Fragments: July 6

According to Martin Fowler, Thoughtworks recently held a Future of Software Development Retreat in Europe—the second of these gatherings. The mood has transformed since the first retreat, held in Utah just months before. In Utah, developers debated whether agentic engineering might work. In Europe, they came to report: we're shipping it. As one attendee noted, the whole debate about whether this changes software engineering is over. People aren't arguing about whether anymore. They're arguing about how. Several patterns emerged. Harness engineering—practices for guiding AI agents—became central to discussions, though the field remains young. Teams are now measuring code quality by token efficiency: how many tokens does the agent need to understand and modify a system? Traditional architecture and design principles still matter. Some worry AI has become powerful enough to handle messy code. Others argue the opposite: clear, modular design helps agents understand code just as it helps humans. Agent thinking and human thinking, they suggest, follow similar principles. One thread tied it together: build code that's clear to both people and machines. The profession is learning to design for an era where agents are partners in development.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton