically AI**, we have two talks exploring those changes from very different perspectives. First, we'll follow the evolution of AI-assisted software engineering from prompts and context through to autonomous loops and graphs and the engineering practices required to make them reliable. Then we'll look at a different kind of engineering challenge: the hidden **Inclusion Debt** we create when the assumptions built into our software and AI systems don't hold true for everyone. Two talks, one common theme: as AI becomes a bigger part of software engineering, **good engineering matters more, not less**. \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\- **Phil Whittaker** \- Prompt \> Context \> Harness \> Loop \> Graph: The Evolution of Agentic Software Engineering Loop and graph engineering are the phrases of the moment, but getting started can feel like learning to run before you can walk. Phil will work through the stages engineering teams experience as they move from simple prompting to managing context, building effective agent harnesses and eventually creating reliable autonomous development loops and graphs. Rather than jumping straight to autonomy, we'll look at what each stage adds, the problems it solves and the new engineering challenges it introduces. Along the way, we'll examine something equally important: the **traditional software engineering practices that make all of this actually work**. You'll learn: * How AI-assisted development evolves from prompting to context engineering * Why the agent harness is becoming an important part of the engineering stack * How autonomous development loops work and what makes them reliable * Where graph-based workflows fit as systems become more complex * Why testing, architecture, observability, feedback and engineering discipline become more important as autonomy increases The tools are changing. Feedback loops are

Description supplied by the event organiser.